University of Southern California Dissertations and Theses

Budget institutions and the positive theory of fiscal policy

(USC Thesis Other)

Budget institutions and the positive theory of fiscal policy

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

BUDGET INSTITUTIONS
AND THE
POSITIVE THEORY OF FISCAL POLICY

by

Michael David Pérez

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(POLITICS AND INTERNATIONAL RELATIONS)

December 2013

Copyright 2013 Michael David Pérez
ii
ACKNOWLEDGMENTS

This dissertation would not have been possible were it not for the continual
assistance and support of numerous people. First, and foremost, I would like to thank my
family. Without the support and encouragement of my parents, David and Kathleen, and
my brother Joshua, I would never have thought it possible to attend college and graduate
school. Throughout this process I was always comforted by the fact that I could always
count on them to be there for me. I would also like to thank my father- and mother- in-
law Paul and Pat and sister-in-law Kristine, for their backing and encouragement these
last few years. I am lucky to have a second family that stood behind me the entire way.
Thanks are also due to my dogs Einstein and Aesop for providing company during the
latter stages of the writing process and much need respites from work.
I am especially grateful for my advisor, Dr. Carol Wise, who convinced me to
attend USC and has been one of my biggest supporters. This dissertation would never
have been completed were in not for her patience, understanding, and assistance
throughout the entire process. More than an advisor she has also been a friend who
helped guide me through graduate school’s many hurdles. I also want to thank Dr. Patrick
James for being a close friend and advisor. Dr. James provided invaluable advice on
methods and presentation and my dissertation is better on account of our interactions. I
would also like to thank Drs. Manuel Pastor, Gerardo Munck, and Nick Weller for their
advice on this dissertation and for their encouragement at various stages of this project.
To my friends Ryan Duff, Alissa Duff, and Leif Iverson, I appreciate the support
and encouragement you have shown me over the years. You have always been my closest
and most dependable support group. Whether I was away at school in Berkeley, New
iii
York, Los Angeles, or Albuquerque, you were always only a phone call away and your
visits helped keep me grounded and focused on the light at the end of the tunnel.
To the many friends I made while at USC: because of you, my tenure in Los
Angeles was made all the more pleasurable. To Justin Berry, Amy Below, Mariano
Bertucci, Fabian Borges, Dave Bridge, Jen Buzza, Juve Cortes, Denise Gonzales, Eva
Weerts Hevron, Parker Hevron, Josh Jacobson, Rebecca Kimitch, Deniz Kuru, Jessica
Liao, Jesse Mills, Gloria Nava, Mark Paradis, Simon Radford, Jamie Simcox, Phil
Wilcox, and Nick de Zamaroczy, thank you. We have many memories together and you
are all my friends for life. Know that you can count of me to be there for you as you have
been for me.
Most importantly, I want to thank my wife Jillian for her unyielding patience,
understanding, and support. More than anyone else she has understood this process and I
could not have finished without her. She was not only my primary source of advice and a
pair of ears to listen to my ideas but she believed in me and my abilities. This dissertation
is just as much her accomplishment as it is mine.
iv
TABLE OF CONTENTS

Acknowledgements……………………………………………………………………..ii

List of Tables…………………………………………………………………………..vii

List of Figures…………………………………………………………………………..ix

Abstract………………………………………………………………………………….x

Chapter 1:
Introduction and Overview………………………………………………………………1
1 Introduction....….…………………………………………………………...............1
2 Overview……………………………………………………………………………7

Chapter 2:
Understanding Fiscal Policy……………………………………………………………13
1 Introduction...……………………………………………………………………...13
2 Recent Trends in Fiscal Policy Outcomes………………………………………...15
3 The Importance of Fiscal Policy in Economic Theory……………………………19
4 Politics and the Institutional Determinants of Fiscal Policy………….…………...25
4.1 The Bureaucracy……………………………………………………………26
4.2 The Branches of Government………………………………………………28
4.3 Political Parties……………………………………………………………..30
4.4 Elections and Electoral Systems……………………………………………32
4.5 Levels of Government……………………………………………………...34
4.6 Budget Institutions…………………………………………………………35
5 Critique……………………………………………………………………………36
6 Conclusion………………………………………………………………………...40

Chapter 3:
A Theory of Budget Institutions and Fiscal Policy…………………………………….43
1 Introduction……………………………………………………………………….43
2 Budget Institutions and Fiscal Policy……………………………………………..46
2.1 Fiscal Rules....................................................................................................46
2.2 Fiscal Transparency………………………………………………………...55
2.3 Procedural Rules……………………………………………………………63
3 Hypotheses………………………………………………………………………...71
3.1 Fiscal Rules…………………………………………………………………73
3.2 Fiscal Transparency………………………………………………………...74
3.3 Procedural Rules……………………………………………………………76
4 Conclusion………………………………………………………………………....78
v
Chapter 4:
Research Design…………………………………………………………………………80
1 Introduction………………………………………………………………………...80
2 Data and Measures………………………………………………………………….82
2.1 Independent Variables……………………………………………………… 82
2.1.1 Budget Institutions………………………………………………...82
2.1.1.1 Index Construction………………………………………91
2.1.2 Control Variables…………………………………………………..97
2.1.2.1 Economic Control Variables……………………………..97
2.1.2.2 Political Institutions Control Variables…………………100
2.2 Dependent Variables………………………………………………………..102
2.2.1 Government Expenditures, Revenue, and Size of Government….103
2.2.2 Government Deficits and Debt…………………………………...105
2.2.3 Composition of Government Expenditures………………………107
3 Case Selection……………………………………………………………………..109
4 Missing Data: Issues and Solutions……………………………………...………..120
4.1 The Causes of Missing Data………………………………………………..121
4.2 The Consequences of Missing Data.………………………………………..123
4.3 Multiple Imputation………………………………………………………...124
5 Method: Econometrics…………………………………………………………….130
6 Conclusion………………………………………………………………………...135

Chapter 5:
Size of Government and the Composition of the Budget………………………………137
1 Introduction……………………………………………………………………….137
2 Preliminaries………………………………………………………………………139
2.1 Missing Data and the Imputation Model…………………………………...139
2.2 Budget Institutions Indexes: Spearman Rank Correlations………………...143
3 Size of Government…………………………………………………………….....145
4 Budget Composition………………………………………………………………157
5 Robustness Tests…………………………………………………………………..168
5.1 Alternative Indicators………………………………………………………170
5.2 Alternative Democratic Samples...................................................................170
5.3 Level of Development……………………………………………………...173
6 Conclusion……………………………………………………………...................174

Chapter 6:
Budget Deficits and Government Debt………………………………………………...177
1 Introduction……………………………………………………………………….177
2 Budget Deficits……………………………………………………………………178
3 Government Debt…………………………………………………………………189
4 Robustness Tests…………………………………………………………………..200
4.1 Alternative Democratic Samples…………………………………………...200
4.2 Level of Development……………………………………………………...207
5 Conclusion………………………………………………………………………...209
vi
Chapter 7:
Conclusion……………………………………………………………………………...211
1 Introduction……………………………………………………………………….211
2 Budget Institutions and Fiscal Policy Outcomes: Theory and Empirical Results...211
2.1 Fiscal Rules…………………………………………………………………212
2.2 Fiscal Transparency………………………………………………………...215
2.3 Procedural Rules……………………………………………………………217
2.4 Preliminary Policy Prescriptions……………………………………………219
3 Limitations and Future Research………………………………………………….220

Bibliography....................................................................................................................229

Appendix A: Data Appendix…..……………………………………………………….253
vii
LIST OF TABLES

Table 3.1 Budget Institutions and Fiscal Policy: Theoretical Predictions………………77

Table 4.1 Descriptive Statistics of Dependent Variables………………………………109

Table 4.2 Correlations Between Democracy Indices…………………………………..117

Table 4.3 Country Scores on Four Indices of Democracy……………………………..118

Table 5.1 The Extent of Missing Values among Observations………………………...140

Table 5.2 Variables with Missing Observations………………………………………..141

Table 5.3 Spearman Rank Correlations………………………………………………...144

Table 5.4 Size of Government: Economic Models……………………………………..146

Table 5.5 Size of Government: Political Institutions Models…………………………..148

Table 5.6 Size of Government: Budget Institutions Models……………………………153

Table 5.7 Size of Government: Nested Models………………………………………...154

Table 5.8 Composition of Expenditures: Economic Models…………………………...159

Table 5.9 Composition of Expenditures: Political Institutions Models………………...161

Table 5.10 Composition of Expenditures: Budget Institutions Models………………...164

Table 5.11 Composition of Expenditures: Nested Models……………………………..166

Table 5.12a Robustness Tests…………………………………………………………..169

Table 5.12b Robustness Tests (cont.)…………………………………………………..172

Table 6.1 Budget Deficits: Economic Models………………………………………….179

Table 6.2 Budget Deficits: Political Institutions Models……………………………….182

Table 6.3 Budget Deficits: Budget Institutions Models………………………………...184

Table 6.4 Budget Deficits: Nested Models……………………………………………..186

Table 6.5 Government Debt: Economic Models.............................................................191

viii
Table 6.6 Government Debt: Political Institutions Models…………………………….193

Table 6.7 Government Debt: Budget Institutions Models……………………………...195

Table 6.8 Government Debt: Nested Models…………………………………………..197

Table 6.9a Robustness Tests……………………………………………………………202

Table 6.9b Robustness Tests (cont.)…………………………………………………....203

Table 6.9c Robustness Tests (cont.)…………………………………………………....204

Table 7.1 Budget Institutions and Fiscal Policy: Theory and Results…………………212

Table 7.2 Pathway Case Example: Size of Government……………………………….227
ix
LIST OF FIGURES

Figure 2.1 Central Government Expenditures (% of GDP) 1960-1998……………….....15

Figure 2.2 Central Government Expenditures on Social Services and Welfare
(% of GDP) 1972-1998…………………………………………………………..16

Figure 2.3 Central Government Budget Surplus (% of GDP) 1960-1998……………….18

x
ABSTRACT

There is much interest in the factors determining fiscal policy in order to explain
existing differences in policy outcomes between countries and within countries over time.
Recently, political institutions have emerged as the set of determinants thought to be most
influential in explaining these differences. However, extant studies are problematic since
they have not derived the full range of implications from theory, they lack sufficient
evidentiary support for their claims, and/or they do not pay particular attention to the
actors and institutions most influential in determining fiscal policy. This study examines
the effects of budget institutions on fiscal policy outcomes. These are the set of protocols
and practices used to draft, approve, and implement budgets. Since the budget explicates
the Government’s fiscal plans, budget institutions are the most proximal institutions
determining fiscal policy. This study uses survey responses to develop indices of three
budget institutions: fiscal rules, fiscal transparency, and procedural rules. Using Ordinary
Least Squares regression this study examines the effects of these institutions on the size
of government, the composition of expenditures, budget deficits, and government debt
among a sample of 83 countries. The results suggest that countries with stringent fiscal
rules, transparent budget procedures, and centralized procedural rules, are associated with
smaller governments, less transfers and subsidies spending, smaller budget deficits, and
less debt. While this is not the first study to examine budget institutions and fiscal policy,
it contributes to this research agenda by being the largest such study to date, in terms of
both country and regional coverage, and the most comprehensive in scope with regard to
the set of policy outcomes examined. Most importantly, its conclusions provide countries
with a set of initial policy prescriptions for improving fiscal policy performance.
1

CHAPTER 1: INTRODUCTION AND OVERVIEW

1. Introduction

The proper role of government in the economy is perhaps one of the most debated
topics among political economists, policymakers, and the public as a whole. In 1959
economist Richard Musgrave published the seminal text in public finance titled The
Theory of Public Finance in which he delineated three distinct (although not mutually
exclusive) economic functions for government. The first role was to improve the
allocation of resources. The proper role of government was simply to establish and
maintain the basic framework within which the market economy could freely operate by
providing for the defense of the nation, basic law and order, the protection of property
rights, and the provision of public goods. In this view, the government’s role in the
economy should be minimal and limited to this small set of essential functions. This was
the perspective popular among 19
th
century economists and which prevailed between
1870 and 1913. Over this 43-year period the level of public spending relative to national
income remained between 11% and 12% even though this was a modernizing period for
most of the now-industrialized countries (Tanzi and Schuknecht 1997).
However, events such as World War I, the Russian Revolution, The Great
Depression, World War II, and the rise of totalitarian regimes marked significant changes
in both the social and economic environment that would galvanize the remarkable
expansion of government activity in the economy up to the present day (Tanzi 2005). It
was during this period that the economic role of the state expanded from attempts to
influence allocative outcomes to the redistribution of income in order to achieve an
equitable wealth distribution and the stabilization of economic activity in order to attain
2
full employment of resources and price stability.
1
As a result of these changes the average
level of public spending as a percentage of GDP rose from 18.2% in 1920, 22.4% in 1937,
and 27.9% in 1960.
2
Between 1960 and 1990, the acceleration of expenditure growth was
even more dramatic. Over this period, public spending increased from an average of
27.9% of GDP to 44.9% of GDP. Expenditure in the decade between 1960 and 1970
alone grew by about the same percentage share of national income as the previous 90
years combined (Tanzi and Schuknecht 1997). Importantly, the growth in government
spending in the post-World War I period was the result of increased involvement of the
state in providing education, health care, and social security and welfare for its citizens.
Governments influence the allocation of resources, redistribute wealth, and
attempt to stabilize the economy through (among other things) the use of expenditures
and taxes. The process of determining the level and composition of expenditures and the
structure of taxation is known as fiscal policy. The study of fiscal policy is important
because governments differ in their decisions of how much of the national income to
spend, what programs to spend on, and whom and how much to tax in order to finance
such expenditures. These fiscal policy decisions merit study because of their impact on
economic indicators such as economic growth, employment, productivity, inflation, and
income distribution, as well as social indicators such as life expectancy, the incidence of
crime, literacy rates, the incidence of illness, and the quality of the physical environment
(Tanzi 2006).

1
Ruggie (1982) refers to the compromise of establishing an international free-trade regime so long as
countries were free to pursue welfare provision and regulate their domestic economies as “Embedded
Liberalism.”
2
These figures pertain to 17 industrialized countries sampled in Tanzi and Schuknecht (1997).
3
The study of fiscal policy typically focuses on either the effects of fiscal policy on
economic and/or social outcomes, or the analytically prior question of what factors
determine fiscal policy. The focus of this dissertation is on the latter, which is not as well
understood (Edwards and Tabellini 1991).
3
Specifically, I am interested in fiscal policy
outcomes, by which I mean the fiscal decisions that result from the policy process. These
are the decisions made by governments pertaining to how much to spend, what to spend
on, how much to tax relative to the level of spending, and how much borrowing will be
required given the difference between planned revenue and earmarked expenditure.
A substantial amount of scholarship has examined this topic in recent years.
However, I will argue that a non-trivial amount of recent work has not focused on the
most consequential factors determining fiscal policy outcomes. For example, scholars
have emphasized political polarization and instability, administrative turnover, and
incidence of coalition governments as factors that determine fiscal policy outcomes.
However, these factors are artifacts of specific political institutions and/or are
epiphenomenal to the political system as a whole. Conversely, studies that emphasize
such factors as electoral systems, forms of government, political parties, and/or the
various branches of government fail to underscore the precise institutional context within
which fiscal decisions are actually made. Instead, I argue that in order to understand what
determines fiscal policy outcomes one must look to budget institutions. The “budget” is a
planning tool for public policies and “budget institutions” are the set of regulations,
protocols, and practices used to draft, approve, and implement budgets (Filc and

3
While in Chapter 2 I briefly discuss some of the economic effects of fiscal policy, I do not give the topic a
thorough treatment, leaving it to other authors and future research. The interested reader should consult the
studies of Tanzi and Schuknecht (1997, 2000), Afonso et al. (2005), Tanzi (2005) and the sources cited
therein.
4
Scartascini 2006: 157, 158). As such they (should) determine directly the level and
composition of expenditure, the fiscal deficit, and the level of public borrowing (debt).
Thus, the question motivating this study is, “what are the effects of budget institutions on
fiscal policy outcomes?”
If the normative theory of fiscal policy accurately described how fiscal policy is
determined, then there would be no need for budget institutions. The normative theory of
fiscal policy posits that policymakers’ single objective is to maximize the “social
welfare” of citizens. These policymakers are cognizant of the fact that policy instruments
such as the structure of taxation, categories of expenditures, and (indirectly) fiscal deficits
can influence the economic and social indicators that comprise overall social welfare
(Cordes 1997; Tanzi 2006).
4
Thus, policymakers attempt to optimize the level of social
welfare by manipulating the available fiscal policy instruments.
In reality, fiscal policy is not formed by a single social planner that internalizes all
the interests and objectives of citizens before churning out a budget. The truth is that the
budget process involves the interaction of numerous parties, each with its own interests
and motivations that are often at odds with one another. For example, during the
negotiation stage of the budget the finance minister has an interest in minimizing the size
of the deficit but must negotiate with individual ministries that would like to maximize
the number of appropriations to their respective departments. Similarly, the executive
probably prefers a budget with benefits that are widely dispersed. However, in order for
the budget to be approved it must negotiate with the legislature whose individual
members likely prefer targeted expenditures that benefit their constituencies. These
interactions influence the size and allotment of state resources. Nevertheless, these

4
I discus in detail how these processes operate in the chapter that follows.
5
negotiations are shaped by the governing institutional framework (Filc and Scartascini
2006). Thus, the budget process is inherently political since by definition it determines
“who gets what, when, and how” (Lasswell 1936). Consequently, the budget that results
from this negotiation process more often than not considerably differs from that likely to
have been produced by a social planner. For instance, budget decisions can generate
wasteful and ineffective expenditure decisions, unsustainable deficits, or expenditures
that follow the election cycle (Filc and Scartascini 2006). Budget institutions may
mitigate these issues by imposing limits on expenditures, deficit levels, and/or the ability
of the government to borrow, or by granting substantial agenda-setting power to actors
(e.g. the finance minister) with an incentive to maintain fiscal discipline. Budget
institutions are therefore important because they prescribe the “rules of game.” Budget
institutions perform this function by imposing restrictions on the entire budgetary process,
by distributing power and assigning roles and responsibilities among policymakers, and
by disseminating information to the various actors and to the public (Filc and Scartascini
2006). Accordingly, budget institutions determine fiscal policy outcomes by delimiting
the use of fiscal policy instruments and granting prerogative to the policy preferences of
certain groups of actors over others.
The study of the effects of budget institutions on fiscal policy is not new. For
example, Von Hagen and Harden (1995) and Hallerberg et al. (2007) studied the effects
of budget institutions on fiscal policy outcomes in the European Union (EU); Alesina et
al. (1999), Filc and Scartascini (2004), and Hallerberg and Marier (2004) examine these
effects within Latin America; Perotti and Kontopoulos (2002) focus on the OECD;
Gleich (2003) examines these institutions among the East and Central European
6
countries; and Gollwitzer (2011) studies Africa. Thus, there is a rich empirical tradition
from which to draw. My contribution to this research agenda is that I examine the effects
of budget institutions on a sample of 83 countries from every region of the world. This
constitutes the largest such study to date of which I am aware. Second, I examine the
effects of budget institutions on a larger set of fiscal policy outcomes than is commonly
done. Most studies focus on budget deficits and public debt. The more ambitious studies
may include the size of government (i.e. level of expenditures). I go beyond these fiscal
outcomes and include for the first time aspects of the composition of the budget as well.
Specifically, I investigate how budget institutions influence how much countries spend
toward redistributing income via transfers and subsidies expenditure, and how much
governments spend toward allocation by means of public good provision.
The emphasis on political institutions as possible explanations for fiscal policy
places this study within the recent move toward the “positive theory of fiscal policy.”
5

This research agenda accepts the public choice tenant that policymakers pursue their own
interests. However, the primary concern is for the institutions and arrangements that
determine policy outcomes.
6
The belief among scholars working within tradition is that if
certain institutional arrangements are in place “good policies could be pursued and better
objectives could be achieved” (Tanzi 2006: 6). Therefore, the goal of the positive theory
of fiscal policy is to isolate the institutional improvements that would make fiscal policy
more effectual. The objective of this study, then, is to demonstrate that (1) the observed
variation in fiscal policy outcomes is (partially) the result of differences in budget

5
Examples of recent work in this vein include Alesina and Rosenthal (1995), Drazen (2000), and Persson
and Tabellini (2000).
6
Public Choice is the application of the analytical tools of economics to the theory and practice of politics.
As such it argues that self-interest motivates economic as well as political decisions (Butler 2012).
7
institutions; and (2) consequently, fiscal discipline is possible provided that the correct
budget institutions are in place.

2. Overview

I conclude this introduction with a brief outline of the chapters to follow. Chapter
2 provides a starting point by familiarizing the reader with the general importance of
fiscal policy and why fiscal outcomes merit continual study. It begins by demonstrating
recent fiscal policy trends and briefly reviews the theory developed by economists to
explain the effects of fiscal policy on essential macroeconomic outcomes. It is on the
basis of this discussion that I argue importance of studying the determinants of fiscal
policy. Once the significance of fiscal policy is understood, I introduce popular
explanations for why fiscal policy varies between countries and within countries over
time. Special attention is paid to recent explanations that emphasize political and
institutional factors. The chapter ends with a critical appraisal of the current literature
based on the ability to accurately and satisfactorily explain spatial and temporal variation
in fiscal policy. I argue that the current literature is deficient in four important aspects: (1)
its theoretical leverage; (2) its explanatory power; (3) its failure to focus on the proper set
of actors; and (4) its neglect of the most consequential institutions that determine fiscal
policy. This critique foreshadows the argument that emphasizing budget institutions
provides a solution to these shortcomings and sets the agenda for the theoretical work to
follow.
Two inherent features of public finance are the agency and common-pool
resource problems. The agency problem arises because policymakers in charge of
expenditure and revenue decisions may have policy preferences that are at odds with the
8
interests of the voters. The common-pool resource problem results from policymakers
failing to fully internalize the tax implications of their fiscal decisions. It is widely
recognized that these features create the incentive for governments to excessively spend
and result in large deficit and debt liabilities. In Chapter 3 I introduce and defend the
assumption that voters are fiscally conservative and therefore, on average, they prefer less
pending, less taxes, and lower debt and deficit levels. Based on this assumption I then
develop an institutional theory describing budget institutions as solutions to the agency
and common pool resource problems and explain how these institutions determine fiscal
policy outcomes.
7

I delineate three budget institutions in particular: fiscal rules, fiscal (or budget)
transparency, and procedural rules. Fiscal rules are numerical targets that place limits on
budgetary aggregates such as expenditure or debt levels, or the size of debt. Such rules
make fiscal policy both credible and predictable and diminish the agency and common-
pool resource problems by limiting the extent to which policymakers are able to exploit
fiscal policy instruments. I hypothesize that where such rules are in place, expenditure
levels, budget deficits, and debt will be lower compared to where they are not.
Fiscal transparency refers to the level of openness regarding the government’s
fiscal policy intentions and the state of public sector accounts. Since policymakers
possess informational advantages over the public with regard to the health of public
finances they are in a position to misuse policy instruments for personal and/or political
gain. Fiscal transparency mitigates the agency and common-pool resource problems by
revealing the state of public finances and opening the policy actions of politicians to

7
Diermeier and Krehbiel (2003: 127) define an “institutional theory” as “…a theory that seeks an
understanding of the relationships between institutions, behavior and outcomes.”
9
scrutiny. Consequently, I predict that the greater the level of fiscal transparency in the
budget process the better the budget balance, the lower the level of expenditures, and the
smaller the size of public debt.
Procedural rules dictate the preparation, approval, and execution of the budget.
Electoral incentives mean that spending ministers and the typical legislator are more
responsive to the groups that benefit from specific spending programs whereas the
executive is relatively more responsive to the interest of the average taxpayer. Where the
interests of the individual spending ministers and the legislature prevail in determining
the budget structure the common pool problem is likely to be exacerbated. Therefore, the
problem can be reduced by granting budgetary power to the actors (e.g. the treasury
minister) that have an incentive to internalize the entire costs and benefits of policy. This
is because they set total expenditure and draft the revenue budget and are therefore fully
cognizant of the government budget constraint. I label these types of procedural rules
hierarchical and argue that such procedures will result is greater fiscal discipline in the
form of lower expenditure levels, and smaller budget deficits and debt. Furthermore, I
expect that since the spending priorities of the executive prevail under hierarchical
procedural rules that more will be spent on public goods whose benefits are widely
dispersed and less will be spent on transfers and subsidies, the benefits of which can be
targeted to special interests and specific socioeconomic and/or demographic groups.
Chapter 4 describes the study’s research design. I begin by discussing the
empirical measures and sources of the budget institutions variables, fiscal policy
variables, as well as the set of controls. Afterward, I describe case selection. The theory
presented in Chapter 3 rests on the assumption that fiscal policy is made within the
10
confines of democratic institutions. Thus, cases are selected based on their “level” or
“quality” of democracy. Previous work has shown that empirical results are often
sensitive to the particular definition and/or attributes of democracy utilized by different
measures. To account for this possibility the empirical tests in subsequent chapters are
each conducted using four different measures of democracy. Sample sizes thus range
from 66 to 83.
8
Missing data is an issue in this study and threatens to adversely affect the
validity of the results. I include a subsection dedicated to familiarizing the reader with the
causes and consequences of missing data for valid statistical inference. I then describe the
multiple imputation solution to the missing data problem and discuss how it affects the
subsequent interpretation of the empirical results. The chapter concludes with a
presentation of the econometric methods utilized to test the previously derived
hypotheses.
Chapter 5 presents the results of the first set of empirical tests. I apply the
econometric methods from the previous chapter and estimate the effects of budget
institutions on the size of government and budget composition. I find that procedural
rules, and to a lesser extent budget transparency, lower the size of government as
expected. The results for fiscal rules are inconclusive. In terms of budget composition,
procedural rules are demonstrated to lower the level of spending on transfers and
subsidies but do not appear to increase expenditures toward public good provision.
Surprisingly, I discover that fiscal rules reduce transfers and subsidies spending and
transparency lower public good spending, although the theory developed in Chapter 3 is
unable to explain these findings. Overall, these effects are not only statistically but

8
In actuality the dataset contains 97 total countries. However, 15 cases must be excluded because they do
not meet the minimum requirements of democracy on any of the four measures.
11
economically meaningful. Moreover, robustness tests reveal that effects are generally
insensitive to specification.
I continue with the empirical tests in Chapter 6 which explores the effects of
budget institutions on budget deficits and government debt. I find that fiscal rules
improve the budget balance as expected. I also find that the magnitude of this effect
increases as the level of transparency becomes sufficiently high. I do not find any effects
on budget deficits of transparency and procedural rules. Both fiscal rules and procedural
rules are shown to induce smaller debt levels. There is some evidence that budget
transparency is also related to lower levels of government debt, although the evidence is
not as compelling. Nevertheless, while most of the results are economically significant
and robust to specification, budget institutions are unable explain a substantial amount of
the variance in budget deficits and government debt.
Finally, Chapter 7 evaluates the empirical findings of this study. I begin by
reviewing which results were consistent with theory and which were not supported by
empirical evidence. On the basis of this assessment I discuss what, if any, lessons may be
offered to policymakers interested in ways to improve their country’s fiscal policies.
Given the limitations of this study and since some unexpected patterns were uncovered in
the data I close the dissertation by suggesting further extensions. In particular, I argue
that future empirical work should concentrate on improving the quality of data and
standardizing the measures of budget institutions. The study of budget institutions and
fiscal policy may also be advanced using alternative research methods. Lastly, I argue
that future theoretical work should focus on explaining the ways in which budget
institutions, especially fiscal rules and budget transparency, impact the composition of
12
spending since these results could not be explained by the theory developed in this study.
All in all the positive theory of fiscal policy is a very active area of research and this
study hopes to demonstrate the central role of budget institutions within this research
agenda.
13
CHAPTER 2: UNDERSTANDING FISCAL POLICY
1. Introduction
Since the Great Depression of the 1930s governments and their constituents
around the world have come to expect, and in varying degrees to accept, an active role for
government in the economy. While the popularity and scope for such actions tends to ebb
and flow over time, the fact is that levels of government expenditures and taxation have
steadily increased for nearly a century, as has the share of national income devoted to
transfers and programs of social protection (Samuelson and Nordhaus 1989: 409). In
addition, the last three decades have witnessed an appreciable increase in public deficits
and debt accumulation across the world (Alesina et al. 1999: 253). Nevertheless there is
substantial variation among fiscal policies between and within countries. For example,
average expenditures can differ between countries in the range of approximately 30% to
45%. Likewise, spending on social services and welfare can differ on average by as much
as 20% and deficits by approximately 8%. This variation is noteworthy because the
nature of taxes and amount that a government spends has consequences for the
macroeconomy, most importantly on productivity and national income. Thus, this
variation and the factors determining fiscal policy require explanation.
The purpose of this chapter is fourfold. The first objective is to present
indisputable evidence of the variation in fiscal policy, specifically government (total and
functional) expenditures and deficits in recent history. These trends are summarized and
discussed in Figures 1-3 below using data developed by Persson and Tabellini (2003b).
The data clearly show significant variation both between and within countries over time.
It is this growth and variation in fiscal policy outcomes that requires explanation.
14
Before turning our attention to existing theories of fiscal policy, however, our
second objective is to briefly review economic theories on the effects of fiscal policy.
This is done in section three. Such a review is necessary because without a solid
understanding of why fiscal policy is important for such macroeconomic outcomes as
productivity, employment, inflation, interest rates, and the trade balance, just to name a
few, explaining the variation in policies across and within countries would be of no great
concern.
The third goal is to introduce existing and prominent explanations for the
determinants of fiscal policy. In particular, the focus is on political, and especially
institutional, explanations, which are argued to hold the most promise. This is due to the
fact that economic theories of fiscal policy have not held up well against the empirical
evidence and non-institutional political explanations turn out to be simply the
mechanisms through which institutional factors have had their effect.
The forth and final objective is to critically appraise the state of current literature,
particularly with regard to its ability to accurately and satisfactorily explain variation in
fiscal policy. The assessment reveals that the current literature is deficient in four
important aspects. Specifically, the literature (1) falls short in its ability to completely
theorize the effects of political institutions on an exhaustive list of fiscal policy outcomes;
(2) lacks explanatory power; (3) fails to focus one of the most consequential institutions
responsible for determining fiscal policy, which is the budget; and, (4) it does not pay
adequate attention to the proper set of actors and their roles in determining fiscal policy.
In light of these shortcomings a brief conclusion foreshadows the chapter that follows
which develops a theory that attempts to remedy these inadequacies.
15
2. Recent Trends in Fiscal Policy Outcomes
10 20 30 40
cgexp
1960 1970 1980 1990 2000
year
cgexp lb/ub

Figure 2.1: Central Government Expenditures (% of GDP) 1960-1998
Figure 1 shows central government expenditures as a percentage of gross
domestic product (GDP) between 1960 and 1998 for a sample of sixty countries using
data collected by Persson and Tabellini (2003b). The curve represents the average
expenditure levels for each year over the given period and the vertical bars denote the
inter-quartile range (the highest and lowest 25% of the data). From the graph it is clear
that during the sample period there is both an upward trend in average expenditure levels
as well significant variation in government spending. In fact, outlays vary both between
countries and within countries over time. For example, the average expenditure level for
the complete sample for the 1960s is 17.8% of GDP compared to 28.6% of GDP during
16
the 1990s.
9
Over the entire sample period expenditures were highest on average in
Belgium (48.1% of GDP) and lowest in Guatemala (10.5% of GDP). In the 1960s
expenditures between countries varied from a low of 6.5% of GDP to a high of 38.7% of
GDP (with a standard deviation of 7.5%). Within that same period expenditures within a
given country varied between 11.1% of GDP to 24.2% of GDP (with a standard deviation
of 1.7%).
10
For the 1990s, the comparison figures for between country variation are a
minimum of 9.4% of GDP and a maximum expenditure level of 51.2% of GDP (with a
standard deviation of 10.9%) and within variation of 20.1% and 38.3%, respectively
(with a standard deviation of 2.2%).
11
In sum, the general trend has been for the
proportion of national income spent by the central government to increase over time.
Nevertheless, it is clear that there remains significant variation between countries.
0 5 1 0 1 5
s s w
1970 1980 1990 2000
year
ssw lb/ub

Figure 2.2: Central Government Expenditures on Social Services & Welfare (% of GDP) 1972-1998

9
Note that the curve conveys sample averages per year whereas here I am talking about averages over ten
year periods.
10
This is not to suggest that a specific country deviated from its average by approximate 24% of GDP or
11.1% of GDP. After accounting for the global average (17.8%), what these figures convey is that some
country deviated from its average by increasing expenditures 6.4% over the period while another decreased
its expenditures by 6.7%.
11
In other words, one country deviated from its average by increasing expenditures 9.7% over the period
while another decreased expenditures by 8.5%.
17
Figure 2 displays central government expenditures on social services and welfare
as a percentage of GDP for the period 1972 to 1998 (Persson and Tabellini 2003b).
Examination of transfer payment expenditures is particularly important because over the
last thirty years such expenditures have grown at a faster rate than government purchases
of goods and services. In addition, the constituencies created by increased spending on
transfer payments make the reduction of such expenditures difficult when and if fiscal
adjustment becomes necessary (see discussion below) (Alesina and Perotti 2008: 13).
There is a similar, albeit less dramatic, trend to total outlays. Social services and welfare
spending have an upward trajectory, increasing from a global average of 6.3% of GDP
during the 1970s to 8.1% of GDP by the 1990s. In this spending category there is
significant variation as well. On average, Belgium spent the most on social services and
welfare over the entire period, spending 20.4%, whereas Peru spent the least, at .04%.
Between countries, transfer payments increased from a minimum of .04% and maximum
of 19.1% during the 1970s to a minimum of .26% and a maximum of 23% (with
respective standard deviations of 5.6% and 6.7%) by the 1990s. Likewise, over the
sample period within country expenditures varied between a minimum and maximum,
respectively, of -.68% and 10.4% of GDP for the 1970s (with a standard deviation of
1.2%) and a minimum of 2.1% of GDP and a maximum of 12.1% of GDP (with a
standard deviation of .97%) between 1990 and 1998. Thus, perhaps to be expected,
expenditures on social services and welfare have increased over the last forty years and
display a similar variation in overall central government outlays.

18
-8 -6 -4 -2 0
spl
1960 1970 1980 1990 2000
year
spl lb/ub

Figure 2.3: Central Government Budget Surplus (% of GDP) 1960-1998

Figure 3 displays central government budget surplus as a percentage of GDP over
the nearly four decades between 1960 and 1998 (Persson and Tabellini 2003b). Average
deficits (negative surpluses) grew from a low of 1.6% of GDP during the 1960s to a high
average of 3.9% of GDP during the 1980s, with countries showing signs of fiscal
consolidation during the 1990s as the global average deficit reduced to 2.6% of GDP.
Nevertheless the data show a significant amount of variation. Overall, for the entire 1960-
1998 period Sri Lanka has, on average, the highest deficits (approximately 8.1% of GDP)
while Botswana has the largest average surplus (about 4.4% of GDP). Nevertheless with
the financial crisis that began in 2007 the financial standing of many countries has
worsened. Perhaps the most prominent example is the current European sovereign debt
crisis where countries such as Greece, Ireland, and Spain are running deficits between
19
8.5% and 10% of GDP and with public debt levels approximating 100% to 160% of GDP
in countries such as Belgium, Greece, Ireland, Italy, and Portugal (CIA World Factbook
2012). Crises such as these are likely to only exacerbate the differences in fiscal policy
outcomes that already exist.
In sum, the descriptive statistics above clearly demonstrate that fiscal policy
outcomes differ both across countries and within countries over time. Before taking stock
of scholarly explanations for this variation, it is important to first be aware of the reasons
why fiscal policy is of interest in its own right. The next section presents a brief review of
the purpose of fiscal policy as well as a discussion of the economic theory of its effects.
After I develop an understanding as to why fiscal policy is of concern to economists,
politicians, and citizens alike, I then turn to the theories that have been put forward to
explain why countries vary in their fiscal policies and thus their overall economic
performance.

3. The Importance of Fiscal Policy in Economic Theory

In a macroeconomic sense, governments across the world have assumed the role
of mitigating the deleterious effects of the business cycle on the economy. Particularly,
governments seek to positively influence employment and output levels as well as the
rate of inflation, and correct “unfair” and/or unequal market outcomes through the
redistribution of income (Samuelson and Nordhaus 1989: 414; Hockley 1992: 30). One
mechanism with which governments attempt to do so is through the use of fiscal policy,
or, the use of taxes and expenditures to influence the economy.
12
Governments tax
economic activity not only to raise revenue to finance public operations but to affect

12
Another method is through the use of monetary policy.
20
aggregate demand. Likewise, governments make purchases in the hopes that such
expenditures will spur employment and increase productivity.
Textbook economic theory argues that governments should increase
expenditures and/or cut taxes when there is insufficient aggregate demand to maintain
full-employment levels in the economy. Falls in aggregate demand result in a reduction
of the number of goods and services produced causing firms to lay-off workers, thus
increasing unemployment and causing national income to fall (Mankiw 1998: 482).
Increased government expenditures should be made in order to put underused and unused
production factors to work in an effort to raise employment and productivity levels back
to their natural rates (Alt and Chrystal 1983: 60). In addition, government expenditures in
one area may increase consumption in other areas as well, leading to a “multiplier effect,”
raising national income by perhaps several dollars more than each original dollar that is
spent (Mankiw 1998: 439-440; Blinder and Solow 1972: 1). Should higher levels of
demand lead to increased levels of investment as well, particularly in capital goods,
expansionary fiscal policy could have additional positive effects on economic growth as
well (Easterly and Rebelo 1993).
Similar effects may take place when the government cuts taxes. Tax cuts increase
a household’s take-home pay. The money that is spent (as opposed to saved,) increases
consumer spending and uplifts aggregate demand. Once more, the increase in consumer
spending causes earnings and profits to rise, further stimulating economic activity and
triggering additional multiplier effects (Mankiw 1998: 442-443). While there is debate
over which is more expansionary, increased spending or cutting taxes, most economists
21
agree that there is nonetheless a potential role for fiscal policy in stabilizing demand
(Blinder 2004; Alesina and Ardagna 2009).
Theory and practice, however, point to a number of potentially negative effects of
discretionary fiscal policy. One possible effect of expansionary fiscal policy is that
increased public outlays “crowd out” private sector activity. For instance, government
expenditures may go toward producing goods and services that otherwise would be
produced and supplied by the private sector. In this sense, public spending simply
replaces private investment (Blinder and Solow 1972: 2-3; Hemming et al. 2002: 4).
Increased public spending may also raise interest rates thus crowding out private
investment. Essentially, expansionary fiscal policy, by increasing income, raises the
demand for money. Interest rates rise to balance supply and demand. The result is a
reduction in the quantity of goods and services demanded and because borrowing is more
expensive, the demand for investment declines (Mankiw 1998: 441-442). In much the
same way, tax cuts may crowd out private investment as well. Over time, reduced
investment can lead to reduced economic growth. Whether or not fiscal policy is
effectively expansionary depends on whether the spending and/or tax multiplier causes a
shift in aggregated demand that is larger than the initial expenditures and/or cuts.
13

Perhaps more worrisome, however, are the potentially adverse effects of
government budget deficits and debt that may result from expansionary fiscal policies.

13
A few caveats are in order. First, the effectiveness of expansionary fiscal policy in the form of multiplier
effects is only likely to be realized in an economy with idle resources (Alt and Chrystal 1983: 61). If the
economy is characterized by full employment, then the effect of fiscal expansion will be inflation as
increased demand raises prices. Second, the ability of a tax cut to influence consumer spending depends on
consumer expectations. If the cut is perceived as temporary the additional income contributes little to one’s
permanent income and thus affects consumer spending very little. If, on the other hand, the cut is perceived
as permanent, it will be viewed as increasing one’s lifetime financial resources and thus likely to have a
more positive effect on consumer spending (Mankiw 1998: 442-443; Kopcke et al. 2005: 5-6; Although,
see Blinder 2004:22-24 for a contrary argument and supporting evidence).
22
First, when a government runs a deficit it reduces national saving. The result is a decline
in the supply of funds available for lending, which drives up interest rates and once more
crowds out investment (Mankiw 1998: 390).
14
High interest rates increase the returns to
domestic saving and attract foreign investors. The combined effect is for net outgoing
foreign investment to fall (Mankiw 1998: 391-392). The fall in investment abroad means
an increase in demand for the domestic currency, which results in the appreciation of the
real exchange rate. Currency appreciation makes domestic goods more expensive relative
to foreign goods causing imports to rise and exports to fall (Mankiw 1998: 392; Gramlich
1989: 23). The cumulative effect is for the trade balance to move into a deficit as well.
15

There are also possible distributional consequences related to running persistent
deficits. Insofar as deficits imply increased borrowing or later tax hikes future
generations can expect a reduced standard of living due either to the diminished
productive capacity and/or the increased burden of servicing the outstanding debt
(Mankiw 1998: 489-490; Kopcke et al. 2005: 5-6).
16
However, unlike the immediate
effect of, say, high interest rates, borrowing and prospective tax increases primarily affect
future, rather than current, generations who are unable to affect policy until it is perhaps
too late.

14
It should also be noted that because running a deficit lowers investment, the long term effect is to lower
future productivity and growth (Gramlich 1995: 171).
15
Not all economists agree that running budget deficits will have negative repercussions. For example, the
so-called “Ricardian Equivalence” view (Barro 1974; 1979) argues that deficits have no real effect on the
economy because private actors anticipate the future tax costs of borrowing and increase their present rate
of saving, effectively neutralizing the deficit (Gramlich 1989: 32; Blinder 2004: 18-19). This is part of the
“New Classical” school of economics, which generally views stabilization policy as having no effect on the
real economy because systematic macroeconomic policies will come to be anticipated by private economic
actors (Alt and Chrystal 1983: 69).
16
However, if the fiscal expansion were a success, i.e., led to higher current and future incomes, it would
alleviate such costs.
23
Nevertheless, even if one supposes that the above effects can be avoided, one
must still take into account the political processes involved in proposing and executing
fiscal policy. Fiscal policy, irrespective of its intended and unintended effects, suffers
from what economists call “inside lags.”
17
These are delays between the time the relevant
actors recognize the need for fiscal stimulus or contraction and the implementation of the
appropriate policies (Blinder 2004: 7). As a consequence of inside lags, by the time a
change in fiscal policy is recognized and executed, the state of the economy may have
changed (Mankiw 1998: 446-448). A policy intervention intended to be countercyclical
may turn out to be pro-cyclical and exaggerate the business cycle (Alt and Chrystal 1983:
62; Feldstein 2009: 2). It merits recognition that some economists find political lags,
rather than the possible negative economic effects, to be the most persuasive argument
against discretionary fiscal policy (Blinder 2004: 8, 27-28).
18

In sum, it is possible for the instruments of fiscal policy to be effective in
smoothing the rough edges of the business cycle by preventing chronic unemployment
and overall economic stagnation. However, such stabilization policy, when imprudently
implemented, runs the risk of adversely affecting investment behavior, the exchange rate,
the trade balance, and long-term economic growth. The consequence is that future
generations are likely to be born into an economy with a significantly reduced standard of
living. The essence of the problem, then, is striking a balance between the conflicting

17
Fiscal policy also suffers from “outside lags,” or “the time that elapses between a fiscal policy shock and
its effects on the economy” (Blinder 2004: 7). However, inside lags are considered more of a problem for
fiscal policy.
18
To avoid inside lags many governments institute “automatic stabilizers,” which are fiscal policy changes
that take effect automatically as the economy goes into a recession or an expansion. Examples include the
tax system, which collects less taxes as economic activity declines during a recession and collects more
taxes as activity expands during a boom, and government spending in the form of social protections such as
unemployment and welfare benefits as more (less) people become and eligible as a result of the state of the
economy (Mankiw 1998: 448).
24
goals of short-run stabilization and long-run growth. After all, it is national saving and
capital formation that ultimately determines a country’s level of output and thus prospects
for growth (Mankiw 1998: 240-241; Kopcke et al. 2005: 7-8). Nevertheless, even if such
a compromise can be made, lags between the time a change in policy is recognized and
actually implemented may render the intervention ineffective.
In light of this analysis, the question becomes: “what determines whether a
government makes effective and sensible use of fiscal policy?” Before turning to the
current state of scholarship, which seeks to answer this question, it is important to first
note that the formulation and implementation of fiscal policy is a highly politicized
process. This is in contrast to monetary policy, which governments by and large have
delegated to technocrats operating within central banks and finance ministries.
19
Fiscal
policy, on the other hand, remains firmly in the grasp of politicians since electoral
prospects often times depend on party-related decisions on spending programs and the
adjustment of tax rates (Alesina and Giavazzi forthcoming: 2). Tax increases or spending
reductions, for example, are politically unpopular and politicians often believe that such
actions could lead to their being voted out of office (Gramlich 1995: 171). This is the
widely held belief, despite the fact that recent evidence suggests that fiscal adjustment
rarely affects the electoral prospects of politicians adversely.
20
Electoral incentives,
nevertheless, make it far more likely that politicians will trade the long-term benefits of
deficit reduction and economic growth for the short-term benefits of high expenditures

19
This decision is of course a political one as well. For an introduction to this issue see Goodman (1991)
and the contributions to the special issue of International Organization volume 56 issue 4 (2002).
20
Specifically, recent work by Alesina, Carloni, and Lecce (2011) examines whether governments that
reduce budget deficits are subsequently voted out of office and find no systematic relationship between
decisive fiscal adjustments and electoral defeat, even after controlling for the “strength” and popularity of
the administration.
25
and/or low taxes (Gramlich 1995: 178). Having established the importance and political
foundations of fiscal policy, I now shift attention to the numerous explanations that have
been put forward to explain the observed variation in fiscal policy outcomes.

4. Politics and the Institutional Determinants of Fiscal Policy

There are substantial differences in fiscal policies among countries and within
countries over time. As was shown, scholars have a fairly comprehensive understanding
of the effects of fiscal policy on key macroeconomic variables, but only recently have
they begun to address the factors that determine a country’s fiscal policy (Edwards and
Tabellini 1991: S16). Accepted economic theory however, has been unable to explain
why countries differ in their composition of revenues and expenditures, for example, or
why some countries run persistent deficits while others run surpluses (Mulas-Granados
2004: 22).
21
The inability of economic theory to explain such differences has provided
scholars with an opportunity to explore the political determinants of fiscal policy.
There exists a substantial body of scholarship on this topic, to which this
dissertation seeks to contribute. However, a non-trivial amount of this literature happens
to focus on factors that are artifacts of specific political institutions or factors
epiphenomenal to the political system as a whole. Alesina and Tabellini (1989, 1990) and
Tabellini and Alesina (1990) for instance, argue that political polarization and frequent
government turnover are closely related to higher debt levels. Alesina (1989) finds a
similar relationship between high levels of national debt and political instability. Others,
notably Roubini and Sachs (1989a,b), argue that the frequency and type of coalition

21
For surveys of economic explanations see Berry and Lowery (1987), Borcherding (1985), Cameron
(1978), Larkey et al. (1981), Lowery and Berry (1983), and Peltzman (1980), to name a few.
26
government (e.g. the frequency of minority government) affects debt levels and the
ability to respond quickly to fiscal shocks.
22
With regards to the latter, Poterba (1994) and
Alt and Lowry (1994) find that divided governments also postpone fiscal adjustments.
Nevertheless, factors such as polarization, change in administration, political instability,
frequency and type of coalition government, are endogenous to such institutions as
established electoral rules, the various forms of government, and political parties. Given
this priority, I shall focus on the ways in which political institutions impact fiscal policy.

4.1 The Bureaucracy
Since the classic work of Niskanen (1971) a number of scholars have attributed
the growth of government to the alleged inefficiencies of bureaucracies. The focus of
Niskanen’s study is the negotiations that take place between a public agency or bureau
and its legislative sponsor over the bureau’s budget.
23
The relationship is characterized as
a bilateral monopoly, where the bureau is the sole supplier of a particular good or service
and the legislature is the singular buyer (monopsonist). It is assumed that the bureau’s
objective is to maximize its budget. Given that it is the monopoly supplier of the good or
service, only the bureau is aware of the true costs of production. In addition, the fact that
the bureau is knowledgeable of how much the sponsor values the good or service allows
it to present the legislature with a take-it-or-leave it output schedule for a given budget
that the legislature is sure to accept. Based on these facts, Niskanen concludes that
bureaus will tend to overproduce and supply their good or service at levels above what is

22
See also Edin and Ohlsson (1991) and De Haan and Sturm (1994, 1997).
23
A “bureau” is an organization whose output is not sold in the marketplace, nor sold at a per-unit price
(Downs 1967: 25; Niskanen 1971: 25).
27
considered socially optimal. As a result, fiscal expenditures, and hence government, is
excessively large.
Critics of Niskanen’s (1971) and other early studies of public agencies argue that
the theoretical foundations do not accurately describe the bureau’s motives, nor does it
accurately depict the relationship between the public agency and its legislative sponsor.
Migue and Belanger (1974) and Wyckoff (1990), for instance, argue that bureaus seek to
maximize their “discretionary” or “slack” budgets, that is, the difference between total
budget and the costs of production, rather than budgets per se. However, empirically,
there is little difference between how each type of bureau behaves (Wyckoff 1990). In
contrast, Romer and Rosenthal (1978, 1979) demonstrate that the ability of the bureau to
exploit the sponsor and extract excess levels of rent depends on the “reversion level,” or
the policy instituted in the instance that the bureau’s offer is rejected (see also MacKay
and Weaver 1979). High expenditures occur only when the reversion level is particularly
low (Romer and Rosenthal 1978; 1979). Others take issue with the fact that in early
models of bureaucracy the legislature plays only a passive role in this process. In fact,
when one accounts for the legislature’s ability to audit (Banks 1989, see also Banks and
Weingast 1992), penalize the bureau for deception (Breton and Wintrobe 1975; Bendor et
al. 1985, 1987ab), and/or conceal its level of demand for the bureau’s good or service
(Miller and Moe 1983; Bendor et al. 1985, 1987ab), then the Niskanen result of over-
productive bureaus emerges as an extreme (high) case and in reality the size of
government can vary considerably (Moe 1997; Wintrobe 1997). Since the legislature is
responsible for imposing the structure that determines governmental supply, it is more
28
reasonable to attribute the problem of big government (if it exists) to it (Miller and Moe
1983: 320).

4.2 The Branches of Government
With regard to legislatures, theory and empirical evidence suggests a direct
relationship between the size of the legislature and the overall level of government
spending. When legislators earmark expenditures for projects that benefit their districts
from a general tax fund, the district bears only a fraction of the total costs. As the number
of districts increases, the per district tax burden decreases, thus creating an incentive for
legislators to raise the overall level of expenditures (Weingast 1979; Shepsle and
Weingast 1981; Weingast et al. 1981). This is known as the “common pool resource
(CPR) problem.” The underlying mechanism causing such growth is the preference of
each legislator for “universalism,” rather than majority rule. The idea is that under
majority rule a given legislator is uncertain as to whether she will be a member of one of
the many possible minimum winning coalitions that may form and set policy. However,
under the norm of universalism she and every other legislator are assured support for
their preferred projects. Thus, each legislator trades the expected benefits of majority rule
for the certain benefits of universalism. Consequently, there are pressures to pass
legislation with large majorities (if not with universal approval) and therefore more
projects are generated out of this complex political process (see also Niou and Ordeshook
1985; Baron 1989; and for evidence at the state and local levels see Gilligan and
Matsusaka 2001).
29
Research has shown, however, that the scale of expenditure increases depends on
the number of legislative chambers. The original theory and preliminary empirical tests
were in relation to unicameral legislatures where the relationship has been consistently
confirmed. However, research on bicameral legislatures designed to represent different
constituencies has found that they mitigate the effect of adding an additional seat by
reducing the likelihood that the two chambers will agree on which groups to tax and
subsidize; moreover, power asymmetries inhibit the ability of each legislative chamber to
institute its own preferences (Bradbury and Crain 2001: 315). In fact, recent work has
shown that as the seat ratio between the lower to upper chamber increases, spending
actually decreases. The logic behind this finding is that lower chamber districts are
typically embedded geographically within each upper chamber districts; that is, each
upper chamber district contains multiple lower chamber districts. As each upper chamber
district is divided into additional lower chamber districts the constituency size of each
lower chamber district decreases. The effect of a smaller constituency is to reduce the
lower chamber members’ benefit from over utilizing common pool resources to fund
projects (Chen and Malhotra 2007: 658). The implication of these studies is that the
number of legislators affects expenditure levels; however the effect will be greatest in
unicameral legislatures.
The common-pool resource problem identified within the legislature exists within
the executive branch of government as well. Specifically, each minister within the cabinet
is partially responsible for creating and making demands on the overall budget. As such,
just as within the legislature, as the number of decision-makers increases, each minister
pays a decreasing share of the revenue costs of each dollar of expenditure he or she
30
proposes (Kontopoulos and Perotti 2008: 82). A number of studies have examined the
effects of the number of spending ministers within the cabinet on fiscal policy. The
evidence is mostly derived from samples of OECD countries, however, the consensus is
that an increase in the number of ministers causes expenditures to increase (Kontopoulos
and Perotti 2008; Perotti and Kontopoulos 2002; Ricciuti 2004) as well as budget deficits
(Volkerink and de Haan 2001; Wehner 2010). While limited in scope, there is some
evidence that the partisan makeup of the executive branch conditions the effect of adding
an additional minister to the cabinet. Wehner (2010) argues that political parties may be
able to force their ministers to internalize a greater proportion of the costs than they
otherwise would, in which case the less political fragmentation that exists within a
cabinet (i.e. the more each spending minister belongs to the same political party), the less
severe is the tendency to overspend. The author finds that both expenditures and the
deficit increase as the level of political fragmentation within the cabinet rises.
Nevertheless, the results suggest that the direct effect of the number of spending ministers
is more pronounced (Wehner 2010: 642-644).

4.3 Political Parties
Scholars have extended the above analysis to examine the effect of political
parties on fiscal policy. Two arguments exist to account for observed systematic
differences in fiscal policies. First, evidence suggests that if a majority party exists, its
size can affect expenditures. Specifically, above some threshold the proportion of the tax
burden shared by members of the majority party rises since the costs can no longer be
exported to the constituencies of nonparty members. Therefore, up to a point
31
expenditures will rise as the size of the majority party increases, falling thereafter (Inman
and Fitts 1990). Where no party holds a legislative majority, an increase in the number of
parties in the legislature makes membership in the governing coaling uncertain and
makes the coalition that does form less stable. In order to avoid defeat of their budget
proposals the coalition will include the preferred projects of other parties.
24
Thus, similar
to the addition of a legislator, an increase in the number of parties raises expenditures
(Mukherjee 2003; see also Perotti and Kontopoulos 2002).
It may be, however, that the difference in electoral accountability between
governments comprised of single-party majorities and multiparty coalitions is responsible
for the differences in fiscal policy decisions. Single-party governments typically
represent a broad array of societal interests and are thus held accountable for all of their
policy decisions (Cox 1990). Each party in a multi-party government, by contrast, usually
represents a narrow group or smaller subset of interests and is therefore held accountable
only for the policies for which they care the most. Since each party is accountable for the
policy areas in which they have the largest stake, they are given priority over these areas
(in the form of ministries). Moreover, since these smaller parties represent smaller groups
most of the costs of proposed policies are externalized to other groups. As a result,
spending is larger the more parties there are in the government coalition (Bawn and
Rosenbluth 2003, 2006).
25

In terms of the composition of expenditures, since parties in multiparty
legislatures typically represent constituencies with specific demographic or geographic

24
Note that this is simply an extension of the “universalism” assumption found in Weingast (1979) and
Shepsle and Weingast (1981).
25
Note that the two arguments posit different causal mechanisms for larger government but are nonetheless
empirically equivalent.
32
characteristics, there is less incentive for each party to spend on public goods whose
benefits are dispersed across constituencies and more incentive to spend on transfers and
subsidies which can be targeted. As a result, with an increase in the number of parties
represented in the legislature, the proportion of expenditures on public goods will decline
and expenditures on subsidies and transfers will increase (Mukherjee 2003).

4.4 Elections and Electoral Systems
Electoral systems have also been shown to affect a country’s level and
composition of expenditures. Recent studies by Grilli et al. (1991), Persson and Tabellini
(1999, 2000, 2002, 2003a,b, 2004), Persson (2002, 2004), Persson et al. (2000, 2007),
and Milesi-Ferretti et al. (2002) find that majoritiatian governments and presidential
systems lead to smaller governments compared to proportional and parliamentary
systems. Theoretical models suggest that the reason for this disparity is that majoritarian
elections focus competition to key swing districts, which leads politicians to target
smaller but critical geographic constituencies. Likewise, presidential regimes foster more
intensive political competition than parliamentary regimes because legislative coalitions
are less cohesive—on account of the lack of an executive confidence requirement—
which results in minorities fighting over different issues with spending typically being
targeted toward the constituencies of the powerful office-holders (Persson and Tabellini
2000). The confidence requirement of parliamentary regimes, however, results in more
legislative cohesion characterized by a stable majority of legislators supporting the
executive and voting together on legislation while pursuing the joint interests of its
collective constituency (Diermeier and Feddersen 1998). As such, spending programs
33
benefit a majority of the electorate in the form of broad transfer programs (Persson and
Tabellini 2004).
The electoral formula is also important in determining the composition of
government spending. The minimum winning coalition needed to be electorally
successful is greater under proportional representation than under plurality rule.
Consequently, politicians in a system characterized by proportional representation are
required to internalize the policy benefits of larger segments of the population. As a result,
the composition of government expenditures is oriented toward spending programs that
benefit large segments of the population, such as encompassing social security and
welfare programs. Conversely, under plurality rule the candidate that receives the most
votes wins, creating the incentive for politicians to target small, geographically
concentrated groups of voters in the form of localized goods and services (Milesi-Ferretti
et al. 2002; Persson 2004; Persson and Tabellini 2004).
It has also been argued that elections—or rather the timing of elections—affects
fiscal policy. Specifically, a number of authors have posited that prior to elections
politicians attempt to stimulate the economy as part of their attempt at reelection.
According to one strand of theory, politicians care only about reelection and voters base
their voting behavior on the performance of the economy leading up to an election. As
such, the government will use such policy instruments as taxes, outlays, interest rates,
and so on in order to stimulate the economy. As a result, the theory predicts politically-
driven business cycles with pre-electoral high growth, low unemployment and high
inflation around the election and a recession thereafter (Nordhaus 1975; MacRae 1977;
Rogoff and Sibert 1988; Rogoff 1990; Alesina and Roubini 1992; Paldam 1997). Others
34
argue that such cycles appear because politicians have partisan preferences over policy
issues and voters support parties based on the implied income distribution of such
policies. Specifically, left-wing parties are assumed to prefer low unemployment and
right-wing parties are assumed to prefer low inflation. Since these macroeconomic
outcomes have redistributional consequences the lower class tends to support the left and
the upper middle class tends to support the right (Alesina 1989). Thus, political business
cycles are explained by government turnover; politicians have partisan preferences over
policy issues and therefore inflation can be expected to be high and unemployment low
during the tenure of left-wing governments, and vice versa during right-wing
administrations (Hibbs 1977, 1987; Alesina 1987). To date, the evidence on the existence
of political business cycles is mixed, at best, but has generally been overshadowed by the
so-called “partisan theory” (Paldam 1997); that is, there is sound evidence of a
relationship between the share of left-parties in government and the size of the public
sector (Hibbs 1977) and the growth of budget deficits (Alesina 1989; Alesina et al. 1992).

4.5 Levels of Government
A separate group of scholars emphasizes the relationship among levels of
government, specifically the degree of fiscal federalism or centralization.
26
Authors such
as Wildavsky (1974), Freeman (1975), and Tarschys (1975) argue that over time
increases in public expenditure tend to be greater at the regional and local levels. These
authors contend that centralization allows government decision-makers to oversee
spending and because they are aware of the relative trade-offs among alternative policy

26
“Fiscal centralization” refers to the proportion of all governments’ revenue generated by the central
government.
35
expenditures they are in a better position to limit aggregate spending. Others, such as
Brennan and Buchanan (1977, 1978, 1980), operating under the assumption that the
central government seeks to maximize its revenue share, argue that fiscal decentralization
has the effect of forcing local, state, and national governments to compete with one
another for tax revenues, thus eliminating the central government’s monopoly over
taxation and ability to spend. While collectively the empirical evidence of fiscal
centralization is mixed, Rodden (2003) finds that the effect of decentralization is
dependent upon the location of expenditure authority and taxation authority. Where
decentralization is funded by intergovernmental transfers or revenue sharing schemes,
there is faster growth in overall government spending. This is because the benefits of
public goods are localized, whereas only a fraction of the costs are internalized.
Therefore, there is an inherent bias to overspend. Slower government growth requires that
public goods be funded by local taxation, which is the only possible way to induce the tax
competition mechanism posited by Brennan and Buchanan.

4.6 Budget Institutions
Finally, a sizable literature has examined the effects of budget institutions on
various measures of fiscal performance, namely debt and deficits. Three categories of
budget institutions have received the most attention. The first is numerical targets, which
are meant to impose constraints on certain budgetary aggregates. Examples include
balanced-budget laws, numerical debt-ceilings, and limits on the growth of taxes and
expenditures (von Hagen 2006). Empirical work has show that such fiscal restraints can
lead to lower deficits and more rapid adjustment to fiscal shocks (Eichengreen 1992; Alt
36
and Lowry 1994; Poterba 1994). The second is procedural rules, which dictates the
preparation and legislative approval of the budget. Studies have found that procedures
that grant the treasury minister substantial power in preparing the budget and restrict the
legislature’s ability to make amendments produce budgets that favor the preferences of a
minimal majority and result in more fiscal restraint, lower and less persistent deficits, and
more rapid adjustment to fiscal shocks (Alesina and Perotti 1996, 1999). The final
category of budget institutions is the rules determining the degree of transparency of the
budget process. A transparent budget process provides the public with all of the necessary
information regarding government functions, fiscal policy intentions, public sector
accounts, and projections in a timely and systematic manner (Kopits and Craig 1998;
Gollwitzer 2011). Transparency is said to ease the asymmetry of information between
politicians and voters, reducing the ability of the former to strategically overspend in
pursuit of opportunistic goals (Cukierman and Meltzer 1986; Alesina and Cukierman
1990)
27
, and the tendency of the latter to overestimate the benefits of public expenditures
and underestimate their present and future costs (Buchanan and Wagner 1977). Evidence
suggests that transparent budget procedures are indeed associated with lower levels of
public debt and deficits (Alt and Lassen 2006).

5. Critique
While the preceding literature has enhanced our understanding of the impact of
political institutions on fiscal policy, there is ample room for improvement. There are
four shortcomings, in particular, that need to be addressed: (1) the effects of political

27
Rogoff and Sibert (1988) and Rogoff (1990) make a similar argument with regard to political budget
cycles discussed above. Policymakers are able to abuse fiscal policy prior to elections if voters are unable
to observe the level and financing of budgetary expenditures.
37
institutions on an exhaustive list of fiscal policy outcomes has been under-theorized; (2)
the explanatory power of existing theories is still weak; (3) the most influential
institutions for determining fiscal policy merit much more scholarly attention; and, (4)
the proper set of actors and their roles in determining fiscal policy must be more carefully
considered. While there is variation in the extent to which the above theories suffer from
these deficiencies, each of the theories is negligent in at least one of these aspects.
First, consider the problem of adducing the full range of implications from a
theory. If any theory that seeks to explain the relationship between some factor (here
political institutions) and some fiscal outcome is to be especially useful, it should imply
something about other fiscal outcomes as well. However, the extant literature tends to
limit the analysis to one explanatory factor and only one or two fiscal outcome variables.
In other words, the literature on political institutions and fiscal policy suffers from low
the “leverage” problem, in that it has not explained (or cannot explain) a host of effects
by reference to one or a few variables (King et al. 1994: 29). To elaborate, the collective
literature on bureaucracy, legislatures, and fiscal federalism only focuses on a singular
outcome (e.g. expenditures), while the remaining scholarship (with the exception of the
literature on electoral systems and fiscal policy outcomes) primarily focuses on two fiscal
outcomes at most, such as expenditures and deficits. Either the full implication of these
various theories has yet to be fleshed out, or the theories have little to say about the
relationship between the specific political institution and other fiscal policy outcomes. In
either case our knowledge of the determinants of fiscal policy has suffered as a result.
For some theories, the lack of evidentiary support presents a major problem. For
example, Johnson and Libecap (1989) examined 45 bureaucracies within the U.S. federal
38
government to determine whether agency growth led to an increase in bureaucratic
salaries and found no significant relationship. Likewise, in her extensive studies of
bureaucratic growth Carroll’s (1989, 1990) evidence best supported the view that
bureaucratic growth was the result of an increase in the demand for the bureau’s services
rather than increases in the bureau’s budget. Similarly, the evidence for electoral effects
on fiscal policy is mixed. Persson and Tabellini (2003b) find evidence of a political
revenue cycle, but no politically induced cycle in expenditures, transfers, or deficits.
Brender and Drazen (2005) argue that political business cycles are present during the first
few elections in countries that have made the transition to democracy, but they find no
statistically significant evidence for political business cycles in established democracies
(however, see Alt and Lassen (2006) for contrary evidence). Those electoral effects that
have been found relate to the composition of spending rather than in the overall level
(Peltzman 1992; Drazen and Eslava 2010). Thus, in many cases the extant literature is
unable to offer empirical support even for the narrow questions that have been posed.
However, the primary shortcoming of the preceding theories is that they all suffer
from a serious conceptual flaw. The vast majority of extant theories stop just shy of
explicating the institutional context within which revenue and expenditure decisions are
actually made. In reality, revenues and expenditures are determined by the budget, and
such decisions are governed by a set of rules and regulations that structure how budgets
are prepared, approved, and implemented, and which assign roles and responsibilities to
the various actors at each stage of the process (Dabla-Norris et al. 2010: 4). Such rules
and regulations constitute “budget institutions” (Alesina and Perotti 2008: 14) and any
theory that seeks to explain fiscal outcomes such as central government revenues and
39
expenditures, deficits, debt, and budget composition is remiss if proper consideration is
not paid to these institutions.
To the extent that the literature has examined budget institutions it has only
emphasized a subset of institutions and examined a limited number of outcomes. The
remainder of the literature either focuses on macro-institutions, whose effect on fiscal
policy formation and implementation is only indirect, or only concentrates on a subset of
relevant decision-makers. Such institutions as elections, federalism, and the type of
electoral system certainly affect fiscal policy outcomes. For instance, electoral systems
may affect fiscal policy through their effects on representation, the number of political
parties, and the types of governments that are formed (Lijphart 1994, 1999), as well as
their effects on electoral competition (Persson and Tabellini 2000).
28
However, these
effects do not take into account how fiscal policy is actually made. Similarly, the
literature on the executive and legislature typically neglects the strategic interaction
between branches of government when determining fiscal policy, i.e., budget formulation
within the executive and submission to/subsequent amendment and approval by the
legislature. In the case of the literature on the bureaucracy, in many instances other
important actors are not only neglected (i.e., the executive), but those branches of
government that are considered, often play only a passive role (i.e. the legislature).
Furthermore, in much of this literature the formation fiscal policy (in this case
expenditures) is assumed to be made by the bureau, when in reality it is much more
common for final choices to be made by the legislature (Miller and Moe 1993: 299).

28
In their defense, Persson (2004) and Persson et al. (2007) acknowledge that the fiscal policy effects of
electoral systems are indirect. Spending, for example, is argued to be higher under proportional
representation because PR systems typically have a larger number of effective parties which typically raises
the incidence of coalition governments that spend more.
40
In light of these various methodological shortcomings, this dissertation seeks to
explain how institutions determine fiscal policy outcomes and how the relevant actors
interact within these institutions when formulating policy. If the level and composition of
government expenditures, deficits, and debt levels are the result of the budget process, it
stands to reason that the rules and procedures which determine how the budget is formed,
authorized and implemented are likely to have an appreciable effect on fiscal policy.

6. Conclusion
The purpose of this chapter has been to introduce the reader to the generalities of
fiscal policy: recent trends, its economic effects, and existing causal explanations. The
fact that governments are more involved in the management of their economies since the
Great Depression is a reality of modern political economic management. However, as this
chapter has demonstrated, the ways and degree to which governments intervene through
the exercise of fiscal policy differs greatly both between countries and within countries
over time. The fact that fiscal policy displays such variation can mean the difference
between policymaker’s ability to stimulate a lagging economy or to further drive it into
recession (or depression). The current sovereign debt crisis in Europe is evidence of this
fact.
Given the importance of fiscal policy, scholars have generated a number of
theories which seek to determine those variables that most influence fiscal policy choices.
This analysis of the state of the literature, while brief, has hopefully convinced the reader
that despite the contributions made to our understanding of the formulation of fiscal
policy, the extant literature is deficient with respect to four important criteria. First,
41
posited theories have yet to satisfactorily explain fiscal outcomes by reference to a
handful of key variables. In all likelihood this problem is a result of incomplete
theorizing and can be remedied by fleshing out the additional findings in the data. Second,
the empirical track record of a plurality of these theories is poor. This is probably due to
the third and forth criteria, in that theories by and large do not take into account the most
consequential institutions and the proper set of actors which determine fiscal policy.
This chapter has focused on the political determinants of fiscal policy, with a
special emphasis on political institutions. In the following chapter I attempt to build on
the positive aspects of extant theories and construct a more comprehensive theory of
those factors which determine fiscal policy. To foreshadow that discussion, I build on the
budget institutions literature; however, I also draw from the literature on agenda
formation and voting in legislatures, fragmentation, electoral systems, and political
budget cycles. In developing this theory, I am cautious to avoid the four deficiencies
previously discussed. Specifically, in emphasizing budget institutions I am able to focus
precisely on those institutions and actors which are most directly responsible for
formulating and executing fiscal policy. Furthermore, I derive thirteen hypotheses for a
full set of fiscal outcomes, i.e., expenditures, debt, deficits, and expenditure composition.
In subsequent chapters these predictions are tested against the data, using both
quantitative econometrics and qualitative case study methods. In contrast to existing
studies, I use a more comprehensive dataset; whereas previous work uses data limited to
a single region or to the OECD countries, the dataset used here is one that samples from
all regions and from both developing and developed countries alike. In addition, the case
studies are selected from the statistical results, thus avoiding potential issues of selection
42
bias. While these improvements cannot guarantee that the theory put forward here will be
more successful empirically than previous theories, this study does promise to test the
most relevant variables in their entirety.

43
CHAPTER 3: A THEORY OF BUDGET INSTITUTIONS
AND FISCAL POLICY

1. Introduction
Any theory that seeks to explain the determining factors of fiscal policy must
contend with two adverse features of public finance: the agency problem and the
common-pool resource problem. Both issues arise from the fact that those responsible
for public spending and those who benefit from such expenditures are not necessarily
those that bear the burden of the costs. The agency problem surfaces because budgeting
invariably involves delegation of some sort; typically voters delegate budget authority to
elected politicians and elected politicians delegate implementation powers to bureaucrats.
The problem arises when the preferences of the “agent” (i.e. the person entrusted with
authority) conflict with the preferences of the “principal” (i.e. the person delegating
authority). That is, there will be an agency problem when the agent in a relationship does
not behave as the principal would like that person to behave (Hallerberg 2003).
In essence, the agency problem boils down to an asymmetry of information
between the two participants. Either the principal does not possess complete information
regarding the agent’s abilities and preferences, or the principal cannot observe the agent’s
actions (Dixit 1998: 85-86).
29
This asymmetry of information is precisely what allows the
agent to behave differently from what is optimal from the point of view of the principal.
In the budget process, the agency problem is said to adversely affect fiscal policy;
expenditures and deficits may be larger than, and the budget composition different from,

29
When principals are ill-informed about agents’ attributes, such as their skills or tastes, the agency
problem is one of “adverse selection.” When principals are unable to completely monitor their actions, the
problem is one of “moral hazard” (Dixit 1998: 85-86).

44
that preferred by the average voter and from that which is considered economically
prudent.
A common-pool resource (CPR) is a resource system that, because of its sheer
size, makes it costly to prohibit potential beneficiaries from participating in its use
(Ostrom 1990: 30). In terms of fiscal policy, the CPR is government revenues. The
common-pool resource problem results when policymakers fully internalize the benefits
of their spending decisions and fail to consider fully the effect of their decisions on the
common pool (Hallerberg 2003). That is, politicians oftentimes spend money from a
general tax fund on public policies that only benefit subgroups of the population. Due to
the fact that the size of the group responsible for footing the bill of such policies (i.e. the
general taxpayer) is substantially larger than the group of beneficiaries of such policies
creates a discrepancy between the net benefits to society versus the net benefits to the
targeted group (Von Hagen 2006). The ability of policymakers to ignore much of the tax
implications of their decisions permits them to be less fiscally responsible. Recent
theoretical work has shown, in fact, that when current spending can be financed by
borrowing, the common pool resource problem leads to higher expenditure levels as well
as larger deficit and debt liabilities (Hallerberg and Von Hagen 2008; Velasco 2000,
2008).
The subject of this chapter is the institutional solutions to the agency and common
pool resource problems. Specifically, this chapter discusses how budget institutions
address these issues and develops a theory of the effect of budget institutions on fiscal
policy outcomes. It discusses three budget institutions in particular: fiscal rules,
transparency, and procedural rules, which were introduced in Chapter 2.

45
Section 2 of this chapter examines each institution individually. In each case the
institution is first defined and its purpose in addressing either the agency or common-pool
problem is discussed. It is here that I present my main theoretical claims. Briefly, these
are that: (1) fiscal rules can mitigate the common-pool resource problem by limiting the
extent to which policymakers are able to exploit fiscal policy instruments; (2) fiscal
transparency eases the agency problem by removing the informational advantage
policymakers possess over the general public and diminishes the common-pool problem
by making their policies and actions known; and, (3) procedural rules soften the
common-pool problem by granting agenda-setting power to policymakers with an
incentive to act in the interest of the general taxpayer and, presumably, to those who are
in a position to understand how policy decisions affect the government’s budget
constraint. Given that these institutions are meant to reduce the extent of the agency and
CPR problems, we would expect countries that adopt these institutions to show more
fiscal restraint. I then buttress the claims made regarding the effectiveness of each
institution in mitigating these problems by presenting (informally) the results of a number
of formal models. I then discuss the issues of institutional selection. This is particularly
important because as we shall see below, establishing each of these institutions entails a
significant reduction in the policymaking power of the political actors involved.
Therefore, a convincing case must be made as to why policymakers would select such
institutions.
Section 3 presents the hypotheses developed from the theory. In total thirteen
hypotheses are derived for the fiscal outcomes of government expenditures, expenditure
composition, debt, and deficits. An important factor in reaching these conclusions is the

46
assumption that voters are fiscal conservatives. Thus, some time is spent justifying this
claim and explaining how it affects the derivation of the hypotheses. Section 4 concludes
and briefly discusses the empirical tests to be undertaken in the following chapter.

2. Budget Institutions and Fiscal Policy
2.1 Fiscal Rules
A fiscal rule is a policy rule that places a permanent constraint on fiscal policy
and is typically expressed in terms of a summary indicator of fiscal performance (Kopits
and Symansky 1998). Examples of fiscal rules are numerical targets, which impose
constraints on certain budgetary aggregates. These include balanced-budget laws,
numerical debt-ceilings, and limits on the growth of taxes and expenditures (Von Hagen
2006). The purpose of such policy rules is to lend credibility to fiscal policy by making
discretionary intervention less likely. In addition, fiscal rules are meant to ensure that
policies are predictable by guaranteeing that they will be followed irrespective of the
government in power (Kopits 2001). If such constraints are both credible and enforced,
they are thought to eliminate politicians’ ability to pursue opportunistic policies (through,
for instance, excessive deficit spending). Essentially, only the relative spending priorities
and tax structure would be up for debate and not the budget balance and/or expenditure
levels since they would be legally predetermined (Kopits 2001: 8). In other words, fiscal
rules may mitigate the extent of the common-pool resource problem by limiting the
extent to which policymakers are able to exploit fiscal policy instruments.
However, there are a number of arguments asserting the uselessness and
ineffectiveness of fiscal rules. These arguments are based on both theoretical and

47
practical reasoning. From a theoretical perspective, neither macroeconomic theory nor
public finance is constructed on rules-based fiscal policy (Kopits 2001: 6). A central
result from welfare economics is that when a social-welfare-maximizing policymaker is
uninhibited in selecting policy—that is, he or she is able to use discretion—welfare will
be at its highest level. The intuition is that for a given policy guideline (e.g. a fiscal rule),
an unconstrained policymaker can always adhere to that principle if he or she so chooses;
however, allowing the policymaker to stray from a predetermined policy if necessary
should only lead to an increase in welfare (Drazen 2002: 3).
Moreover, standard Keynesian economics suggests that budget deficits should be
run when the economy suffers from insufficient aggregate demand and that budget
surpluses should be run whenever there is an excess of aggregate demand; in the former
case the goal is to combat unemployment and in the latter instance to quell inflation
(Alesina and Perotti 1996). In addition, the “tax-smoothing theory” posits that deficits
and surpluses should be used to correct the distortionary effects of taxation such that
deficits should be permitted when spending is temporarily high or revenues low (for
instance the former should be permitted during wars or emergencies, and the latter during
recessions) in order to implement the optimal tax policy (Alesina and Perotti 2008). In
either case, fiscal rules such as balanced budget laws would hinder governments’ ability
to pursue such stabilization policies. In other words, discretion, rather than rules, is
necessary to achieve the numerous fiscal goals and functions of government; for instance,
stabilization, distributional fairness and allocative efficiency (Kopits 2001).
On a practical level, there is the argument that fiscal rules are neither necessary
nor sufficient to ensure budgetary discipline. Fiscal rules are not necessary to ensure that

48
prudent fiscal policies are followed because it is possible for governments to do so
without such formal commitments. That is, governments can simply develop a reputation
for sound fiscal policy (and macroeconomic policy in general), obviating the need for
official constraints (Kopits 2001). Critics of fiscal rules point to the success of U.S.
macroeconomic policy in the 1990s in keeping inflation low and budgets in surplus.
30

According to this view, fiscal rules are redundant when a strong reputation is equally
capable of establishing the credibility of economic policy.
Furthermore, fiscal rules are not a sufficient condition for prudent budgeting. This
is because it is widely acknowledged that fiscal policy rules establishing stringent
numerical targets create the incentive for policymakers to engage in “creative
accounting” in order to appear to meet such policy requirements when in fact they are
not.
31
Examples of creative accounting practices include privatization of government
assets (which is counted as reducing the deficit but in reality goes toward financing the
deficit), cuts in public investment (which is treated as an expenditure), reductions to
operations and maintenance spending, delayed payments to workers and suppliers, and
shifting taxes forward in time, to name a few (Easterly 1999). In essence, in order to
circumvent restrictions governments will trade asset reductions and increases in implicit
liabilities for increases in explicit liabilities (Easterly 1999).
32
The problem is that

30
Of course, subsequent deficits in the U.S. significantly weaken this argument.
31
Milesi-Ferretti (2003: 379) states that “…a measure implying an improvement in the fiscal balance is
considered to be creative accounting if it does not imply an improvement in the inter-temporal budgetary
position of the government sector at large (an increase in the government’s net worth).”
32
In addition, the true state of the national account can also be obfuscated through the overestimation of the
expected growth of the economy so as to overestimate revenues and underestimate expenditures and by
overestimating the revenue from a new tax so as to stall adjustment until a future period (Alesina and
Perotti 2008). At the sub-national level (at least among U.S. states), state governments circumvent debt
limitations on guaranteed debt by issuing more nonguaranteed debt (Van Hagen 1991). The most common
non-guaranteed debt instruments used by state governments are revenue bonds or the devolution of debt to

49
creative accounting tactics, aside from simply circumventing fiscal restraints, are likely to
induce significant economic distortions. In other words, poorly designed rules are likely
to backfire, perhaps even leading to greater distortions than they were intended to address
(Drazen 2002). Another problem with fiscal rules is that they can simply be removed or
modified in circumstances in which they become too constricting.
For example, in 1985 the U.S. Congress passed the Gramm-Rudman-Hollings
Deficit Reduction Act which laid out a series of annual deficit targets designed to reduce
the U.S. budget deficit to zero within five years.
33
If in any given year the predetermined
targets were not met, then equal cuts in defense and nondefense expenditures would be
made in order to meet the statutory target. When these targets could not be met, creative
accounting practices, such as the selling of assets were undertaken, and in instances when
the targets became especially binding, such as in 1987 and 1990, Congress simply passed
new legislation to revise the targets.
34
In the end, the Deficit Reduction Act failed to
eliminate U.S. budget deficits. Similarly, in the run-up to the European Monetary Union
Greece recorded capital transfers to public enterprises as “equity increases” which were
excluded from the calculation of the budget deficit.
35
Together with the issuance of zero-
coupon bonds intended to reduce interest payments, these measures resulted in the
artificial reduction of Greece’s 1997 budget deficit by 1% of GDP.
Creative accounting practices are also prevalent among developing countries. For
instance, during the period 1989 to 1993 Nigeria was in the process of adjusting to two

local governments (Kiewiet and Szakaly 1996). When revenue bonds were unavailable, debt limitations
were avoided through bond issuance by public authorities (Bunch 1991).
33
This example is taken from Gramlich (1995: 179-180).
34
In their study of indebtedness among U.S. States, Kiewiet and Szakaly (1996) found that twelve states
went so far as to amend their constitutions to alter debt limitations!
35
This example is taken from Milesi-Ferretti (2000: 5-6), which also discusses creative accounting
techniques undertaken in France and Italy.

50
IMF stand-by agreements and two World Bank adjustment loans that limited deficit and
debt levels.
36
To meet its targets the government sold its equity shares in oil ventures. It
was later found that US$12.2 billion of oil money had disappeared, lining the pockets of
state officials and thus lowering the net worth of the public sector. In addition, in a
similar fashion to Greece, Brazil issued zero-coupon bonds in 1988 in an effort to reduce
that year’s expenditure on interest payments. As the argument goes, and the above
examples demonstrate, policymakers will find ways to modify and elude fiscal rules if it
is in their interest to do so.
Nevertheless, budget rules are argued to be optimal over discretion when the
behavior of individuals is dependent upon future policy (Drazen 2002). This has to do
with the possibility that a “time-consistency” problem could arise, whereby a
policymaker has ex post incentive to go back on a policy promise that may have been
previously optimal ex ante (Dixit 1999: 63).
37
In other words, it is possible for there to be
a severance between the incentive to make a policy commitment and the incentive to
keep it.
38
Such a policy is not only lacking in credibility, but is also likely to lead to
suboptimal results (Kydland and Prescott 1977). The reason is that private actors who are
aware of the incentive for policymakers to deviate from the promised policy expect them
to do so and thus will adjust their behavior accordingly. The altered behavior means that

36
These examples are taken from Easterly (1999: 58-60), who discusses creative accounting practices
among countries that have undergone IMF and World Bank fiscal adjustment programs.
37
Persson and Tabellini (2000: 277) define an “ex ante optimal policy rule” as “…the best one at the
starting date of the dynamic policy game. Such a policy takes into account all inter-temporal effects of
economic policies…” Conversely, an “ex post optimal policy rule,” is one that “…describes the policies
appearing optimal from the perspective of any later date, with all previous private decisions…taken as
given.” They go on to state that because rational actors expect policy to be ex post optimal, ex ante policies
lack credibility. Only policies that are ex post optimal are capable of being credible and therefore such
policies will be expected by rational actors.
38
Note that the time-inconsistency problem creates a moral hazard for policymakers and thus is a type of
agency problem. See footnote 1.

51
the policy will not have its intended effect and likely will make all participants worse
off.
39

The primary argument for fiscal rules is that there is a natural tendency for
governments to run budget deficits (Drazen 2002).
40
Some of the political dynamics that
underpin this tendency were discussed in the previous chapter. Thus, the attractiveness
of fiscal rules is that by constraining policymakers they will ostensibly reduce or
eliminate the bias toward deficits. This is made possible, because, unlike discretion, rules
imply commitment and commitment is important if the expectation of future government
policy affects the current choices of private economic actors (Stokey 2003: 10).
Furthermore, creative accounting is not necessarily as grave a concern as some
critics of fiscal rules make it out to be. It is widely recognized that the scope for creative
accounting can be reduced through the proper design and implementation of fiscal rules
and budget institutions more generally (Drazen 2002; Kopits 2001). One of the most
efficient and increasingly common ways to combat creative accounting practices is by
increasing the transparency of the budget process, which I discuss in the following
section. Within a more transparent budgeting process, deviations from the rule are made
more obvious and violators can be more readily punished for their subterfuge.
This raises the question of how fiscal rules may discipline political behavior. First,
government policymakers may not be as omnipotent and benevolent as many economic

39
A theoretical example from monetary policy may help to illustrate this point. Barro and Gordon (1983)
present a model in which a government makes an announcement to commit to a targeted inflation rate of
zero. However, if such a pronouncement were believed, the government would have the incentive to select
a higher inflation rate so as to lower unemployment. Private actors, recognizing this incentive, correctly
anticipate this change in policy and the result is unemployment at its natural rate however with a sub-
optimally high rate of inflation.
40
Hercowitz and Strawczynski (2004), for example, show that among the OECD countries there exists a
“ratchet effect,” whereby expenditure increases during recessions are only partially reduced during
expansions. Likewise, Persson and Tabellini (2003b) find a similar ratchet effect among countries with
proportional-parliamentary electoral institutions.

52
models portray them. Furthermore, as I argued in the previous chapter, the formulation of
fiscal policy is hardly a matter of a benevolent social planner optimizing a social welfare
function. Rather, policymakers are politically motivated in their choice of fiscal policy,
which explains the resultant biases. More to the point, when there exists uncertainty
regarding future economic conditions, governments facing (or that will face) reelection
prefer to hold on to as many policy options as possible (Cukierman and Meltzer 1986b:
378). Barring any restrictions to the contrary, policymakers can be expected to renege on
their promises if they stand to benefit politically from doing so. Rules which specify a
detailed course of action and are difficult to change, then, may protect the public from
government officials who might be imprudent, greedy, or myopic (Stokey 2003: 11).
To illustrate, Cukierman and Meltzer (1986b) present a model in which the
electoral prospects of a politically motivated policymaker depend upon the level of social
welfare provided to the public during his/her term in office. The level of welfare is
affected by the policymaker’s choice of policy and a random state variable (i.e. the state
of the economy). Current policies affect the performance of the economy in the current
and in future periods, so current welfare is determined by policy settings in the past and
current period. Voters base their electoral choice on the level of welfare experienced
under the current administration and on the prospects for future welfare levels, whereas
the policymaker is primarily concerned about reelection. Governments differ from one
another in their ability to accurately forecast economic performance when making policy.
Voters are only able to observe their level of welfare. The electoral incentive and
asymmetry of information induces the policymaker to implement policies solely to
improve his/her reelection probability without regard for the effects such policies will

53
have on the public’s welfare in future periods. The conclusion is that policymakers will
not select the socially optimal choice of policy. The authors show, however, that the loss
in welfare is reversed in the case where government policy is bound by a rule, which
eliminates the ability of government to exploit its information advantage.
41

Second, rules codified into law have penalties attached to them. Thus, there are
explicit costs associated with a violation of the rule.
42
The fact that rules have associated
penalties is partly why they are considered to be more credible than promises alone; the
penalties and costs are the mechanisms which endow rules with force. For example, in
Brazil politicians that violate the Fiscal Crimes Law may be detained for up to four years
(IMF 2001). While it is true that rules cannot force politicians to behave in a fiscally
responsible manner, they do increase the likelihood and raise awareness when rules have
not been followed. In essence, policy credibility is enhanced by rules because they raise
the cost and lower the benefit from deviating from a given policy (Drazen 2002: 11-12).
Nevertheless, one need not possess a cynical view of government to understand
why fiscal rules are useful or even why policymakers may wish to limit their own
discretion. As Stokey (2003) has shown, even far-sighted policymakers may find the
institutionalization of fiscal rules to be advantageous. This is because often times it is
difficult for “good” policymakers to distinguish themselves from politicians that are
“bad” (i.e. myopic, greedy, incompetent) and distrusted by the private sector.
43
Good

41
Incidentally, Cukierman and Meltzer (1986b) also show that the only other instances in which social
welfare is maximized is when a benevolent social planner sets policy (both with and without a policy rule),
or when all voters are fully informed. Since both of these situations are unlikely to be realized in practice,
one may reasonably conclude that fiscal policy rules may indeed be useful if the object is to maximize
social welfare.
42
Of course, there could be costs to reneging on a promise in the form of loss to reputation.
43
In Stokey’s model, a “good,” or “Ramsey” policymaker is one that raises revenue so as to maximize the
expected discounted utility of a representative household. Conversely, a “bad” policymaker is one that sets

54
politicians, when elected, may find themselves forced to deal with the inheritance of bad
policies (e.g. large outstanding debt). Due to the fact that bad policymakers adversely
affect private sector behavior, when in power good governments find it necessary to
offset the policies of the bad types. If the probability that a myopic policymaker will
assume power in the future is sufficiently high, a policy rule that places a restriction on
the policy of the myopic type will lessen or eliminate the potential damage done.
44
In
addition, when the policy rule is in place both types of government follow the same
policy. Thus, by tying the hands of future politicians the fiscal rule makes policy both
credible and predictable. Importantly, the author shows that the effect of the rule far
outweighs any additional benefit that may have come from discretionary policy.
45

Therefore, a fiscal rule is potentially adopted when a particular policymaker (or set of
policymakers) fears that future policy may be distorted by others with divergent policy
preferences and therefore prefers to commit such politicians to a particular policy path.
In summary, a fiscal rule is a policy rule intended to constrain the scope for
discretionary fiscal policy in some way. Typical rules include balanced-budget laws,
numerical debt-ceilings, and limits on the growth of taxes and expenditures. The purpose
of such rules is to lend credibility and predictability to fiscal policy by limiting policy
discretion and ensuring that policy is consistent across administrations. In other words,
fiscal rules are intended to demonstrate a government’s credible commitment to sound
fiscal policy. Fiscal rules solve the time-inconsistency problem because they are codified

current tax rates only in order to maximize current-period utility; in other words, the bad type is myopic
and the good type is farsighted.
44
Similarly to the Cukierman and Meltzer model, a discretionary equilibrium is obtainable, however, only
if it is known with certainty that the government is of the “good” type.
45
This is true so long as there is a positive probability that the government is myopic. Expected utility is
higher under discretion when the government in power is certain to be of the good type. However, as stated
above, the likelihood of a benevolent social planner holding office in the real world is slim.

55
into law and carry penalties for violation. Nevertheless, while not necessarily an inherent
characteristic, fiscal rules do raise the incentive for policymakers to resort to creative
accounting tactics in order to meet fiscal targets. The above discussion alluded to the fact
that creative accounting can be effectively countered by the design of transparent budget
institutions. The effect of transparency and the relationship between fiscal rules and
transparency is discussed in the following section.

2.2 Fiscal Transparency
A second category of budget institution is the rules determining the degree of
transparency of the budget process. Fiscal transparency refers to the level of openness
regarding fiscal policy intentions, formulation, and implementation (OECD 2002: 7).
Broadly defined, fiscal transparency is “…openness toward the public at large about
government structure and functions, fiscal policy intentions, public sector accounts, and
projections” (Kopits and Craig 1998: 1). Fiscal transparency is widely recognized as an
integral part of effective budgeting. As evidence of this fact, the Organization of
Economic Cooperation and Development (OECD) and the IMF jointly publish the Codes
of Best Practice for Fiscal Transparency manual in an effort to guide countries toward
more transparent budget practices. Nevertheless, there remains a large amount of
variation among countries in the degree to which their citizens are permitted to observe
the inner-workings of government and the basis for policy making (Besly 2006: 203).
It is argued that fiscal transparency is necessary for good democratic governance
because politicians possess informational advantages over voters with regard to their
policy preferences and/or the observability of their actions. For example, policymakers

56
may be myopic, and the threat of being replaced in the next election may induce a
preference for tax rates and expenditure levels and/or deficits that are larger than that
preferred by the public itself (Besley and Smart 2007; Milesi-Ferretti 2003; Persson and
Svensson 1989; Tabellini and Alesina 1990). Furthermore, politicians may differ in their
competence and abilities—say, in the provision of public goods—and thus are
encouraged to bias fiscal policy in an attempt to disguise their true level of performance
(Alt and Lassen 2006b; Gonzalez 2002; Rogoff and Sibert 1988; Rogoff 1990; Shi and
Svensson 2006). The purpose of fiscal transparency, then, is to mitigate the agency
problem that may arise during the formulation and implementation of fiscal policy. In
addition, transparency may soften the common-pool resource problem in so far as
transparency raises public awareness of policymakers’ spending priorities and revenue
sources.
According to the International Monetary Fund (IMF 2007), transparent budget
institutions are defined by the extent to which they satisfy four criteria. First, transparent
budget institutions clearly define the roles and responsibilities of the various policy actors
in the budget process. Identifying all the entities providing public goods and services
gives the public an idea as to the true scope of government. In addition, transparent
budget institutions promote accountability by establishing the roles and responsibilities of
government in the collection and use of public resources. Second, transparent budget
institutions are explicit about the policy goals of government. To that end, transparency
implies an explicit statement of expenditure proposals, as well as the means by which
they will be financed. Similarly, transparent budget institutions present detailed proposals
for revenue collection and borrowing and explain how these efforts assist the government

57
in achieving its policy objectives. Third, such institutions make all fiscal information
available to the public. This information must not only be made available in a timely
fashion, but must also include all the relevant information regarding the fiscal activities
and objectives of the government in such a way that facilitates policy analysis and
promotes responsibility. Finally, an essential feature of transparent budget institutions is
that the fiscal data meet basic criteria that can attest to their quality and integrity.
Typically, this implies mechanisms of both internal and external oversight. In essence,
fiscal transparency is a means of providing the public, financial markets, and politicians
themselves with information regarding the plans that motivate the government’s fiscal
policy, the actual policies implemented, and estimates concerning their long-term effects.
As such, fiscal transparency greatly simplifies the task of attributing fiscal outcomes to
specific policies and politicians (Alt et al. 2006: 31).
A number of scholars have modeled the relationship between transparency and
fiscal policymaking and the possible ways in which transparency may moderate the
agency problem. These models assume that policymakers differ in their preferences
and/or abilities; for example, in their ability to provide public goods (Gonzales 2002) or
their preferences for rent-seeking over public good provision (Besley and Smart 2007;
Persson and Tabellini 2000: Chapter 4; Shi and Svensson 2006). Other differences in
preferences and competence manifest themselves through the manipulation of fiscal
policy instruments; the channeling of expenditures toward rents (Besley and Smart 2007);
or, in an effort to appear competent, policymakers can issue debt (Alt and Lassen 2006)
or lower taxes and increase spending (Shi and Svensson 2006) in order to be reelected.
46

46
That is, in order to appear competent by providing the same level of public goods, incompetent
politicians tend to spend more excessively and/or issue debt.

58
When the preferences and actions of policymakers are difficult to observe, these
self-interested incumbent politicians can issue inordinate amounts of debt, induce socially
sub-optimal budget cycles, or allocate revenues toward rent-seeking. In other words, they
take advantage of the inability of voters to discern their preferences or observe their
actions to manipulate voters’ perceptions and achieve reelection. However, the ability of
voters to discipline policymakers increases with transparency; when transparency is high
voters are able to observe the debt levels and other fiscal abuses and therefore able to
punish profligate incumbents. Transparency thus reveals the competency-level of
politicians (or their “types”), their policy actions, and the fiscal consequences of their
policies. In this way, transparency militates against the informational advantage of
politicians.
Thus, it is possible to identify three potential effects of fiscal transparency. First,
transparency reveals information about the underlying types of politicians by providing
information about their past actions. Second, transparency provides the public with
information regarding the fiscal outcomes of policy. Finally, transparency divulges to the
public the true costs of public spending (Besley 2006: 204). Thus, budget transparency
permits the public to understand the inner goings-on of government as well as policy
outcomes and therefore allows the public to hold policymakers more accountable for their
actions. In this sense, transparent budget institutions reduce the informational advantage
that policymakers have over the public. This way, politicians will be less able to mislead
the public with regard to the fiscal activities of government, and, to the extent that they
do, they are more likely to be discovered and punished for their actions.

59
The question that remains to be answered is: if fiscal transparency severely
reduces the ability of policymakers to exploit their informational advantage over the
public when setting fiscal policy, then why would such politicians voluntarily increase
the level of transparency and thereby limit their policy latitude? As Alesina and Perotti
(1996: 403) state, “politicians typically do not have an incentive to adopt the most
transparent practices.” In fact, if politicians are myopic and/or incompetent and do not
wish to be distinguished from their more capable counterparts, then they may even be
motivated to decrease transparency (Alt et al. 2006: 34). Nevertheless, politicians
sometimes do increase transparency. The question remains as to why.
First, such measures may be the result of outside pressure, as with IMF and World
Bank adjustment programs. Second, transparency augmentation may be a part of a
greater fiscal reform effort that follows a period of fiscal distress. For example, in the
midst of a severe economic crisis in the 1980s New Zealand undertook a massive fiscal
reform effort over the next decade in a successful effort to restore financial stability.
47

Two Acts in particular, the Public Finance Act (1989) and the Fiscal Responsibility Act
(1993), were imposed specifically to enhance fiscal transparency. Together these acts
standardized the method of accounting, fully disclosed fiscal information and objectives,
and mandated that each agency specify the outputs they intended to provide over the
fiscal year and for which they could be held accountable. Combined with other fiscal
reforms, between 1983 and 1994 the improved transparency of the budget contributed to
both lowering the expenditure-to-GDP ratio from 38% to 35% and the deficit-to-GDP
ratio from -9% to a small surplus.

47
This example is taken from Campos and Pradhan (2008: 246-253).

60
More importantly for present purposes are the political motives which prompt
policymakers to choose to increase fiscal transparency. First, as mentioned above, it is
often difficult for “good” or policy-effective politicians to distinguish themselves from
“bad” or opportunistic politicians. Therefore, politicians who perform well or implement
socially optimal policy would prefer that voters be able to observe their actions and
correctly attribute positive fiscal outcomes to their policies. A second explanation posits
that political competition induces one set of policymakers to attempt to tie the hands of
others. The informational advantages afforded by low transparency are only exploitable
by those in power. If one expects to remain in power for the foreseeable future, then
maintaining low transparency will allow one to pursue one’s political objectives.
However, if the probability that a given party will spend a significant amount of time
outside of office, then while in power there is the incentive to increase institutional
transparency in order to prevent one’s opponents, who, if elected, would pursue their own
policy interests. This incentive is even greater where ideological polarization between
political parties is particularly high. While the choice of high transparency ties one’s own
hands as well as one’s opponents, if the probability of being replaced is sufficiently high,
the benefits or restricting opponents will outweigh the costs (Alt and Lassen 2006b; Alt et
al. 2006). Finally, incumbent politicians may also want to tie the hands of other
politicians with whom they share power; for example, in instances of divided or
coalitional government (i.e. the other branches of government or the other coalitional
partners).
48

48
Few studies have attempted to sort through the various explanations for increased fiscal transparency. I
am aware of two studies, both of which offer evidence supporting the “political competition” hypothesis
but less robust evidence to support the “fiscal environment” hypothesis (Alt and Lassen 2006b; Alt et al.

61
Having established a general understanding of the hypothetical effects of fiscal
transparency I now return to the question of the relationship between fiscal rules and
transparency. Recall that while fiscal rules may be effective in mitigating the common-
pool resource problem, it was suggested that such rules may be less successful in
addressing, and may even exacerbate, the agency problem. Specifically, it was argued
that while fiscal rules may place limits on fiscal aggregates, it may induce policymakers
to find and exploit loopholes in order to meet legally mandated targets. It should by now
be obvious that in order to address both problems (at least in part), fiscal rules must be
accompanied by fiscal transparency. This is due to the fact that while fiscal rules may
induce creative accounting, transparency increases the likelihood that such behavior is
discovered and thereby raises the cost of misconduct. Absent a sufficient level of fiscal
transparency, fiscal rules may be ineffectual or quite possibly may even lead to fiscal
outcomes contrary to those intended.
Importantly, there exists a theoretical basis for these claims. Specifically, Milesi-
Ferretti (2003) presents a model of fiscal policy formation within government (where it
sets expenditure and revenue levels) and examines the effect on the budget of two
alternative institutional settings. In the first instance government is free to set policy at its
discretion, and in the second, it must adhere to a fiscal rule which in the model takes the
form of a deficit ceiling. The author shows that on average the government will run
deficits under the discretionary framework, a result that provides a rationale for fiscal
policy rules. The institution of a fiscal rule, however, creates the possibility and the
incentive for creative accounting in order to meet the rule. Whether or not policymakers

2006). The proposition that fiscal transparency is adopted by “good” politicians in order to showcase their
sound performance has received the least amount of attention.

62
will engage in creative accounting however, is a function of the level of fiscal
transparency. When transparency is high, creative accounting is detected by the public
and policymakers are punished for their deception and transgression. The central result is
that as budgetary transparency increases, the amount of creative accounting decreases and
the amount of actual fiscal adjustment (deficit reduction) increases.
In summary, a transparent budget process provides the public with all the
necessary information regarding government functions, fiscal policy objectives, public
sector accounts, and macroeconomic forecasts in a timely and systematic manner.
Transparency is important in fiscal policymaking given the fact that politicians possess
information regarding their preferences and actions that the public may not be able to
discern and/or observe directly. This agency problem allows policymakers to exploit their
informational advantage to further their own interests; for example, politicians may
manipulate fiscal policy instruments in order to be reelected and voters may be left
unaware of the full benefits and costs of policies. The purpose of fiscal transparency is to
dispossess politicians of this advantage. This way, the past actions of policymakers and
their consequences can be observed and politicians can be held accountable for their
policies. Finally, it was argued that fiscal transparency adds to the force of fiscal rules by
increasing the likelihood that creative accounting tactics are discovered. Together, these
budget institutions may be useful in mitigating the negative effects of the common-pool
resource and agency problems inherent to the fiscal policy making process. What remains
to be discussed is the role of the overall rules which regulate the entire budget process
vis-à-vis these problems, to which I now turn.

63
2.3 Procedural Rules
The final budget institution is the set of rules stipulating the process to be
followed in the preparation, submission, and execution of the budget. These procedural
rules include both the formal and informal rules regulating budgetary decisions within the
various branches of government. Specifically, they consist of the rules determining the
formulation of the budget within the executive branch, its submission to, and passage
through, the legislature, and its implementation by the bureaucracy. Procedural rules are
important in determining fiscal outcomes because they specify the relative distribution of
roles and responsibilities among the various actors at each stage of the budget process,
thereby distributing strategic influence over fiscal policy (Dabla-Norris 2010: 4; Von
Hagen 2002). The purpose of procedural rules is to alleviate the strains on public finances
caused by the common-pool resource problem, by making policymakers more aware of
the effects of their fiscal decisions in light of the government’s budget constraints.
However, as argued below, because procedural rules differ in the distribution of the roles
and authority granted to the various actors, they vary in the extent to which they are
successful in reducing the common-pool problem.
The budget process may be divided into three phases.
49
The first is the “executive
planning phase.” This phase begins the year prior to the relevant fiscal year and ends with
the submission of the budget to the legislature. During this stage the overall budget
guidelines are set and the various spending ministries submit their appropriations
proposals. Any conflicts between the spending interests of the various departments are
resolved during this phase. Finally, it is during this period that the revenue budget is
constructed. The second phase is the “legislative approval” phase. In this stage the

49
What follows is adapted from Gollwitzer (2011: 119-123) and Von Hagen (2002: 271).

64
legislature deliberates and votes on the annual budget, including the general budget
policies as well as the specific allocations. Importantly, it is possible that this phase may
include a process of legislative amendments to the budget proposal, which also may
involve more than one chamber of the legislature. The legislative approval stage
concludes with the passage of the budget. The final stage is the “implementation phase.”
This covers the fiscal year to which the budget applies and includes the execution,
management, and reporting of budgetary allocations.
The potential to worsen the common-pool resource problem exists at each stage of
the budget process. At the executive planning stage the CPR problem can be exacerbated
if the overall level of spending is determined residually, after the allocation bids have
been received from each of the spending departments. This is because each department
will seek to maximize its appropriation without consideration of the overall effect on the
budget (Ferejohn and Krehbiel 1987). Likewise, the manner in which conflict is resolved
within the executive is of consequence to the budget. If conflict resolution takes place in
an uncoordinated way it opens the door for log-rolling and reciprocity, with similarly
upward effects on the size of the budget (Von Hagen 2002). At the legislative approval
phase, the CPR problem is of even greater concern given the much larger number of
decision-makers. The primary concern at this stage is the ability of the legislature to
make unlimited amendments to the executive’s proposals. Specifically, the danger is that
the legislature can increase spending and the amount of debt issued (Alesina and Perotti
2008). Lastly, the CPR problem can worsen during the implementation stage in so far as
the government is legally able to deviate from the policies stipulated in the budget,
including the ability to change or implement an entirely new budget during the fiscal year

65
(Von Hagen 2002). One of the more common practices is the use of supplementary
budgets, which allow governments to fund additional expenditures during the fiscal year.
The extent to which the CPR problem is manifested in the budget process is
largely determined by the design of procedural rules. Effectively designed rules force
policymakers to internalize the full set of costs and benefits of public policies financed
through the budget. Conversely, inappropriately designed rules do just the opposite; they
encourage policymakers to focus on the rents that they can accrue and the resources that
they can distribute to their constituencies (Von Hagen 2002).
Scholars distinguish between two types of procedures based on which actors
possess agenda-setting power within the executive branch and between the executive and
the legislature. The first type is hierarchical, or centralized procedures (Alesina and
Perotti 2008; Von Hagen 2006). Under hierarchical procedures during the executive
preparation phase some central budget authority (CBA) is given substantial power over
the various ministries in preparing the budget.
50
Typically, the CBA is the prime minister
or perhaps the finance or treasury minister in the government. Similarly, hierarchical
procedures usually favor the priorities of the executive over the legislature. This
customarily occurs by restricting the legislature’s ability to make amendments to the
budget bill, either in number or kind (or both). The second type of procedural rules is
collegial, or fragmented procedures (Alesina and Perotti 2008; Von Hagen 2006).
Collegial procedures have just the opposite features as hierarchical procedures. These

50
The central budget authority (CBA) is defined as the ministry or government agency that has “the leading
role in maintaining aggregate fiscal discipline, ensuring compliance with the budget laws and enforcing
effective control of budgetary expenditure” (Curristine and Bas 2007: 17). Von Hagen (2002: 273) uses a
similar concept, which he calls a “fiscal entrepreneur.” This is the person to whom the budget process
grants the authority “to set the broad parameters of the budget and to assure that all other participants
observe them. An effective ‘entrepreneur’ has the ability to monitor the other participants and to use
selective punishments against any defectors.”

66
procedures instead grant substantially more agenda-setting power to individual spending
ministers than to the treasury and impose much less restrictive limitations on possible
legislative amendments to the budget.
Given that hierarchical procedures grant agenda-setting power to the CBA within
the executive and to the executive over the legislature in the approval stage, it is argued
that such procedures are less democratic and produce budgets that favor the preferences
of the majority. Conversely, collegial procedures are more democratic because they give
greater voice to the spending ministers within government, favor the prerogatives of the
legislature versus the executive, and grant budgetary rights to the minority opposition
within the legislature (Alesina and Perotti 2008: 16-17). Nevertheless, hierarchical
procedures are argued to be better suited at solving the common-pool resource problem.
This is due to the fact that they grant budgetary power to the actors (e.g. the treasury
minister) that (allegedly) have an incentive to internalize the entire costs and benefits of
policies and limit the autonomy of spending ministers and legislators, which, if allowed
to decide their own spending and debt levels, would fail to consider the potential
externalities (Gleich 2003; Filc and Scartascini 2004). The rationale behind this
argument is twofold: on the one hand, the CBA is able to fully internalize the costs of
expenditure because they set total expenditure and draft the revenue budget. Therefore,
they are fully cognizant of the government budget constraint (Perotti and Kontopoulos
2002: 196; Von Hagen 2002: 273). On the other hand, spending ministers are presumably
more responsive to groups who benefit from specific spending programs whereas the
CBA is relatively more responsive to the interest of the average taxpayer (Von Hagen and
Harden 1995: 774; Alesina and Perotti 1996: 401-402; Hallerberg and Von Hagen 2008:

67
215). Thus, while perhaps less democratic, hierarchical procedures are more conducive to
fiscal discipline.
The theoretical underpinnings of these claims derive from the work on agenda
formation and bargaining in legislatures, particularly the work of Baron (1989; 1991) and
Baron and Ferejohn (1989). These authors discuss two types of voting rules which affect
the types of coalitions that are formed, as well as the size and distribution of benefits
among members of the winning coalition. Under “open rules” proposals made by the
agenda setter can be amended from the floor whereas under “closed rules” the proposal
must be voted up or down without amendments. Consequently, the agenda-setter has
more power under a closed rule procedure. For present purposes, the most interesting
result posits that under a closed rule procedure benefits are distributed among a minimal
majority whereas under an open rule benefits may be more equally distributed among a
larger majority.
51
When interpreted in the context of budget institutions, the theory
suggests that a closed rule voting procedure would result in the approval of a budget in
which the benefits of expenditures are concentrated among a minimal wining coalition.
Open rules, by contrast, would spread the benefits to a wider majority.
Under hierarchical procedures the CBA is the agenda-setter within the
government and the executive branch is the agenda-setter during the legislative approval
stage of the budget process. Hierarchical procedures are thus similar to “closed rules” in
that the expenditure preferences of spending ministers are subordinate to the priorities of
the CBA in the executive preparation phase, and the amendments that can be made by the
legislature during the approval phase are strictly limited. There is a similar

51
Whether benefits under an open rule are distributed among members of a minimal majority or a super
majority depends upon their “impatience;” minimal majorities result when impatience is sufficiently low
(Baron and Ferejohn 1989: 1198).

68
correspondence between “open rules” and collegial procedures (Alesina and Perotti 2008).
Thus, hierarchical procedures grant more power to the government, which is, as argued
above, more likely to internalize the entire costs and benefits of policies, as well as result
in smaller-sized majorities.
Once more a reasonable question to ask is why politicians would voluntarily
restrict their policymaking powers, in this case, by delegating agenda-setting powers to a
central budget authority? One explanation is the so-called “costly insulation hypothesis,”
which attributes voluntary relinquishment of the power to set policy to electoral
competition (de Figueiredo Jr. 2002; 2003). Essentially, electoral competition creates
uncertainty among incumbents who are aware that at some point in the future they will be
out of power. With their opponents in office, the policies they set while in power are
under constant threat of being overturned. As such, the threat of future policy reversals
will induce incumbents to protect their policies from their opponents (Moe 1989; 1995).
This threat is especially acute to electorally weak groups that expect to be out of power
more often than in power. Therefore, policymakers seek protection though the use of
“insulation mechanisms” which trade benefits when in office for benefits when out of
office (de Figueiredo Jr. 2002: 326). Of the various mechanisms with which politicians
can insulate their policies, one of the more pertinent is the assignment of policymaking to
a sympathetic part of the government hierarchy or enhancing its role in the process
(McCubbins, Noll, and Weingast 1987). In so doing, the instituted policies will continue
on their course even after their architects will have left office. In reference to budgeting,
institutionalizing hierarchical procedures thus serves as the insulating mechanism.
Specifically, the insulation hypothesis suggests that when in control of the legislature

69
electorally weak groups have the incentive to shift budgetary authority away from the
legislature and toward the executive so as to limit the ability of future legislative
majorities composed of political opponents the pursue their preferred spending priorities.
A second explanation, derived from Thies (2001), is perhaps more relevant to
parliamentary systems but is nonetheless useful in developing further the intuition behind
the choice of hierarchical procedures.
52
In such systems, multiparty coalition
governments are common. One issue that needs to be resolved after a government has
formed is the partition of cabinet portfolios to each of the coalition members. Cabinet
positions are valuable because each ministry is given the authority to set policy in a
particular area and oversee its implementation. The partitioning of portfolios is
potentially problematic because parties are unable to perfectly monitor and control the
cabinet ministers of other parties. In other words, delegating cabinet portfolios among
coalition partners creates an agency problem. Thus, optimal delegation requires that
parties possess mechanisms that facilitate the detection of whether or not the ministers of
other parties are following the agreed-upon governmental program and they must be able
to discipline aberrant ministers if they are discovered. Thies (2001: 582-583) posits two
such mechanisms: first, the requirement that important actions of ministers be submitted
for prior approval. This implies, however, that each party must able to monitor the
activities of the ministers of the other parties.
53
After all, each minister possesses an
informational advantage about the policy implications of their respective proposals.
Second, parties might establish a set of institutional checks to constrain the ability of each
ministry to independently act or propose policy. For example, policy responsibility could

52
I say “more relevant” rather than “only relevant” because as Cheibub (2007: Chapter 3) has shown,
government coalitions can, and do, occur in presidential systems, albeit less frequently.
53
In the budgeting context this task could be accomplished through fiscal transparency.

70
be divided between two or more agents with divergent preferences. In the budget process,
both of these mechanisms can be accomplished by granting prerogative to a CBA, for
example a treasury minister. Specifically, parties can control and monitor the actions of
each other’s spending minister by requiring that all proposals be submitted first to the
treasury minister.
54
Thus, while delegation to a treasury minister may be considered an
abdication of power, it may be a worthwhile cost to pay in order to keep tabs on one’s
coalition partners.
In conclusion, procedural rules are the set of laws and conventions which specify
the process to be followed in the preparation, submission, and execution of the budget.
Procedural rules are important in determining fiscal outcomes because they set the
distribution of roles and responsibilities among the various actors at each stage of the
budget process, and thus determine who commands strategic influence over fiscal policy.
Scholars distinguish between two types of procedural rules; hierarchical rules which
concentrate agenda-setting power within the executive and tend to produce budgets that
favor the priorities of the majority; and, collegial rules which disperse policymaking
power and emphasize the privilege of spending ministers within the government, the
prerogatives of the legislature in relation to the government, and the rights of the minority
opposition in the legislature. While the latter institution is characteristically more
democratic, in contrast to hierarchical rules, it is also less conducive to fiscal discipline.
Hierarchical rules are better able to attack the common-pool resource problem because

54
Of course, this raises the question of which coalition partner will be given the powerful treasury portfolio.
I imagine that this would either be granted to the formateur party, or would be decided through interparty
bargaining; perhaps given to a strategic party member whose participation in the coalition is necessary in
order to form the government. In addition, there is the issue of how the treasury ministry, once it has been
offered to a party, is given to a specific party member. Here, too, I assume that this decision would be made
through some sort of intraparty bargaining procedure.

71
they grant agenda-setting power to a central budget authority, which is considered to be
less encumbered by individual spending interests and thus in a better position to take a
comprehensive view of the entire budget.

3. Hypotheses
In this section I derive from the above theoretical discussion testable hypotheses
which seek to explicate the relationship between budget institutions and various fiscal
policy outcomes. Here, the emphasis is on four outcomes in particular: deficits, debt,
expenditures, and expenditure composition. My approach represents an advance over
existing studies because these have neglected to fully consider the implications of their
theories for all of these policy outcomes.
55
The discussion in this section is separated
according to each by institution and I hypothesize relationships between each of the
institutions and each of the policy outcomes.
Regarding the fiscal policy outcomes, I follow the convention of referring to
expenditure levels as “size of government,” both for ease of exposition and in line with
the use of this terminology within existing works. Similarly, I follow the standard of
dividing expenditure composition into two broad categories: “public goods” and
“subsidies and transfers.” In total, I derive thirteen hypotheses concerning the
relationship between budget institutions and fiscal policy outcomes, which are considered
for the first time under a single theoretical framework.
Before discussing my hypotheses an important caveat is in order. Throughout the
following discussion I maintain one important and perhaps controversial assumption in
developing these hypotheses: voters are fiscal conservatives. This is controversial

55
For a more complete discussion, see Chapter 2 of this dissertation.

72
because perhaps it is the case that voters are willing to tolerate larger expenditures and
higher taxes if they perceive the benefits of public spending to be worthwhile (see for
example, Ferejohn 1999). Or perhaps voters are fiscally neutral and prefer balance rather
than loose or tight fiscal policy. Conceivably, it may even be that voters’ expectations
change with the party in power, whereby they would expect “liberals” to tax and spend
more and expect “conservatives” to do less of both. Accordingly, voters would be
expected to punish parties that deviate from these left-right expectations (Lowry et al.
1998).
While each of these (and other) arguments is valid, they are contradicted by much
empirical evidence. The landmark study is that conducted by Peltzman (1992) who found
that U.S. voters punished candidates in presidential, senatorial, and gubernatorial
elections for increasing spending and deficits.
56
Similar effects have been found in U.S.
states with regard to tax increases (Kone and Winters 1993; Besley and Case 1995a,b;
Niemi et al. 1995). Internationally, Brender (2003) and Drazen and Eslava (2010) found
that incumbents in local elections in Israel and Colombia, respectively, were less likely to
be reelected if deficits and debt increased during the candidate’s tenure in office.
Furthermore, Brender and Drazen (2008) find that among a large panel of both developed
and developing countries, fiscal expansions do not increase the probability of reelection
(contra political business cycle theory); however, increasing deficits does reduce the
likelihood of reelection. Finally, evidence suggests that the converse of the latter finding
is true, in that incumbents are not voted out of office for fiscal tightening (Alesina, Perotti,
and Tavares 1998; Alesina, Carloni, and Lecce 2011). Given the preponderance of

56
Peltzman (1992: 358) claims that the reason for voters’ fiscal conservatism is that fiscal systems are
typically progressive and voters are generally wealthier than non-voters.

73
evidence and lack of data to the contrary, I maintain the assumption that on average,
voters prefer less spending, less taxes, and lower debts and deficits.
57
Thus, my
presumption that voters are fiscally conservative explains why, for example, when fiscal
agenda-setting power is in the hands of actors that prioritize the interests of the general
taxpayer versus special interests the expectation is for lower spending and smaller deficits.
The following sections develop the hypotheses to be tested in the subsequent
chapters.

3.1. Fiscal Rules
With regard to fiscal rules, the expectation is for fiscal rules that require a
balanced budget, limit the growth of expenditures and revenues, and cap sectoral budgets
and/or debt to constrain the excessive appropriation tendencies of spending ministers,
thus mitigating the common-pool problem (Gollwitzer 2011). State-level evidence on
taxation and expenditure, and (certain) debt limitations suggests that such measures may
be successful (Kieweit and Szakaly 1996; Shadbegian 1998). However, it is possible for
governments faced with fiscal restraints to resort to creative accounting to circumvent
such limits, thereby limiting their effectiveness (Von Hagen 1991). In fact, as we have
seen, the scope for creative accounting is a function of the level of budget transparency.
The implication is that fiscal rules may be effective in partially resolving the common-
pool problem but less so in solving the agency problem; especially if politicians’ actions
cannot be observed and monitored. This suggests that the potency of fiscal rules is
contingent on the level of budgetary transparency. Based on these assumptions, I have
generated the following hypotheses:

57
Keep in mind that voters can still prefer more spending targeted toward them but less overall.

74
If the level of budgetary transparency is sufficiently high,
H
1
: Then fiscal rules requiring a balanced budget should lead to lower and less
persistent deficits.
H
2
: Then fiscal rules that impose ceilings on borrowing should lead to lower levels of
debt.
H
3
: Then fiscal rules that limit the growth of revenues and expenditures and/or impose
sectoral budget ceilings on spending ministries should lead to smaller government.
In contrast, there is no reason to expect numerical targets to have a direct effect on
budget composition. By definition, fiscal rules place constraints on budgetary aggregates.
When enforced, the budget balance and/or expenditure levels are legally predetermined
and policy debates are limited to determining the relative spending priorities (Kopits
2001). Thus, the effect of numerical targets is to focus debate toward the relative
spending priorities, but they have no effect over what those priorities will be. This leads
to the following hypothesis:
H
4
: Fiscal rules will have no direct effect on budget composition.

3.2 Fiscal Transparency
In addition to the interactive effect of transparency and fiscal targets, it is possible
to identify direct effects of transparency on the operative variables of interest. As
discussed above, a lack of budgetary transparency obfuscates the true state of public
finances, creating an agency problem which provides politicians with the incentive to use
their informational advantage to mislead voters as to the true benefits and costs of
expenditures, and/or to overspend on opportunistic goals. Conversely, transparent budget

75
institutions provide the public with a clear sense of the government’s fiscal priorities, the
state of public sector accounts, and revenue and performance projections, thereby
permitting citizens to hold public officials accountable for deviant behavior. When
responsibility is clearly established fiscal restrain is more likely. Thus, the expectation is
for debt levels to be lower under transparent budget institutions since policymakers are
unable to feign competence by issuing debt and face electoral punishment for such
practices. In addition, when unable and/or unwilling to excessively borrow a government
is less likely to run deficits and expenditures are likely to be lower than they would be
otherwise under a non-transparent budget institution. It follows that:
H
5
: If the degree of budget transparency is high, then the smaller the deficit.
H
6
: If the degree of budget transparency is high, the lower the level of debt.
H
7
: If the degree of budget transparency is high, the smaller the size of government.
As argued above, the composition of the budget is largely determined by the
spending priorities of the various policymakers involved in preparing and approving the
budget and the decision-making rights and privileges with which they are conferred by
the budget institution’s procedural rules. Therefore, ceteris paribus, increasing
(decreasing) budget transparency should not change ones preference from favoring more
public goods provision to preferring more subsidies and transfers, and vice versa. Thus,
we have the following hypothesis:
H
8
: The degree of transparency will have no direct effect on budget composition.

76
3.3 Procedural Rules
Recall that the purpose of procedural rules is to mitigate the common-pool
resource problem; however, different rules are more or less effective in this regard. The
theoretical discussion of budget procedures distinguished between two types of rules:
hierarchical and collegial rules. Hierarchical rules were shown to concentrate agenda-
setting power in fiscal policy whereas collegial rules dispersed such power. Specifically,
hierarchical rules concentrated agenda-setting power among actors with a self-interest in
maintaining fiscal discipline whereas collegial rules dispersed power among actors,
which, it was hypothesized, were less concerned with achieving this goal. In addition, in
the theoretical section I drew analogies between hierarchical procedures and closed
voting rules and between collegial procedures and open voting rules. The implication of
that discussion was that hierarchical (closed) rules distribute benefits among a minimal-
sized majority whereas collegial (open) rules distribute benefits more equally among a
larger majority. The general conclusion to be deduced is that since power and benefits are
concentrated among a smaller, more fiscally responsible group of policymakers,
hierarchical rules should lead to greater fiscal restraint.
In addition, this theory leads us to expect the spending priorities of the
executive/CBA to prevail under hierarchical procedures and those of the spending
ministers/legislature to hold sway under collegial rules. As mentioned above, the CBA
and executive (in theory) are more likely to be concerned with the interests of the average
taxpayer whereas spending ministers and legislators would be more responsive to groups
who benefit from specific spending programs. Since spending on subsidies and transfers
can be targeted to special interests and specific socioeconomic and/or demographic

77
groups, spending ministers and legislators are more likely to prefer budgets that spend
more on subsidies and transfers so as to maximize political support. This is in contrast to
public goods, the benefits of which are broadly dispersed and do not favor any particular
narrow constituency (Mukherjee 2003: 706-707). The expectation, therefore, is for
hierarchical rules to produce budgets with a larger share of spending allocated to public
goods and for budgets produced by collegial procedures to result in a larger share of
spending disbursed on transfers and subsidies. My hypotheses on procedural rules are the
following:
H
9
: The more hierarchical are budget procedures, the lower and less persistent will be
deficits.
H
10
: The more hierarchical are budget procedures, the lower will be levels of debt.
H
11
: The more hierarchical are budget procedures, the smaller the size of government.
H
12
: The more hierarchical are budget procedures, the more will be spent on public
goods.
H
13
: The more collegial are budget procedures, the more will be spent on subsidies and
transfers.
Table 3.1
Budget Institutions and Fiscal Policy: Theoretical Predictions
Budget Institution
Fiscal Rules Transparency Procedural Rules (HR/CR)
Fiscal Outcome
Government Deficit L L L/H
Budget Deficits L L L/H
Government Size L L L/H
Budget Composition N/A N/A (+PG/-ST)/(+ST/-PG)

78
Table 3.1 presents a summary of the theoretical expectations discussed above.
Each row represents the hypothesized relationship between each of the three budget
institution categories from the columns and the fiscal policy outcome of interest on the
far left of the table. An “L” indicates that the associated budget institution category in the
column in expected to lead to a smaller degree or lower level of the outcome in the
adjoined row and an “H” indicates a higher degree or higher level of the outcome. In the
“procedural rules” column, “HR” signifies “hierarchical rules” whereas “CR” denotes
“collegial rules.” The hypothesized effects of hierarchical rules are listed first, and the
effect of collegial rules, second, in each of the fiscal outcome cells. The abbreviations
“PG” and “ST” stand for “public goods” and “subsidies and transfers” respectively. A
plus (minus) sign next to a given expenditure category indicates that the budget
institution is expected to lead to more (less) spending in this category. Finally, “N/A”
signifies that there is no expected relationship between the budget institution and the
fiscal policy outcome.

4. Conclusion
This chapter began with an overview of how fiscal policy making is beset by the
agency and common-pool problems. The extent to which these two problems are
operable largely determines the shape of fiscal policy. When these problems are prevalent,
fiscal policy is usually imprudent and economically sub-optimal. It was argued that
properly-designed budget institutions---fiscal rules, transparency, and hierarchical
procedural rules, in particular---are each capable of mitigating these problems and thus
likely to lead to sound fiscal policies.

79
Specifically, fiscal rules confront the common-pool resource problem by limiting
the extent to which politicians and policymakers are able to exploit fiscal policy
instruments in pursuit of their own goals. Transparency reduces the common-pool
resource and agency problems by lowering the informational advantage that
policymakers possess and thereby raise public awareness of untoward policies and
actions. Procedural rules, particularly hierarchical procedures, limit the common-pool
problem by placing policy-setting power in the hands of those with an incentive to act in
the interest of the general taxpayer and who are in a position to understand how policy
decisions can exacerbate the government’s budget constraint. Finally, it was argued that
policymakers may seemingly belie their own self-interest and adopt these institutions in
order to tie the hands of future policymakers and/or insulate their own preferred policies.
From these theories I derived a series of hypotheses relating each of the budget
institutions to four fiscal policy outcomes.
In the next chapter, I begin to test the theory against the data. In particular, I
develop measures of budget institutions using the OECD and World Bank’s International
Budget Practices and Procedures Database. The database covers the entire budget cycle
for a total of 97 countries from all regions and at all levels of development. As such, it
constitutes one of the most comprehensive comparisons of national budgeting procedures
available and lays the foundation for all subsequent empirical work.

80
CHAPTER 4: RESEARCH DESIGN
1. Introduction
The purpose of this dissertation is to answer the question, “what are the affects of
political institutions on fiscal policy outcomes, and, in particular, what are the affects of
budget institutions?” Chapter 2 discussed existing answers to the first half of this
question. Particular attention was paid to the influence of the bureaucracy, the various
branches of government, political parties, electoral systems, and levels of government.
Nevertheless, these explanations were shown to be deficient and budget institutions were
argued to be more appropriate units of analysis. Chapter 3 explicated the theorized
relationship between the various budget institutions and several fiscal policy outcomes.
The theoretical relationships were then distilled into thirteen falsifiable hypotheses.
The task of this chapter is to explain how these hypotheses are to be tested and
address how the results answer the specific research question. Explaining how the
hypotheses are to be tested, however, requires first a discussion of the available data and
sample size. Only once these issues have been address can an appropriate method of
analysis be selected.
This chapter is organized in the following manner. In Section 2 I introduce the
data and measures used for each variable. The section begins with the budget institutions
variables, discussing the sources of data and explaining how indicators for each
institution will be constructed. Subsequent subsections present the measures and data
sources of the control variables. In particular, I introduce and briefly explain the theory
behind the economic controls argued to influence fiscal policy and discuss the measures

81
of the political-institutional control variables that were put forward in Chapter 2. Then,
the sources and measures of the dependent variables are discussed.
Section 3 discusses case selection. An underlying assumption of the hypotheses in
Chapter 3 was that budget institutions and processes were embedded within an overall
framework of democratic institutions. Thus, the democratic countries in the sample need
to be separated from the non-democratic ones. This section discusses the pros- and cons-
of the four measures of democracy used to sample countries and justifies the use of each
based on the three objectives: 1) maximizing the number of observations, 2) producing
comparable results, and 3) testing the sensitivity of subsequent analyses to the particular
definition and measure of democracy. This last objective is especially important since
research shows that results may vary according to the measure utilized and since there is
no consensus regarding which indicator is to be preferred over others.
One problem that plagues all data and measures used in this study is missing data.
Section 4 discusses the causes and consequence of missing data in general and with
regard to this study in particular. This section also discusses multiple imputation as a
possible solution to the problem of missing data and how it may be applied to this study.
Section 5 discusses method of analysis. Here it is argued that given the specific
research questions to be answered, the nature of the available data, and the size of the
sample, that econometric analysis is preferred method of analysis. It then proceeds to
introduce the series of nested econometric models to be tested and discusses how such
analyses will move us one step closer toward answering how political institutions,
particularly budget institutions, affect fiscal policy outcomes. This section also discusses

82
the limitations of the cross-sectional approach as well as the feasibility of alternative,
longitudinal, methods of analysis. Section 6 concludes.

2. Data and Measures
58

2.1 Independent Variables
2.1.1 Budget Institutions
Under ideal circumstances, budget institutions would be directly observable. In
other words, it would be preferable to be able to determine, according to some metric, the
degree of transparency of a given budget institution, for example, and/or the level of
hierarchy or collegiality of the procedural rules. Unfortunately, no such set of metrics
exist. Consequently, the standard practice has been to develop indicators of budget
institutions based on responses to surveys sent out to a variety of experts involved in the
budgeting process in a given set of countries.
This study continues this practice and constructs measures of the three budget
institutions based on information from two international surveys. The first is from the
International Budget Practices and Procedures Database (OECD n.d.). The database
contains joint OECD and World Bank survey results of budget practices and procedures
conducted between 2007 and 2008. The respondents were the identified budget officials
(Central Budget Authorities – CBAs) in each participating country. The database covers a
total of 97 countries, including 33 OECD members and 64 non-members from Asia,
Africa, Eastern Europe, Latin America and the Caribbean, and the Middle East.
59

Supplementing the OECD survey is the International Budget Partnership’s (IBP) Open

58
Full descriptions of each variable as well as their sources can also be found in the data appendix. All
observations are averaged over the period 1997-2007 (or the sub-period for which data are available).
59
The actual sample of countries to be used in the subsequent analyses is discussed in Section 3.

83
Budget Survey conducted in 2008.
60
The survey provides comprehensive data pertaining
to the level of transparency of the budget process in 85 countries based on questionnaires
conducted by civil society partners of the IBP’s parent NGO, the Center on Budget and
Policy Priorities within each country.
61
Overall, the surveys contains 99 and 123
questions, respectively, that cover the entire budget cycle, from the budget formulation
stage within the Executive, to the budget approval phase within the Legislature, on to the
execution of the budget by the bureaucracy. As such, it facilitates a comprehensive cross-
sectional comparison of national budgeting procedures. In addition, the use of
questionnaires has advantages over other means of collecting data on budget institutions,
such as relying solely on written legislation. For instance, in this case the survey will
permit the evaluation of budget “practices,” rather than just the letter of the law.
Furthermore, the use of questionnaires allows for a greater amount of information to be
collected than otherwise, such as the case in which information is gathered independently
(Alesina et al. 1999: 258 fn. 8).
62

In what follows I describe each item selected from the surveys and discuss how
each component is to be scored.
The first category of items endeavors to capture the existence, depth, and scope of
various fiscal rules. The initial item inquires as to whether there are any rules limiting
fiscal policy. Specifically, it asks if any and/or all of the following rules exist: an

60
The IBP has published surveys in 2006, 2008, 2010, and 2012. Only information from the 2008 survey is
used in this study because the others do not coincide with the OECD survey and/or (as in 2006) contain a
smaller sample.
61
The two surveys cover only 50 of the same countries (≈ 52%). This creates a missing data problem,
especially for the indicators of transparency. How missing data will be dealt with in this study is discussed
in Section 4.
62
With the exception of Fiscal Transparency, the primary source of data on the indicators of the budget
institutions is the OECD database. Thus, unless otherwise noted, the source of the variables discussed in
this section is from the OECD survey.

84
expenditure rule, a revenue rule, a budget balance (surplus/deficit) rule, and/or a debt rule.
Each of these items receives a score of 1 if the rule exists, and 0 otherwise.
The next set of items asks about the coverage of each of the fiscal rules in place.
The second item indicates the components of the general government that are covered by
the rule: the entire government sector, the central government, regional government, local
government, extra-budgetary funds, or other. For reasons discussed below, all the fiscal
policy outcome indicators are measured at the level of the central government. Therefore,
it is only meaningful to examine fiscal rules that apply to this level. Thus, countries with
rules that cover either the entire government sector or the central government receive a
score of 1. Rules that only cover the other levels of government receive a score of 0. The
next item inquires about the institutional definition of the rule. The highest possible score
of 3 is assigned to countries whose rule is defined within the constitution. This is because
constitutions are typically difficult to amend and thus constitutionally defined rules are
(presumably) the most binding. Following this logic, rules defined by legislation receive
a score of 2, those defined by a political commitment of the government or by a formal
agreement between parties in government receive a score of 1, and where the rule is not
explicitly defined it is scored a 0.
Of course, for rules to be effective there must be some semblance of oversight.
The next item asks about the body or bodies in charge of monitoring compliance with the
fiscal rule. The budget officials were permitted to check all that apply and countries
receive a score of 1 for each selected for a possible score between 0 and 5. The options
are the Central Budget Authority (CBA), the Supreme Audit Institution (SAI), the

85
Legislature (or some legislative body), an independent body, and/or some other body.
63

Here, the assumption is that the more actors involved in the monitoring process the
stronger the fiscal rule. The final item asks about the enforcement procedures in instances
of non-compliance with the fiscal rule. Once more, countries receive a score of 1 for each
option selected, for a range of possible scores between 0 and 6. Possible responses
include a proposal of corrective measures sent to the legislature, corrective measures
implemented by the responsible government or ministry, the existence of an automatic
correction mechanism, the possibility of sanctions, automatic sanctions, and/or other
measures.
Naturally, ex ante fiscal rules will be ineffective if the government has recourse to
side-step such restrictions during the budget execution. Two final items ask whether the
government has the authority to increase spending once the budget has been approved by
the legislature. The first pertains to increases in mandatory spending, the second, to
discretionary spending. Scores of 0 are assigned to each of these items if increases in
spending are possible without restrictions. If the government may increase spending but
is restricted in its scope for doing so, it receives a score of 1. Instances in which spending
increases after budget approval are not allowed are scored a 2.
The second category of items is intended to capture the degree of hierarchy of the
procedural rules. Specifically, the items seek to measure two aspects: first, the agenda-
setting power of the CBA relative to the individual spending ministries and the executive
vis-à-vis the legislature. Secondly, the extent to which the fiscal prerogatives of the CBA
and Executive are given precedence over the ministries and the Legislature, respectively.

63
According to the OECD (2006: 7), the Supreme Audit Institution is “The public body of a State
which…exercises by virtue of law, the highest public auditing function of that State.” For a definition of
CBA see Chapter 2 footnote 24.

86
The first item signifies the legal vesting of power within the executive branch. Countries
with a single CBA responsible for managing the budget (e.g. the Ministry of Finance)
receive a score of 1 and countries where several ministries or government bodies share
responsibility for managing the budget receive a score of 0. The second item serves as an
indicator of the overall standing of the CBA relative to line ministries. It asks how
disputes between the ministries and the CBA in the budget preparation process are
resolved. Countries are assigned a score of 2 if the CBA, Prime Minister, or President has
the final word regarding the resolution of conflicts, a score of 1 if the issue is resolved by
the cabinet, and a score of 0 if the issue is resolved by a ministerial committee. The third
item inquires as to whether or not the CBA imposes limits (ceilings) for each ministry’s
initial spending requests. Such limits suggest that the CBA has prerogative in drafting the
budget. A country receives a score of 0 when no limits are imposed. A score of 1 is
assigned in cases in which limits are only “suggested” but not imposed. In cases where
limits are imposed for some types of expenditures, but not all, a score of 2 is given to
limits at the chapter level and a score of 3 is assigned to limits at the line item level.
64

Where limits are imposed for all types of expenditures, scores of 4 and 5 are given to
limits at the chapter and line item levels, respectively.
The remaining items pertain to the procedures between the Executive and the
Legislature during and after the budget’s approval. The first item is meant to capture the
overall position of the Executive in relation to the Legislature. It does so by inquiring
about how the Legislature’s own budget is prepared. Extremely hierarchical procedures
are those in which the Legislature is subject to the same procedures and policies as all

64
Limits at the line item level are scored higher because line items are the most detailed level of
appropriations.

87
other governmental organizations included in the government’s proposal. In other words,
hierarchical procedures do not allow for the Legislature to prepare its own budget, either
independently or unchanged as part of the government’s budget proposal. Countries in
the former category receive a score of 1, and those in the latter category receive a score of
0. In a similar vein, the next item asks whether the Legislature first votes on the total
amount of expenditure before it votes on specific appropriations. In line with the theory
outlined in Chapter 3, countries with hierarchical procedures will vote first on the total
level of expenditures. If this is the case, then a score of 1 is assigned. If not, a score of 0
is given. The third question solicits information about the formal powers of the
Legislature to amend the Executive’s budget proposal. Countries where the Legislature
possesses unlimited powers to amend the budget are scored 0. A score of 1 is given to
cases in which the Legislature may make amendments, but only if it does not change the
total deficit/surplus proposed by the executive. Where the Legislature may only decrease
existing expenditures and/or revenues (i.e. the Legislature cannot increase existing items
nor create new ones) a score of 2 is assigned. Finally, scores of 3 are assigned to cases in
which the Legislature may not make any changes but only vote the budget as a whole up
or down.
Conversely, more hierarchical procedures permit the Executive to veto the budget
approved by the legislature if it deviates too far from its ideal point. If a country’s
Executive is granted this authority at both the line item and package level a score of 3 is
given. If veto power exists but only at the chapter or only at the line item level, a score of
1 and 2 is assigned, respectively. On the other hand, countries where the Executive is
prohibited from vetoing an approved budget are scored 0. The final item inquires about

88
the consequences if the Legislature fails to approve the budget prior to the start of the
fiscal year. Here, the highest score of 3 is ascribed to countries where the Executive’s
budget proposal takes effect. Scores of 2 and 1 are assigned to cases where the
Executive’s budget proposal takes effect on an interim basis, and where last year’s budget
takes effect on an interim basis, respectively. The lowest score of 0 is reserved for those
instances in which other interim measures are voted on by the Legislature or cases where
expenditure without prior Legislative approval is prohibited. This is due to the fact that
the weaker the position of the government on this issue, the greater incentive there is for
the Executive to propose a larger-than-preferable budget in order to ensure approval
(Alesina et al. 1999).
The final category of items is intended to tap the overall degree of budget
transparency. To that end, the initial two items seek to measure how clearly defined are
the roles and responsibilities of the various policy actors in the budget process. The first
item inquires about the legal basis for the roles and responsibilities of different parts of
the Executive in budget formulation and execution. The second item asks about the legal
basis for roles and responsibilities of the Legislature and the Executive in the budget
process. Possible responses to each item include the constitution, legislation, internal
rules, or no formal basis, and the budget directors were instructed to check options all
that apply. When there is no formal legal basis determining the roles and responsibilities
of the budget actors, the country receives a score of 0. Otherwise, a country receives a
score equal to the number of areas in which the basis is defined. Thus, if the roles and
responsibilities are defined in the constitution, as well as in legislation and internal rules,
a score of 3 is assigned, and so on.

89
The next item pertains to the budget’s explicitness with regard to the policy goals
of the government. Countries are asked to list the number of the following elements
which are included as part of the budget presented by the central government to the
Legislature: the fiscal policy objectives for the medium-term, the budget priorities,
financial plan that encompasses all revenues and expenditures including a) off-budget
expenditures and extra budgetary funds and b) all levels of government, medium and/or
long-term perspective on total revenue and expenditure, clearly defined appropriations to
be voted by the legislature and a linkage of appropriations to administrative units, and the
text of legislation for policies proposed within the budget. Depending upon the number of
elements selected, possible scores range between 0 and 9, where higher score indicates
more budget transparency.
The succeeding set of criteria evaluates the public availability of fiscal
information contained in the budget.
65
The first item relates to the publication of the
Executive’s budget proposal (draft budget). Publication of the draft budget is required if
the public is to follow and fully understand the parliamentary discussion of the budget
(Gollwitzer 2011). Scores range between 0 and 2, depending on how widely disseminated
and easily accessible is the document. Proposals that are not produced, or are only
produced for internal purposes receive a score of 0. Countries receiving a score of 1
produce the draft budget and make it available to the public, however only upon request.
Countries that produce the Executive’s proposal and distribute the document to the public
(e.g. through libraries, the internet, etc.) are assigned a score of 2. The second item asks
whether the budget proposal presents the macroeconomic forecasts upon which the
budget projections are based and makes explicit the key economic assumptions (e.g.

65
The following items are drawn from the International Budget Partnership’s Open Budget Survey.

90
inflation, GDP growth, etc.). If this information is not included the country receives a
score of 0. Scores of 1 are given to those countries whose budgets present some
discussion of the macroeconomic forecast and assumptions, but important details are
excluded. Scores of 2 are assigned in instances in which the macroeconomic forecast is
discussed and most of the key assumptions are stated explicitly, however some details
remain lacking. Lastly, countries are assigned a score of 3 if the budget proposal includes
an extensive discussion of the macroeconomic forecast and all key assumptions are stated
explicitly.
Next, the final items assess the extent of budgetary oversight. Two items assess
external budget oversight in terms of the frequency and amount of information contained
in in-year (e.g. quarterly reports), and year-end reports. Transparent In-year reports detail
the actual level of all expenditures consumed theretofore, and transparent year-end
reports detail and explain the differences between all enacted and all actual expenditure
levels. For in-year reports, the lowest score of 0 is reserved for countries that do not
produce such documents. A score of 1 is assigned to those countries whose reports are
released at least twice a year but only cover less than two-thirds of expenditures. A score
of 2 is given to countries if the reports are bi-annual and cover at least two-thirds of
expenditures. A score of 3 is allocated to cases in which reports are released at least every
quarter and two-thirds of expenditures are covered. The highest score of 4 is reserved for
instances in which in-year reports are released at least every quarter and all expenditures
are covered. Regarding year-end reports, a 0 score is reserved for cases that either do not
produce such reports or neglect to explain the differences in expenditures. Scores of 1 are
assigned to countries whose reports provide some explanation but lack important details.

91
A score of 2 is given if an explanation is present and key differences are highlighted yet
some details remain excluded. Lastly, countries that produce reports that provide an
extensive explanation of the differences receive a score of 3.
Finally, three items asses the internal audit mechanisms of budgets. The first asks
the proportion of line ministries subject to internal audits. Possible scores range between
0 and 5 and each score differs from the previous one by auditing an addition 20% of line
ministries.
66
The second item asks whether an audit report is produced and inquires as to
the scope of its availability. Once more, scores range from 0 to 2, based on the
document’s degree of dissemination and ease of access. Proposals that are not produced,
or are only produced for internal purposes receive a score of 0. Countries receiving a
score of 1 produce the audit report and make it available to the public only upon request.
Countries that produce the audit report and distribute the document to the public (e.g.
though libraries, the internet, etc.) are assigned a score of 2. The third, and final,
transparency item asks if the audit reports produced by the SAI are scrutinized by the
Legislature (committee). Scores range between 0 and 3, depending upon whether the
reports are not scrutinized, some reports are scrutinized, most are scrutinized, or all audit
reports are scrutinized. Higher scores indicate greater transparency.

2.1.1.1 Index Construction
Once these items have been selected and scores assigned to each country based on
their respective responses, the next task is to aggregate the items into a single composite
measure of each budget institution. Aggregation into a composite measure is preferable to

66
For example, a score of 2 is given if 20%-40% of line ministries are subject to audit and a score of 3 is
given if 41% to 60% of line ministries are subject to audit. The source of this variable is the OECD.

92
the use of each item individually for a number of reasons. First, most concepts, such as
budget institutions, do not have clear and unambiguous single indicators. Second, single
indicators are likely to be invalid and/or unreliable for many respondents, whereas
composite measures can overcome this problem. Furthermore, there may not be enough
variation along a single indicator (especially if the indicator is categorical) to detect
statistical relationships. Lastly, using a number of indicators allows for a more
comprehensive and accurate indication of a concept compared to a single indicator.
Importantly, when these indicators are summarized by a composite measure oftentimes
the measure maintains the details of the individual indicators (Babbie 2010: 161).
A natural candidate for developing composite measures is factor analysis. The
aim of factor analysis is to explain relationships among a number of variables in terms of
much simpler relations (Cattell 1965). It can reveal, for example, the existence and scope
of any general influence underlying the set variables being examined. In addition, factor
analysis may uncover the presence and structure of a more restricted set of common
factors underlying groups of items or variables (Walkey and Welch 2010). Specifically,
factor analysis can determine whether the items thought to be indicators of the various
budget institutions are in fact related to one another though some underlying and
unobservable factors, where these latent variables are interpreted as measures of the
budget institutions themselves. It would then be possible to extract estimates of these
factors (called “factor scores”) which could then be used as explanatory variables in
subsequent statistical analyses.
However, the nature of the data presents a number of challenges for performing
factor analysis. First, one of the underlying assumptions of traditional factor analysis is

93
that the level of measurement of the data is interval or ratio (Kim and Mueller 1978a).
Yet, while it is certainly true that methods exist for performing factor analysis on
dichotomous and ordinal variables, the interpretation of the resulting factors is not as
straightforward as is the case with continuous variables (Kim and Mueller 1978b;
Gorsuch 1983).
67

In addition, there is a second, and much larger, problem of performing factor
analysis with variables that contain missing observations. For instance, factor analysis
assumes a sufficient number of cases. Most statistical procedures eliminate cases with
missing data and it may be the case that there is not enough data to run the analysis, or, if
it can be run, the results may not be statistically significant as a result of the small amount
of input data.
Perhaps more worrisome, is that it has been shown that performing factor analysis
in the presence of missing data leads to grossly distorted factor structures and thus
misleading results. The severity of factor structure distortion is largely determined by the
amount of missing observations and the pattern of missingness (Mackelprang 1970;
Rummel 1970).
68
In fact, Mackelprang (1970) estimates that distorted factor structures
result when the level of missing data is as low as five percent. This is problematic since
the level of missing data ranges between 3% and 48% among the budget institutions
questionnaire responses.

67
See Muthen (1978), Mislevy (1986), and Joreskog and Moustaki (2001) for discussions of these methods.
68
A more detailed discussion on the effects of missing data can be found in Section 4.

94
One possible solution to this problem would be to impute for the missing data.
However, multiple imputation in Stata® does not work with factor analysis.
69
Truxillo
(2005) has developed an approach to imputing for missing observations and performing
factor analysis using maximum likelihood (ML) with the expectation-maximization (EM)
algorithm. However, the EM algorithm assumes that the data are multivariate normal and
tests for normality based on skewness and kurtosis reveal that the indicators are not
normally distributed. Consequently, because the input variables are categorical and
because there are missing observations on these variables, factor analysis cannot (and
ought not) be performed.
70
As a result, an alternative aggregation method must be
sought.
71

A simple, yet effective, aggregation method is the summation of weighted
individual indicators (OECD 2008). In fact, this method is the most widely used
aggregation technique found in the literature (See Von Hagen and Harden 1996; Alesina

69
I focus on Stata® because it is the statistical package with which I am most familiar and the program
with which all statistical analyses will be performed. Regarding this particular issue, Stata’s mi
estimate or mim commands do not allow one to subsequently run the factor command.
70
Specifically, standard methods of performing factor analysis are based on a matrix of Pearson’s
correlations, which assumes that the variables are continuous and multivariate normal. Factor analysis with
categorical variables can be performed in Stata® using a polychoric matrix of correlations. However, when
data are missing, the Expectation Maximization algorithm and the specific imputation method (mi
impute mvn) assumes the data are multivariate normal. To verify that the assumption of multivariate
normality was in fact meaningful, I applied Truxillo’s approach using both the standard and polychoric
methods, setting aside considerations of whether doing so was statistically valid. The results showed that
there were in fact significant differences in factor structures between the two procedures. For this reason I
believe it is inadvisable to use factor analysis in this particular instance.
71
A number of researchers have noted that factor analysis is essentially equivalent to item-response theory
(IRT) models (Takane and de Leeuw 1987; Reckase 1997; Treier and Jackman 2008). Once it became clear
that factor analysis would no longer be an option, I also tried to estimate an ordinal item response model. In
particular, I tried the Mokken model, however preliminary examination of the items comprising each
budget institution showed that they did not form a required Guttman scale (van Schuur 2003). In addition,
since some of the items are multi-category and not dichotomous, I was unable to estimate a standard Rasch
model as well (Bond and Fox 2012). Finally, other IRT models, such as the Partial Credit Model (Masters
1982), were not attempted due to the difficulty of estimating such models in Stata® (See Rabe-Hesketh et
al. 2004: Chapter 8 and Hardouin 2007).

95
et al. 1999; Gleich 2003; Filc and Scartascini 2004; Wehner 2006; Dabla-Norris et al.
2010; Gollwitzer 2011). This type of index takes the following form:
∑
=
=
n
i
j
i i j
c w I
1
(1)
where the c
i
are the values of the i = {1,…,n} different components (items) of the index
and the w
i
are the weights attached to each component. The superscript j is a power term
indicating the degree of substitutability among items (Wehner 2006). When j = 1, the
items are simply added together under the assumption that the items are perfect
substitutes for one another. Perfect substitution, or compensability, means that low scores
on some criteria can be offset by large scores on another (OECD 2008). In other words,
the index will not differentiate between countries that possess intermediate scores on all
items and those that have high scores in some components and low scores in others
(Dabla-Norris et al. 2010). If 0 < j < 1, then the items possess low substitutability. Under
these circumstances, countries with consistently intermediate scores in all categories will
be ranked higher than those with a mixture of high and low scores. When j > 1, the
opposite is true; a high degree of substitutability is assumed and high scores will be
rewarded (Alesina et al. 1999).
72

Altering the values of j will allow for the testing of the robustness of the index
results. Following convention, I will use the simple sum index of j = 1 and indices with

72
If the assumption were that the components were perfectly non-substitutable then a multiplicative
aggregation method would be warranted. This method can be expressed in the following formula:
∏
=
n
i
i
c G . Nevertheless, this method has a number of undesirable properties. First, this form of
aggregation usually results in highly skewed distributions since low scores on items drag down the index.
As a result this method is especially sensitive to specification error since miss-specified items would result
in a larger error in the overall score (Dabla-Norris et al. 2010). What is more, since many of the cases here
have values of zero on at least one of the items, this method would not produce informative results. Lastly,
it seems implausible to assume non-substitutability among all components (Wehner 2006). For these
reasons, it is inadvisable to use multiplicative aggregation.

96
arbitrary values of 0.5 and 2 for the power term (i.e. half and double the value of the
simple sum version).
73
This will permit the consideration of the impact of different
substitutability assumptions.
Regarding w
i
, I, like most of the preceding literature, am agnostic about the
importance of one item in each index over another. Thus, I set w
i
= 1 for all items. Under
this assumption, all items in a given index are of equal importance. The only exception to
this practice with which I am aware is the study by Dabla-Norris et al. (2010). These
authors argue that based on “expert assessment” certain categories, especially “rules and
controls,” are more critical than other categories (Dabla-Norris 2010: 53).
74

Unfortunately, I do not have access to these expert assessments and theory does not
suggest a clear importance ordering. In any case, their analyses reveal a statistically
significant correlation (r > 0.95, p < 0.01) between the un-weighted and weighted indices,
leading the authors to conclude that their original, and more parsimonious, additive
aggregation procedure was appropriate. Thus, an item weight of 1 is likely an innocuous
assumption.
Finally, it is worth noting the way in which the index for fiscal rules differs from
those of fiscal transparency and procedural rules. Recall from Chapter 3 that hypotheses 1
though 3 refer to particular rules, i.e. balanced budget (deficit), borrowing (debt), and
expenditure rules and restrictions. As such, it makes little sense to construct a composite

73
This is the approach taken by Wehner (2006). Alesina et al. (1999) and Dabla-Norris et al. (2010) choose
0.4 and 2 as alternate values for j. Alesina et al. (1999) justify this choice by reasoning that it is plausible
that the “true model” of how the different items interact fall within this range. Values higher than 2 and
lower than 0.4, they show, result in highly unreasonable scoring schemes.
74
Interestingly, these authors, as well as Gollwitzer (2011) conflate fiscal rules and procedural rules,
organizing them into a single category of “Rules and Controls,” in the case of the former and “Fiscal and
Procedural Rules and Controls,” in the case of the latter.

97
measure comprised of all the fiscal rules mentioned above.
75
Instead, a separate index is
appropriate for each rule. Each of these indices will be comprised of the items denoting a)
the level of government covered by the rule, b) the institutional definition of the rule, c)
the actors in charge of monitoring the rule, and d) the enforcement mechanisms in place,
as discussed above.
76
Furthermore, it was argued that a necessary condition for
hypotheses 1 through 3 is that fiscal transparency is sufficiently high. Thus, it is these
individual indices, rather than their amalgamation, that are to be conditioned on the
transparency index.

2.1.2 Control Variables
2.1.2.1 Economic Control Variables
Chapter 2 made mention of the existence of potential economic determinants of
fiscal policy outcomes. However, their discussion was postponed in order to give
precedence to political-institutional explanations. In this section we introduce and briefly
explain these possible economic influences as well as how they are to be measured and
controlled for.
Demographics: The demographic composition of the population may be related to
fiscal outcomes. First, the total population must be taken into account. A number of
authors have posited that country size affects economies of scale in the provision of
public goods and services and thus larger countries may be associated with smaller

75
For example, hypothesis 2 states, “fiscal rules that impose ceilings on borrowing should lead to lower
levels of debt.” Thus, it is the relationship between a debt rule and levels of debt that we want to test and
not the affects of other rules on debt since it would not be a fair test and it is likely to lead to misleading
results.
76
The expenditure index will also include the mandatory spending and discretionary spending items
mentioned above.

98
governments and (possibly) lower deficit levels (Alesina and Wacziarg 1998; Hallerberg
et al. 2007). Therefore, the analyses include the natural log of the total population
(LPOPTOT) as part of the vector of controls. In addition, the age structure of the
population may affect fiscal outcomes. For instance, the larger the proportion of the
population that is retired the more political pressure there may be to spend and
redistribute income (Wilensky 1976: 47; Persson and Tabellini 2003b: 39).
77
The age
structure of the population is measured as both the proportion of the population between
the ages of 15 and 64 (POP1564) and the proportion of the population over the age of 65
(POP65). The source of these variables is the World Bank’s Word Development
Indicators database.
National Income: According to “Wagner’s Law,” as national income increases
citizens both demand, and are willing to finance, increased levels of publicly-provided
services. In other words, the demand and willingness to pay for such services are income-
elastic. The conclusion, then, is that government expenditure and revenue levels will
grow proportionally to national income (Cameron 1978: 1245; Peltzman 1980: 218; Alt
and Chrystal 1983: 184; Borcherding 1985: 365; Holsey and Borcherding 1997: 569). To
capture this potential effect, the analyses include the natural log of per capita gross
domestic product (LGDPPC) as a control. The source of this variable is the World
Bank’s WDI database.
Trade: A number of scholars have documented the empirical regularity that the
more open an economy the larger its public sector. Explanations for this relationship

77
As Wilensky (1976: 47-48) comments, “If there is one source of public spending that is most powerful—
a single proximate cause—it is the proportion of old people in the population. The welfare state is a symbol
of the ambiguous position of the aged in modern society, both dependent and independent, a minority of
strategic importance in public spending.”

99
include compensatory social insurance programs to the sectors and industries exposed to
increased external risk (Cameron 1978; Katzenstein 1985; Rodrik 1997; 1998 Adsera and
Boix 2002; see also Ruggie 1982), or as a precondition to trade liberalization by buying
the cooperation of the non-tradable sector to maintaining slow wage growth in order to
ensure the overall competitiveness of the economy (Katzenstein 1985; Adsera and Boix
2002). The increased size of government may also be due to the easily accessible tax
bases ensuing from taxes on exports and imports (Persson and Tabellini 2003b). Thus,
the expectation is for trade openness to be positively related to both measures of
government size. In addition, it is also possible for trade openness to be related to deficits
and debt seeing as an increased level of imports may lead to balance-of-payments deficits
that require greater amounts of borrowing (Cameron 1978). Trade openness (TRADE) is
measured as the sum of exports plus imports as a proportion of GDP. The source of the
trade openness indicator is the World Bank’s WDI database.
Resources: Some have suggested that the so-called “resource curse” shown to
hinder economic growth also may adversely affect fiscal policy outcomes (Coutinho
2011). Specifically, high natural resource prices may lead to overspending since countries
can easily access cheap credit on international capital markets (Frankel 2011).
Furthermore, this borrowing can lead to a debt problem in periods of lower natural
resource prices (Barnett and Ossowski 2002). The dataset contains two alternative
variables to account for these possible effects. The first is RESOURCES which is a
dummy variable equal to one for countries classifying as hydrocarbon- and/or mineral-
rich. The other indicator RESRENTS measures a country’s total natural resource rents as a
percent of GDP. It is the sum of oil rents, natural gas rents, coal rents, mineral rents, and

100
forest rents. The sources of these respective indicators are the IMF’s Guide on Resource
Revenue Transparency (2007) and the WDI database.

2.1.2.2 Political Institutions Control Variables
78

Branches of Government: The subsequent analyses include four indicators
intended to capture and control for various aspects of the government branches in each
country. To control for the potential impact of the size of the legislature on fiscal policy
outcomes, included is a variable of the total number of seats in the legislature
(TOTSEATS).
79
Also included is an indicator variable for countries with bicameral
legislatures (BICAMERAL). In the case of a bicameral legislature, the variable RELSUC
measures the relative size of the upper chamber vis-à-vis the legislature as a whole.
Finally, to control for the degree of political fragmentation within the executive branch,
the variable GOVFRAC is the probability that two deputies picked at random from among
the government parties will be of different parties. With the exception of BICAMERAL,
the source of these variables is Beck et al. (2001). The source of the BICAMERAL
variable is Teorell et al. (2011).
Political Parties: Two variables control for the effects of political parties. The
first, LGSTPARTY, counts the largest party’s number of seats divided by the legislative
assemblies’ total number of seats. In cases of bicameral legislatures, the lower house is
counted. The second, ENLP, counts the “effective” number of parliamentary or
legislative parties (Laakso and Taagepera 1979; Taagepera 1997). The effective number
of legislative parties counts not only the number of political parties in the legislature but

78
See Chapter 2 for a discussion of the expected relationships between political institutions variables and
fiscal policy outcomes.
79
In the case of bicameral legislatures, this variable counts the total number of seats in the lower chamber.

101
accounts also for their relative size or strength, which, in this case, is based on the
parties’ share of legislative seats.
80
The formula for said measure is ( )
1
1
2
−
=
∑
=
n
i
i
p ENLP ,
where
2
i
p is the square of party i’s proportion of seats in the legislature and n is the total
number of parties. The source of the party size variable is Teorell et al. (2011). The
source of the number of legislative parties is Gallagher and Mitchell (2008) and Teorell et
al. (2011).
Elections and Electoral Systems: In controlling for the electoral rules of each
country I follow Persson and Tabellini (2003b) and first classify cases based on the
electoral formula for electing the lower chamber. Two binary indicators distinguish
between countries with “majoritarian” (MAJ), “mixed” (SEMI), and “proportional”
(comparison group) electoral rules.
81
Countries also are classified according to their form
of government. A dichotomous presidentialism variable (PRES) distinguishes between
parliamentary regimes where the executive is subject to a legislative confidence
requirement (PRES = 0), and presidential regimes (PRES = 1), where such confidence
requirement is absent (Persson and Tabellini 2003b: 97).
82
Lastly, the electoral formula
and regime variables are combined to capture the political system as a whole. Thus,
countries are distinguished according to whether they have majoritarian and presidential

80
A similar measure is the “effective number of electoral parties,” which weighs parties according to their
vote share as opposed to their share of legislative seats (Laakso and Taagepera 1979). The theory outlined
in Chapter 2 regarding legislature size related to the number of parties specifically in the legislature rather
than in the system as a whole. For this reason the more appropriate measure is the effective number of
legislative parties.
81
Given the small number of countries with mixed electoral rules, in the subsequent analyses they are
combined with proportional as part of the comparison group.
82
According to some, the absence of the legislative confidence requirement is but one defining
characteristics of presidential regimes. Shugart and Carey (1992: 19), for example, identify four defining
criteria of presidential regimes: (1) the executive is popularly elected; (2) The terms of the executive and
the legislature are fixed and do not require mutual confidence; (3) The executive is solely responsible for
appointing and directing the composition of the government; and (4) The president has some
constitutionally granted lawmaking authority. Persson and Tabellini (2003b: 97) defined cases according to
the second criterion in order to simplify the data-collection process.

102
systems ( ) PRES MAJ MAJPRES × = or majoritarian and parliamentary systems
( [ ] PRES MAJ MAJPAR − × = 1 ), and between systems that are proportional and
parliamentary ( [ ] [ ] PRES MAJ PROPAR − × − = 1 1 ) and systems that are proportional and
presidential ( [ ] PRES MAJ PROPRES × − = 1 ). The data sources for these variables are
Persson and Tabellini (2003b), Beck et al. (2001), and Teorell et al. (2011).
Levels of Government: Finally, Chapter 2 discussed the debate within the political
economy literature as to whether centralized or diffuse fiscal decision-making was more
likely to result in higher levels of spending. The analyses include two variables in order
to account for either of these two possibilities. The first is an indicator variable,
FEDERAL, equal to one for federal political structures and zero otherwise. The second,
FISCCEN, is a measure of the degree of fiscal centralization and equals the ratio between
central and general government spending. Higher values indicate greater fiscal
centralization. The source of the FEDERAL variable is Teorell et al. 2011. The source of
FISCCEN is the IMF’s Government Finance Statistics database and the World Bank’s
Fiscal Decentralization Indicators database.

2.2 Dependent Variables
The following section describes the data sources and measures of the fiscal policy
outcome variables to be explained in subsequent analyses. All measures pertain to the
central government in each country. The reasons for focusing on the central, as opposed
to the general government, are threefold. First, data are more widely available for the
central government than for other levels of the public sector (e.g. general, regional, local,
etc.). Second, measures taken at the central government level are more comparable. As

103
Persson and Tabellini (2003b: 37-38) note, the definition of the relevant government
units and the exact definition of government expenditures and revenues at the various
levels of government are often not spatially or temporally comparable. In any case,
indicators measured at the central and general government levels are often highly
correlated. Third, fiscal measures at the central government level are used so that the
results of subsequent analyses may be compared to those of previous studies. Much of
the previous research of budget institutions and the determinants of fiscal policy more
generally, rely on measures taken at the central government level.
83
Thus, measures are
taken at the central government level on account of data availability and comparability.
Furthermore, all indicators are measured as a proportion of gross domestic
product at current prices. The purpose of doing so is to capture the relative, rather than
absolute, size of these fiscal outcomes using GDP as a means of standardization.

2.2.1 Government Expenditures, Revenue, and the Size of Government
The first set of fiscal policy outcome indicators seek to measure the size of
government. Conceptually, “size of government” describes the scale and scope of
government activity (Berry and Lowery 1984a). In particular, its purpose is to capture the
size of the public sector relative to the rest of society (Lewis-Beck and Rice 1985). Most
analyses, beginning with the literature examining the growth of post-War government,
have defined “government size” as the ratio of government expenditures (purchases of
goods and services and transfer payments) to the total output of an economy (Berry and

83
Important exceptions include Von Hagen and Harden (1995), Gleich (2003), Filc and Scartascini (2004),
and Stein et al. (2008) to name of few. With the exception of Stein et al. (2008), which uses measures from
the “consolidated public sector” (i.e. the central government, the social security system, public enterprises,
and local governments), these studies use fiscal measures from the general government.

104
Lowery 1984b). The reasons for such selection are possibly due to the fact that data for
such an indicator are the most available and easily measured (Larkey et al. 1981). In
addition, perhaps for this reason, it is the most widely used measure of government size.
There are, of course, other measures intended to capture the scope of government activity
such as revenues collected, the number of governmental units, public employment as a
percentage of the total workforce, or the number of regulatory agencies, to name a few
(Larkey et al. 1981: 163). However, none of these, for better or worse, has gained as wide
an acceptance as a preferred indicator of government size in such a way as has the
proportion of expenditures to GDP. Furthermore, with the exception of revenues, none of
these measures is simultaneously an indicator of fiscal policy, which is the present focus.
Thus, for consistency with previous work and consistency with the theory outlined in the
previous chapter, the primary measure of government size used here is the ratio of
government expenditures to gross domestic product (GDP) at current prices (EXP).
84

Nevertheless, as an alternative, government size is also measured equivalently as the ratio
of revenues to GDP (REV).
85
The two alternative measures correlate highly with one

84
Admittedly, focusing solely on the central government may bias downward subsequent results. To
“truly” measure the size of government, i.e. the scale and scope of government activity, one should include
expenditures at all levels of government. Otherwise, to focus exclusively on central government
expenditures is to disregard other key areas of government involvement in society (Lewis-Beck and Rice
1985: 3). Nevertheless, for the reasons stated above, the analyses are confined to measures of the central
government.
85
Despite its wide usage, as a measure of government size, expenditures as a percentage of GDP is not
without its controversies. Specifically, there is debate around which measure of national product is more
useful, GDP or Gross National Product (GNP), whether transfer payments should be included in the
definition of government expenditures, and how to handle inflation (Lewis-Beck and Rice 1981).
As measures of national wealth, GNP and GDP differ in how each of these measures considers the
goods and services produced by foreigners. GNP considers only the value of production of a nation’s
permanent residents whereas GDP measures the value of production within the geographic confines of a
country regardless of whether that value is generated by foreigners or nationals (Mankiw 1998: 203-204).
For a number of reasons GDP is the preferred measure of national wealth. First, it is standard practice
among economists to use GDP to measure the value of economic activity. Second, in most cases, the
distinction between the two measures in unimportant as in most countries domestic residents are primarily
responsible for a greater proportion of domestic production so the two measures are very close (Mankiw

105
another (r = 0.94) and will permit sensitivity analysis to test the robustness of the results.
For simplicity, I refer to EXP and REV as GOVSIZE in the methods section below.
The source of this data primarily comes from the International Monetary Fund’s
Government Finance Statistics database (GFS). However, the GFS database suffers from
missing data for a number of countries and years. This is true for a number of European
countries and especially countries in Asia and Africa. Therefore, the GFS data is
augmented with equivalent data from the OECD’s National Accounts Statistics database,
the Asian Development Bank’s (ADB) Statistical Database System, and the African
Development Bank’s (AfDB) Statistical Data Portal.
86

2.2.2 Budget Deficits and Government Debt
The second set of fiscal policy outcome indicators are intended to measure the
central government’s budget balance and its long-term consequences. The first indicator

1998: 204). Third, a country’s fiscal policy affects economic activity primarily within its borders, thus
making GDP the preferred measure.
A second issue is whether transfer payments (e.g. social security, welfare, unemployment benefits,
etc.) should be included in the calculation of government expenditures. The argument against including
transfer payments states that transfers do not divert resources from the private economy to the government
in the same way as does the purchase of goods and services. In this case, the government is simply a liaison,
redistributing income from one group of private citizens to another (Brown and Jackson 1991: 86). The
argument in favor of including transfers posits that the taxes required to finance transfers impose real costs
on those from whom they are levied; in fact, as much of a cost as any other outlay (Buchanan and Flowers
1980: 40-41). Thus, transfers must be included as a part of government expense (Lewis-Beck and Rice
1981: 5). Most importantly, what the government spends money on is just as important as how much
money is spent, and is every bit an issue of fiscal policy. Since the general analysis includes how the level
of expenditure is composed, the analyses include transfer payments as part of overall expenditures.
The final issues concerns whether and how to account for inflation. A number of scholars argue
that in order to obtain an accurate account of public outlays it is necessary to deflate expenditure figures by
a price deflator (See Beck 1976, 1979; Lowery and Berry 1983; Berry and Lowery 1984b). This is because
over time governments will spend more than they did in the past, however this does not necessarily mean
that the level of goods and services they provide will have increased, because prices have increased. In
other words, governments may be spending more simply to provide the same level of goods and services.
Others note that government expenditures as a percentage of GDP already accounts for price fluctuations
since price increases for the entire economy are reflected in rising prices to government (Herber 1975: 23;
Lewis-Beck and Rice 1981: 5). For this reason, and because more recent analyses of government
expenditures (particularly those emphasizing institutional explanations) use the unadjusted measure the
analyses here will follow this practice.
86
The earliest date for which data is available from the African Development Bank is 2001.

106
is the measure of budget balance itself, measured as the size of the budget deficit of the
central government, as a proportion of gross domestic product (DEFICIT). When this
value is negative, the budget is said to be in deficit, whereas a positive value indicates a
budget in surplus (a negative deficit). A value of zero indicates that the budget is
balanced.
When the government’s budget is in deficit the revenue shortfall is typically
compensated by borrowing (i.e. issuing debt). Governments borrow in a number of ways,
for instance by issuing bonds and securities, as well as borrowing from international
government and/or financial institutions. The second indicator is a measure of the total
outstanding borrowings of the central government, that is, the national, or public, debt.
Whereas the government deficit measures the annual increase in total borrowing, the
national debt represents the total stock of borrowing (Bannock et al. 2003: 270). In other
words, the debt measures the accumulation of all past budget deficits (Mankiw 1998:
277). The measure of the total stock of borrowing is straightforward and simply the ratio
of central government debt to gross domestic product (DEBT).
A number of data sources are utilized in constructing the deficit and debt
indicators. For deficits, the primary source of data is the IMF’s GFS and International
Financial Statistics (IFS) databases. As mentioned above, the GFS database is subject to
missing data. The same is true for the IFS database. Therefore, additional institutional
sources were consulted. These sources include the OECD and AfDB databases mentioned
above, as well as the World Bank’s World Development Indicators (WDI) database, and
the European Union’s (EU) Eurostat database.

107
Four sources contribute to the data on central government debt as a proportion of
GDP. These are the World Bank’s WDI database, the OECD’s StatExtracts database, and
the publicly provided data found in Reinhart and Rogoff (2009) and Jaimovich and
Panizza (2010).
87

2.2.3 Composition of Government Expenditures
The final set of fiscal outcome indicators are the various components of
government expenditure composition. As discussed in Chapter 3, the specific focus is on
two types of government expenditure. The first type of expenditure is transfers and
subsides to households and firms. Transfers are payments made by the government to
individuals, however not in exchange for some productive service (Mankiw 1998: 206).
Examples include pensions, unemployment benefits, and other forms of income support.
As such they represent a form of income redistribution (Bannock et al. 2003: 386).
Subsidies are defined as direct government payments to suppliers of goods and services
(Schwartz and Clements 1999: 120).
88
The second type of expenditure is expenditure on
public goods. Public goods spending is defined as the sum of current and capital spending
on goods and services, that is, the sum of government consumption and capital spending.
In measuring transfers and subsidies and public goods, I follow Milesi-Ferretti et
al. (2002) and Mukherjee (2003). Transfers and subsidies (TAS) are the measured as the
sum of two components: social security benefits and other transfers to households,
including subsidies to firms, over GDP. Public goods (PGOOD) are measured as the sum

87
The Jaimovich and Panizza (2010) data covers central government debt only up to 2005.
88
This definition applies to subsidies to firms. Subsidies may also be made to consumers in the form of
cash payments in order to purchase goods and services a below market prices.

108
of government wage consumption, and non-wage consumption, plus government
investment, net of depreciation, again, relative to GDP.
An alternative indicator of expenditure composition is the size of welfare state
programs such as those found in Persson and Tabellini (2003b). These authors measure
such programs using the level of social security and welfare spending (SSW) by the
central government. This measure is distinct from, albeit similar to the measure of
transfers and subsidies, and in fact the two indicators correlate highly with one another (r
= 0.90). Similarly to the other fiscal outcome variables, social security and welfare
spending is expressed as a ratio to GDP. Unfortunately, I am unaware of an alternative
indicator to public good spending. Thus, sensitivity analyses of the results will
necessarily be limited to the effect of budget institutions on expenditure composition in
terms of transfers and subsidies and social security and welfare.
The primary source of data for the composition of government spending is the
IMF’s GFS database. In addition to the GFS, the variables for transfers and social
security and welfare are augmented by the IMF’s World Economic Outlook (WEO)
database, and the Asian Development Bank, the Economic Commission for Latin
America and the Caribbean (CEPALSTAT), and Persson and Tabellini (2003b),
respectively. Table 4.1 depicts the summary statistics of the dependent variables used in
this study.

109

Table 4.1
Descriptive Statistics of Dependent Variables
Standard
Mean Deviation Minimum Maximum
EXP 29.87 8.82 13.01 52.68
REV 28.46 8.90 11.66 54.09
TAS 8.71 5.91 0.97 24.98
SSW 8.52 5.75 0.06 26.53
PGOOD 12.34 5.85 3.71 38.56
DEFICT -1.50 2.76 -7.56 8.64
DEBT 56.72 31.68 2.46 187.27

3. Case Selection
An apparent, if unstated assumption, underlying the theoretical arguments in
Chapter 3 is that the focus of this study is in fiscal policy as it is made within the confines
of democratic institutions. Thus, the theories herein should only apply to, and only be
tested among, democratic regimes. Nevertheless, it is not an innocuous or uncontroversial
choice in selecting the criteria by which to classify regimes as democracies and non-
democracies. There is ongoing debate in the literature over how democracy is properly
conceptualized, operationalized, and consequently how individual cases should be
scored.
89
The result is that there is an abundance of measures and no consensus among
practitioners as to which is to be preferred. I proceed conservatively and follow Collier
and Adcock (1999:539) who argue that the choice between measures of democracy
should be based on the goals and context of research. The specific measures are selected
with three specific goals in mind. First, given the relatively small sample size, the aim is
to maximize the number of prospective observations contained in the dataset. Thus, the
criteria used to classify as a democracy may require being quite charitable. Second, it is

89
For discussions of these debates see Bollen (1980) Bollen and Jackman (1989), Collier and Levitsky
(1997), Collier and Adcock (1999), and Munck and Verkuilen (2002).

110
desirable to place the results of this study within the larger research agenda of the effects
of political institutions. Therefore, it is useful to select the most widely-used measures.
Finally, previous work has demonstrated the volatility of analytical results to the use of
specific measures. Accordingly, it is important to test the sensitivity of the results to
alternative measures.
Four distinct democracy indices will be utilized in subsequent chapters. The first
two are the Freedom House indices of political rights and civil liberties and the Polity IV
measure of “authority patterns.” These are two of the most wide used measures and are
graded which allows for varying thresholds so as to maximize sample sizes. The second
two measures are the dichotomous indices of Cheibub et al. (2010) and Boix et al.
(forthcoming). These indices are useful because not only are they improvements in many
respects over the Freedom House and Polity IV datasets (given their clarity in terms of
conceptualization and measurement. See Munck and Verkuilen 2002), but they classify
cases differently from Freedom House and Polity IV and from one another. Thus, they
contribute to testing the sensitivity of the analytical results. Each of these measures is
discussed in detail below.
The data used to classify democracies come from four diverse sources. The first
two are the well-know Freedom House and Polity IV data sets (Marshall and Jaggers,
2010). These are two of the most widely-used measures in the literature. The Freedom
House index measures democracy across two dimensions—political rights and civil
liberties. Countries are assigned values on a discrete scale from 1 to 7 along each of these
dimensions, with low values signifying better quality of democratic governance. These
scores are assigned by a team of experts who consider a checklist of factors. The

111
checklist is comprised of 62 items for political rights (separated into ten categories plus
an additional two discretionary categories) and 80 items for civil liberties (divided into 15
categories).
90
The checklists include questions such as: are the head of government and
legislature elected through free and fair elections? Are the electoral laws and framework
fair? Is the government accountable to the electorate between elections, and does it
operate with openness and transparency? Are there free and independent media? Is there
freedom of assembly? And, is there equality of opportunity and freedom from economic
exploitation? For each of the categories coders assign values from zero to four and
countries can obtain a maximum score of 40 in the political rights and 60 in the civil
rights. These scores are then condensed into the two seven-point scales of political rights
and civil liberties. Countries with values of 1 or 2 on a scale are considered “free,”
countries with scores between 3 and 5 are considered “partly free,” and countries with
values of 6 and 7 are labeled “not free.” Following Persson and Tabellini (2003b) I take
the average of these two scores to create a combined measure of democracy (GASTIL).
To be included in the democracy sample, a country must score lower than an average of 5.
Using this threshold, 83 countries (≈ 86%) are classified as democracies. However, this
generous classification permits countries such as Burkina Faso, Jordan, Namibia, and
Zambia to qualify as “democracies,” which prima facie appears to run counter to any
casual understanding of democracy. For this reason, a stricter definition of democracy is
also used. In this case, countries must score below an average of 3.5 to be considered a
democracy. This classification reduces the sample by 17 countries and includes

90
For a more complete description of the checklist and methodology Freedom House uses to construct its
index see http://www.freedomhouse.org/report/freedom-world-2011/checklist-questions.

112
approximately 68% of the total sample (66 of 97). These thresholds represent
approximately the upper and lower limits of the “partly-free” category.
91

The Polity IV index is a measure of the “authority patterns” of political regimes
and includes indicators for democracy and autocracy (Marshall et al. 2002). These
indicators are comprised of five components: (1) a competitiveness of political
participation (2) regulation of political participation, (3) competitiveness of executive
recruitment (4) openness of executive recruitment, and (5) constraints on chief executive.
Countries are assigned values on each of these dimensions for a total score between 0 and
10 for both the democracy and autocracy variable. These scores are combined (by
subtracting the autocracy score from the democracy score) to form a 21-point scale of
democracy ranging from -10 to +10 (POLITY). According to Marshall and Jaggers (2010),
Polity IV scores between -10 and -6 are to be considered “autocracies,” scores between -5
and +5 are “anocracies,” and scores between +6 and +10 are considered “democracies.”
92

91
These thresholds are admittedly arbitrary. However, they are the same thresholds as those used in
Persson and Tabellini (2003b) and Wehner (2006) and as it stands no consensus exists as to the proper
threshold to select when attempting to distinguish between democracies and non-democracies. In fact,
Bogaards (2010) surveys the literature on democracy and war and finds at least 12 different thresholds for
Freedom House and at least 14 different Polity thresholds used by scholars with little or no justification. As
a general matter, I institute specific thresholds and use dichotomous measures (and dichotomizations of
graded measures) of democracy given that the research aim is to separate the cases into distinct and
identifiable groups of democracies and non-democracies, rather than, say, determine what affect democracy
or the quality of democracy has on some other outcome of interest, a research question that may be better
suited to an interval measure of democracy. Specifically, I am invoking Collier and Adcock (1999:539)
who suggest that research that utilizes a concept of democracy as “bounded whole,” i.e. constituted by
multiple attributes individually necessary and collectively sufficient in order to be considered an instance of
the concept, they suggest, are more amenable to dichotomous measures (543, 561-562).
92
Hegre et al. (2001: 33) define “anocracies” as semi-democracies that are “partially free yet somewhat
repressive,” and Fearon and Laitin (2003: 75-76, 81) describe anocracies as possessing “financially,
organizationally, and politically weak central governments,” and essentially as countries that mix
democratic and autocratic features (See also Vreeland 2008: 403-404).

113
Thus, only countries with POLITY scores above +5 are included in the analyses. This
leads to a sample of 62 countries, or approximately 64% of the total.
93

Despite their wide usage, both of these measures have been subject to heavy
criticism. Freedom House, for example, has been criticized for including too many
superfluous attributes, failing to describe the relationship between components and
attributes, neglecting to provide a clear set of coding rules and a justification for the use
of specific indicators and level of measurement, and finally for not providing an explicit
theoretical justification for the choice of aggregation rule (Munck and Verkuilen 2002: 9,
14, 19, 21, 25). Similarly, Polity IV has been criticized for omitting participation as part
of the index’s attributes while including other redundant attributes, and for failing to
provide a justification for the specific weighting scheme used in the aggregation
procedure (Munck and Verkuilen 2002: 11, 14, 26; Treier and Jackman 2008: 202). The
aforementioned rules of aggregation of both measures are perhaps the most troubling to
scholars. For example, given that the Freedom House has ten categories for political
rights and 15 categories for civil liberties, each with a possible rating from zero to four,
there are 5
10
= 9,765,625 possible ways to obtain a sum of scores between 0 and 40 in
political rights, and 5
15
= 30, 517,578, 125 possible ways to obtain a score between 0 and
60 in civil liberties. These scores are then arbitrarily compressed into the two 7-point
scales of political rights and civil liberties (Cheibub et al. 2010: 75). The same is true for
the Polity IV index. Cheibub et al. (2010: 76) show that there are 1,512 possible
combinations that lead to the various Polity IV scores (see also Gleditsch and Ward 1997:
366-367; Treier and Jackman 2008: 205-206). Thus, in both instances, multiple scoring

93
Polity IV only rates countries with populations over half a million inhabitants. Thus, countries such as
Iceland, Luxembourg, and Malta are excluded from the data. Polity also does not have data for Hong Kong
and Bosnia and Herzegovina.

114
patterns lead to the same assigned score (Gleditsch and Ward 1997; Treier and Jackman
2008).
To address such issues, two strictly dichotomous measures of democracy are also
used.
94
First, I employ the measure of democracy first developed by Alvarez et al. (1996)
and Przeworski et al. (2000) and recently updated by Cheibub et al. (2010). Their concept
of democracy is explicitly procedural, or “minimalist,” in that democracy is simply
defined as a means of selecting rulers through competitive elections (Przeworski 2003:
12). According to their classification, a regime qualifies as a democracy if and only if it
meets all of the following criteria: (1) the chief executive must either be chosen directly
by popular election or indirectly through a body that was itself popularly elected; (2) the
legislature must be popularly elected; (3) there must me more than one party competing
in the elections; and (4) an alteration in power under electoral rules identical to the ones
that brought the incumbent to office must have taken place (Cheibub et al. 2010: 69).
When conditions (1) through (4) are met, cases are considered democracies and are coded
1 on the variable CGV and 0 otherwise. The biggest discrepancies are likely to arise with
those cases that fulfill conditions (1) through (3) but fail to meet condition four.
95
In these
cases, there have been multi-party elections held under conditions widely believed to
have been free and fair, however each time the incumbent has won and thus there is no
way to determine with absolute certainty whether there would have been a peaceful
turnover of power had the incumbent lost, or whether the government would have utilized
extralegal means to remain in power. In these instances there is simply not enough

94
I use the phrase “strictly dichotomous” because using thresholds to separate values on the graded
Freedom House and Polity IV measures is equivalent to dichotomizing the measures.
95
There are 14 such cases in the present dataset. They are: Botswana, Burkina Faso, Cambodia, The
Republic of Congo, Ethiopia, Guinea, Lesotho, Mozambique, Namibia, Russia, South Africa, Tajikistan,
Tunisia, and Zambia.

115
information to code such cases with the above rules. Thus, Cheibub et al. (2010:70-71)
take a cautious approach and code such cases as dictatorships.
96
However, these cases are
identifiable with an indicator variable TYPE2, coded one if the dictatorship corresponds
to a possible Type II error—i.e., a false negative—and zero otherwise. Using this
approach it is possible to examine the robustness of the results to the inclusion or
exclusion of such cases. Excluding the possible 14 type II cases (or some subset of them),
conditioning on CGV results in a sample of 68 countries (≈70%).
The fourth and final measure of democracy comes from Boix et al. (forthcoming).
According to their measure, countries are classified as democratic if they meet all of the
following conditions: (1) The executive is elected (directly or indirectly) in popular
elections an is responsible either directly to voters or to the legislature; (2) the legislature
(or the executive if elected directly) is chosen in free and fair elections; where (2a) “free”
signifies that voters are given multiple options on ballots and (2b) “fair” denotes the
absence of fraud and/or government electoral abuse; and (3) suffrage for a majority of
adult males (Boix et al. forthcoming: 8-9). These features are captured in the variable
BMR. Countries that satisfy these requirements are assigned values of BMR = 1, and 0
otherwise.
Note first the similarities between CGV and BMR. Both require the Executive and
Legislature to be elected (conditions (1) and (2) for both CGV and BMR) in politically
contested contests (condition (3) for CGV and (2a) for BMR). Boix et al. (forthcoming)
even use Cheibub et al. (2010)’s executive alternation condition as an indicator for free

96
Specifically, for this rule to apply conditions (1) through (3) must be met and the incumbent must have
held office by virtue of elections for more than two electoral terms or initially held office without being
elected. If the incumbents subsequently held elections but never lost, then the regime is classified as
authoritarian (Alvarez et al. 1996: 13). For a more in-depth discussion of this rule see Alvarez et al. (1996:
10-13) and especially Przeworski et al. (2000: 23-28).

116
and fair elections. Their main points of difference are with regard to the male suffrage
requirement and the extent to which executive turnover is required to qualify or
disqualify a country as a democracy.
97
Boix and collaborators check the histories of those
cases with no electoral turnover and judge a country to be a democracy if there was no
evidence of prolonged executive control by a single party due to events such as internal
coups, external interventions, or acts such as abuses of state power, or electoral fraud.
They are also aware that turnover may occur in non-democratic regimes, for example in
so-called “competitive authoritarian” regimes, and thus check whether alternation occurs
consensually and within a genuinely competitive political system (Boix et al.
forthcoming: 9-10).
98
For the 1997-2007 period these coding rules lead to four of
Cheibub et al.’s potential type II cases to be classified by Boix and associates as
democracies (Botswana, Lesotho, Mozambique, and South Africa) and two democracies
according to Cheibub and colleagues to be classified as non-democracies (Nigeria and
Solomon Islands). Finally, Boix et al. require a majority of adult men to have the right
to vote. While this requirement is absent from the criteria used to code democracies in
Cheibub et al. (2010) and Polity IV, its effect is minimal in the sample years used in this
study since, as the authors acknowledge, universal male suffrage was a fixture of nearly
all nations with free competitive elections after 1946 (Boix et al. forthcoming: 10). Under

97
Note that another way to distinguish between the measures of Cheibub et al. (2010) and Boix et al.
(forthcoming) is to think in terms of Dahl’s (1971) classification of democracies along the dimensions of
“contestation” and “participation.” Cheibub et al. focus solely on the contestation, or political competition
dimension, whereas Boix et al. include both dimensions since they consider majority male suffrage as one
of their requirements.
98
“Competitive Authoritarian” regimes are hybrid regimes that violate the basic standards for democracy
(e.g. free and fair elections; civil liberties and political rights, etc.) in order to give the government an
advantage over the opposition. Violations to minimum democratic standards occur frequently and to such
an extent that such regimes cannot be characterized as democracies. Nevertheless, meaningful democratic
institutions do persist and are taken seriously by all political actors. Thus, the persistence of genuine
“arenas of democratic contestation” and sometimes alternation also means that such regimes cannot be
classified as fully authoritarian (Levitsky and Way 2002: 52-54). See also Linz (2000) and Levitsky and
Way (2010). Gasiorowski (1996: 471) refers to such regimes as “semidemocracies.”

117
these authors’ coding scheme 70 countries in the total sample classify as democracies (≈
72%).
Table 4.2
Correlations Between Democracy Indices
GASTIL POLITY CGV BMR Average
GASTIL 1 0.82
POLITY 0.90 1 0.83
CGV 0.76 0.77 1 0.79
BMR 0.81 0.82 0.84 1 0.82
Note: Correlations are absolute values. GASTIL = Freedom House;
POLITY = Polity IV; CGV = Cheibub et al. (2010); BMR = Boix et al. (Forthcoming).

Despite the differences between the four measures with regard to how democracy
is conceptualized and measured, they remain highly correlated with one another. Table
4.1 lists the pair-wise and average correlations between indices. The pair-wise
correlations range from between a correlation of 0.76 (between Freedom House index
and that of Cheibub et al. (2010)) and 0.90 (between Freedom House and Polity IV). The
column on the far right lists each index’s average correlation with the others. All have
about a 0.80 correlation coefficient with the other indices. The Cheibub et al. index has
the lowest (although still relatively high) correlation coefficient of 0.79, although this is
likely due to the coding of some of the type II cases (e.g. Botswana and South Africa).
Nevertheless, correlation does not imply interchangeability. As Casper and Tufis (2003)
and Bogaards (2010) have shown, despite high correlations between democracy indices,
different indices can produce different analytical results.
99
For this reason, I follow these
authors’ advice and test the robustness of the analytical results to the inclusion and

99
Casper and Tufis (2003) examine the effect of using different democracy indices (specifically Polity IV,
Freedom House, and Vanhanen (2000)’s Polyarch 1.2 index) for studying democratization and Bogaards
(2010) examines the implications of using different indices (specifically, Reich (2002)—an update of
Gasiorowski (1996), Freedom House, Polity IV, Vanhanen (2000) and Doorenspleet (2000, 2003)) for the
relationship between democracy and war.

118
exclusion of cases based on the different measures of democracy (Casper and Tufis 2003:
203).
Table 4.3 lists the complete sample of 97 countries according to their scores on
the four measures of democracy. The column on the far-right labeled “ALL” lists those
countries that meet the minimum criteria to be classified as a democracy on all four
measures—i.e. scores lower than five on GASTIL, greater than five on POLITY, and
scores of one on both CGV and BMR. There are 60 such cases, or approximately 62% of
the total sample.
100
The variability in samples based on index scores suggests that the
results may differ according to the specific measure of democracy used. This reinforces
the decision to test the results against the various measures.
Table 4.3
Country Scores on Four Indices of Democracy
COUNTRY GASTIL POLITY CGV BMR ALL
Albania 3.545455 7 1 1 yes
Argentina 2.318182 8 1 1 yes
Australia 1 10 1 1 yes
Austria 1 10 1 1 yes
Belgium 1.227273 10 1 1 yes
Benin 2.136364 6 1 1 yes
Bolivia 2.5 9 1 1 yes
Bosnia and Herzegovina 4.136364 N/A 0 N/A no
Botswana 2 8 0 1 no
Brazil 2.727273 8 1 1 yes
Bulgaria 1.909091 9 1 1 yes
Burkina Faso 4.136364 -1 0 0 no
Cambodia 5.727273 1 0 0 no
Canada 1 10 1 1 yes
Chile 1.545455 9 1 1 yes
Congo 5.181818 -5 0 0 no
Costa Rica 1.318182 10 1 1 yes
Croatia 2.590909 8 1 1 yes
Cyprus 1 10 1 1 yes
Czech Republic 1.318182 10 1 1 yes
Denmark 1 10 1 1 yes
Ethiopia 4.863636 1 0 0 No
Fiji 3.772727 4 0 0 No

100
When the GASTIL threshold is lowered to 3.5, 56 cases qualify as democracies on all four indices. The
countries included above but excluded by this threshold are Albania, Sierra Leone, Turkey, and Venezuela.

119
Table 4.3 (continued)
Country Scores on Four Indices of Democracy
COUNTRY GASTIL POLITY CGV BMR ALL
Finland 1 10 1 1 yes
France 1.227273 9 1 1 yes
Germany 1.227273 10 1 1 yes
Ghana 2.272727 5 1 1 no
Greece 1.727273 10 1 1 yes
Guinea 5.5 -1 0 0 no
Haiti 5.454545 2 0 0 no
Hong Kong N/A N/A N/A N/A no
Hungary 1.318182 10 1 1 yes
Iceland 1 N/A 1 1 yes
Indonesia 3.636364 5 1 1 no
Ireland 1 10 1 1 yes
Israel 1.818182 10 1 1 yes
Italy 1.227273 10 1 1 yes
Japan 1.5 10 1 1 yes
Jordan 4.545455 -2 0 0 no
Kenya 4.318182 3 1 1 no
Korea, Republic of 1.818182 8 1 1 yes
Kyrgyzstan 5 -1 0 0 no
Latvia 1.409091 8 1 1 yes
Lesotho 3.181818 8 0 1 no
Liberia 4.772727 2 0 0 no
Lithuania 1.409091 10 1 1 yes
Luxembourg 1 N/A 1 1 yes
Madagascar 3.136364 7 1 1 yes
Malawi 3.409091 6 1 1 yes
Mali 2.454545 7 1 1 yes
Malta 1 N/A 1 1 yes
Mauritius 1.409091 10 1 1 yes
Mexico 2.590909 7 1 1 yes
Moldova 3.318182 8 1 1 yes
Mongolia 2.227273 10 1 1 yes
Morocco 4.681818 -6 0 0 no
Mozambique 3.454545 5 0 1 no
Namibia 2.363636 6 0 0 no
Netherlands 1 10 1 1 yes
New Zealand 1 10 1 1 yes
Nigeria 4.363636 3 1 0 no
Norway 1 10 1 1 yes
Papua New Guinea 2.772727 4 1 1 no
Peru 3.090909 7 1 1 yes
Philippines 2.681818 8 1 1 yes
Poland 1.318182 10 1 1 yes
Portugal 1 10 1 1 yes
Qatar 5.909091 -10 0 0 no
Romania 2.045455 8 1 1 yes

120
Table 4.3 (continued)
Country Scores on Four Indices of Democracy
COUNTRY GASTIL POLITY CGV BMR ALL
Russian Federation 4.909091 5 0 0 no
Rwanda 6 -4 0 0 no
Serbia, Republic of 3.138889 3 1 1 no
Sierra Leone 4.045455 6 1 1 yes
Slovak Republic 1.5 9 1 1 yes
Slovenia 1.227273 10 1 1 yes
Solomon Islands 2.863636 8 1 0 no
South Africa 1.590909 9 0 1 no
Spain 1.227273 10 1 1 yes
Suriname 2.045455 5 1 1 no
Swaziland 5.681818 -9 0 0 no
Sweden 1 10 1 1 yes
Switzerland 1 10 1 1 yes
Taiwan 1.681818 9 1 1 yes
Tajikistan 5.727273 -2 0 0 no
Thailand 3.090909 7 1 1 yes
Tunisia 5.545455 -4 0 0 no
Turkey 3.772727 7 1 1 yes
Uganda 4.681818 -3 0 0 no
Ukraine 3.409091 6 1 1 yes
United Arab Emirates 5.636364 -8 0 0 no
United Kingdom 1.227273 10 1 1 yes
United States 1 10 1 1 yes
Uruguay 1.136364 10 1 1 yes
Venezuela 3.590909 6 1 1 yes
Viet Nam 6.5 -7 0 0 no
Zambia 4.136364 4 0 0 no
Zimbabwe 5.909091 -4 0 0 no
NOTE: POLITY = POLITY IV; GASTIL = Freedom House; CGV = Cheibub et al. (2010);
BMR = Boix et al. (Forthcoming). Values of POLITY missing for countries with less than half a
million inhabitants. Values for Hong Kong are missing for all indices. Values for Bosnia and
Herzegovina are missing for POLITY and BMR. "All" signifies that the country has values lower than
5 on GASTIL, greater than 5 on POLITY, and 1 on both CGV and BMR.

4. Missing Data: Issues and Solutions
Before moving forward to the methodological specification, it is important to first
address the issue of missing data. Many variables of interest in this study suffer from
problems of missing data. For example, some of the indicators for fiscal transparency are
missing as much as 48% of observations. While the degree of missingness is much closer

121
to between 5% and 20% for most variables, these levels still present a number of
statistical problems and could lead to invalid inferences if left unaddressed.
This section offers a brief discussion of the possible causes of missing data as
well as their potential consequences for the strength of the study design as well as the
validity of its conclusions. This section also presents Multiple Imputation (MI) as a
possible solution to the problem of missing data. Also discussed are the issues with MI
particular to this study, such as the selection of an imputation model, determining the
proper number of imputations, and the effect of variable transformations.

4.1 The Causes of Missing Data
According to McKnight et al. (2007: 54-57), there are a number of common
reasons for data to missing. First, data may be missing as a result of the study design.
When a study places a sufficiently great burden on the participant, for example when the
duration of the study is long, there are multiple repeated measures, long questionnaires,
etc., subjects are more likely to fail to respond to questions, comply with research
protocols, or remain in the study. Thus, the likelihood of missing data increases with the
burden placed on the subjects.
Second, there may be missing data as a result of characteristics of the participants
in the form of item non-response. Participants may fail to respond to particular items
because they are unable to process the information or comprehend the material. Or,
subjects may neglect to answer specific items on account of their beliefs and attitudes
toward the research topic. In either case, the result is a lack of information on a subgroup

122
of the population. The result is a biased sample that threatens the generalizability of the
study’s conclusions (see below).
Third data may be missing on account of the conditions under which the data was
collected and managed. If data were collected under pressure, or under time constraints
there may be an increase in the likelihood that data is missing. The measurement
characteristics may also affect the rate of missingness. For instance, missing data may
result from malfunctioning equipment. Furthermore, missing data may arise as data are
transferred from one format to another.
Finally, data may be missing purely by chance. However, data that is missing
randomly does not bias statistical conclusions or causal generalizations. Nevertheless,
chance occurrences are not entirely innocuous. Such events, for example, may lead to
large amounts of missing data which can be problematic.
In this study, it is possible that missing data is due to participant characteristics.
To illustrate, very few countries have complete data for all sample years in the various
databases listed above. Countries with large deficits and high debt may be less likely to
report such information. Developing countries may not have the capacity to collect the
necessary statistics or may be slow to do so. The budget directors may have inadvertently
skipped survey questions. Furthermore, countries arguably self-select into the
international institutions that collect such data (or neglect to join, e.g. Cuba), and in some
cases countries may be excluded from membership for political reasons (e.g. Taiwan).
In all of these instances, it is something particular to a country that causes data to
be missing. In the missing data literature this type of missing data is known as “missing
not at random” (MNAR). This means that the likelihood of a missing value is related to

123
the variable itself and possibly also related to other observed variables. The problem with
data that are MNAR is that the relationship between a variable and its missing values
cannot be known to the researcher since it is impossible (by definition) to observe the
missing values (Allison 2002: 4 McKnight et al. 2007: 48-49). Thus, even though the
missing data in this study may be MNAR, one cannot say for certain. Fortunately for this
study, the method used to handle missing data—multiple imputation—is, in many
circumstances, compatible with data that are MNAR (Schafer and Graham 2002).

4.2 The Consequences of Missing Data
In addition to the myriad reasons for missing data, there are also a number ways
in which missing data can affect the conclusions of a study. In particular, there are three
broad areas of the scientific procedure that can be unfavorably impinged upon by missing
data. First, missing data may adversely affect construct validity. This is because when
data is missing, the ability to capture variables of interest as accurately as possible
becomes impaired. Specifically, for single measures with multiple items, missing data
may lead to measurement bias since the information on the variable only reflects a
portion of the construct of interest. In the case of a single measure with a single item,
when data is missing for that item all information about the construct is missing
(McKnight et al. 2007: 21-23).
Second, missing data may affect the internal validity of a study. When data are
missing, the set of observations used for analyses represents a smaller and potentially
biased subsample of the population, which may lead to inaccurate parameter estimates.
The result is weaker causal inferences concerning the associations between variables

124
(McKnight et al. 2007: 25). Specifically, missing data may affect the internal validity of a
study in two important ways. First, missing data may induce selection bias, such that
cases with complete data possess different characteristics from those with incomplete
data that are relevant for the study’s conclusions. Consequently, the unobserved
differences between groups may be responsible for the observed outcome rather than the
“intervention” itself. Second, missing data affects the data analysis. Many statistical
procedures assume a particular distribution for data (e.g. normal) and such distribution
can be affected by missing data. For example, the ordinary least squares (OLS) procedure
assumes normally distributed errors and when the data violate this assumption it
adversely impacts significance tests and parameter estimates (McKnight et al. 2007: 31-
32).
Finally, such differences may leave the final sample unrepresentative of the
intended population, thus affecting the generalizability of the study’s conclusions
(McKnight et al. 2007: 28). A lack of sample representativeness thus limits the external
validity of a study’s findings.

4.3 Multiple Imputation
In order to compensate for missing values and to be able to make valid inferences,
the analyses will utilize multiple imputation. Multiple imputation will fill in missing
observations with plausible values which maintain the overall variability in the
population and preserve the relationships among variables (Wayman 2003: 4). With
multiple imputation (MI) missing values are predicted using the observed values from the
other variables. These predicted values are then substituted for the missing values,

125
creating a full dataset known as an “imputed dataset.” This process is repeated m > 1
times, generating m complete multiple imputed datasets. The desired statistical analyses
are then performed on each of the m imputed datasets, producing a set of analytical
results. These results are then combined to produce a single set of parameters and test
statistics (Little and Rubin 1989: 294; Collins et al. 2001: 335; Sinharay et al. 2001: 320;
Schafer and Graham 2002: 165; Wayman 2003: 4).
There are a number of possible techniques for handling missing data from which
to choose.
101
However, multiple imputation is the standard method for handling missing
data and arguably the best among the many methods available (Allison 2002: 56;
McKnight et al. 2007: 196).
102
This is due to the fact that MI has a number of advantages
over alternative approaches. For example, unlike common methods such as complete case
analysis (listwise deletion) and single imputation (e.g. marginal mean imputation and
conditional mean imputation), MI appropriately introduces random error into the
imputation procedure which produces unbiased estimates of the parameters. Furthermore,
repeated imputation permits one to obtain accurate estimates of the standard errors
(Allison 2000: 301-303).
103
Finally, multiple imputation is robust to violations of

101
See Sinharay et al. (2001: 318-320), Schafer and Graham (2002: 155-162), and McKnight et al. (2007:
Chapters 7-9) for summaries of the various methods for handling missing data.
102
The closest possible contender is Maximum Likelihood (especially Expectation Maximization).
Nevertheless, MI and Maximum Likelihood (ML) are closely related and will produce similar results under
certain conditions (Collins et al. 2001: 331. See pages 336-338 for these conditions). However, MI has a
number of advantages over Maximum Likelihood. For instance, in most practical applications MI is
computationally simpler than ML. In addition, Maximum Likelihood is problem specific and requires
unique computation procedures for different models, even those applied to the same data set. Conversely,
under MI, the same imputed datasets may be used for various types of analyses (Sinharay et al. 2001: 320).
103
However, one disadvantage of deliberately introducing random variation into the imputation process is
that MI produces different estimates every time it is used. This could potentially lead to problems with
replication even when researchers use the same methods of analysis. Nevertheless, these differences are
likely to be small and without a random component variances and covariances are usually underestimated
and test statistics are generally too high (Allison 2002: 28-29).

126
normality assumptions and results remain adequate even when sample size is low and/or
there are high rates of missing data (Wayman 2003: 3).
There are five steps to the MI process.
104
First, values are imputed for missing
observations using a model that incorporates random variation. Values may be imputed in
one of four ways: (1) unconditionally, (2) by conditioning on other variables, (3) with a
theoretical model, or (4) purely empirically. Second, the imputation procedure is repeated
m times, producing m number of “complete” datasets. Next, standard complete-data
analyses are performed on each of the m datasets. Then, the parameters are estimated
from the aggregated results. Each parameter in the statistical analysis (e.g. each
regression coefficient) is derived by simply averaging across the estimates produced by
each of the imputed datasets. Lastly, the standard errors are calculated by first taking the
mean of the squared standard errors of the m estimates (the within-imputation variance),
then calculating the variance of the m parameter estimates over each of the samples (the
between-imputation variance), and finally by combining the two quantities (to obtain the
total variance) using a simple formula developed by Rubin (1987).
105
After these five
steps, the results are presented.
The first task is to select an imputation model to impute missing values. This
involves two modeling choices (Van Buuren et al. 1999: 687). The first is the set of
predictors included as part of the model. It is generally agreed that at a minimum, all of
the covariates and the outcome from the analysis model must be included as part of the
imputation model. In cases where there will be a number of analytical models (such as

104
See Allison (2000: 301) and McKnight et al. (2007: 201-210).
105
The methods for combining estimates and standard errors are collectively known as “Rubin’s Rules.”
For a technical discussion of how each of these quantities is derived, see Little and Rubin (1989: 304-306),
Rubin (1996: 476-477), Sinharay et al. (2001: 323-324), Schafer and Graham (2002: 165-166), and
McKnight et al. (2007: 202-206).

127
this study), then the imputation model should include every variable included in any of
these models (White et al. 2011: 384-385). Otherwise, the complete data analysis may be
biased, especially if the analytical model contains strong predictive relations (Van Buuren
et al. 1999: 687). This is especially true when considering whether to include the
dependent variable in the imputation model. Failure to include the dependent variable in
the imputation model will result in imputed values for the predictors that will not be
associated with the dependent variable. As a result, the regression coefficients for the
independent variables with substantial missing values will be biased toward zero (Allison
2000: 307; von Hippel 2007: 84, 2009: 266).
What is less agreed upon, however, is whether the imputation model should
include variables not in the analysis model but included solely to improve the
performance of the missing data procedure. Collins et al. (2001) refer to such variables as
auxiliary variables. These variables may be included because they are correlates of the
variable with missing values and/or are a correlate of missingness. These authors
advocate for an “inclusive strategy” of including numerous auxiliary variables. Their
simulations suggest that with an inclusive strategy important correlates or causes of
missingness are less likely to be omitted. Furthermore, when auxiliary variables are
correlated with the variable with missing values, their inclusion can result in a decrease in
standard errors. Consequently, including auxiliary variables may increase efficiency and
statistical power. Perhaps most importantly, the authors argue that when data are MNAR
(a possibility in this study) auxiliary variables that are correlated with the variable with
missing variables my help to alleviate some of the bias because they may serve as useful

128
surrogates for the variable with missing values (Y) when Y is missing (Collins et al. 2001:
338).
106

The second modeling choice is the correct functional form of the model. This
simply means, on the one hand, modeling continuous variables with missing values using
a linear model (or some appropriate variation), modeling dichotomous variables with
missing values using a logistic model, ordinal variables with missing values using an
ordinal logistic model, and so on. On the other hand, the correct functional form requires
possibly dealing with interactions and non-linearities (White et al. 2011: 385). von Hippel
(2009) distinguishes between two approaches for both interactions and non-linearities
(e.g. squares). The first is to calculate the interactions and non-linearities in the
incomplete data, and then impute these transformations just as one would any other
variable. This method von Hippel labels, “transform, then impute.” The second approach
is to impute variables in their “raw” form and then construct transformations in the
complete imputed data set. This technique he labels, “impute, then transform.” For both
logical and methodological reasons, the transform, then impute method is preferable to its
alternative. Logically, if the proper functional form of the imputation model includes
interactions and non-linearities, then they must be included. This would not be possible if
one followed the “impute, then transform” method. Second, von Hippel shows that the
“transform, then impute” method produces unbiased regression estimates whereas the

106
The argument against an inclusive strategy posits that it is possible for the analyst to include variables
that are neither correlated with Y nor a correlate of missingness. In this case it is likely that such poor
predictors when added to the imputation model inflate standard errors (Allison 2002: 54-55). However,
Collins et al. (2001) show that, at worst, the effect of including such variables is neutral and, at best,
beneficial.

129
“impute, then transform” method does not.
107
Thus, choosing the correct functional form
requires transforming, then imputing, interactions and non-linearities if necessary.
After the imputation model has been chosen, the final task prior to imputation is
the selection of the number of imputations, m. Many authors argue that only a few
imputations are necessary to obtain satisfactory results. For instance, McKnight et al.
(2007) follow Rubin (1987) and suggest selecting a value of m as low as three and as
high as ten. Schafer and Olsen (1998) argue that five imputations are sufficient and
suggest that any more imputations will produce little additional value. Nevertheless,
recent work by Graham et al. (2007), Bodner (2008), and White et al. (2011) concludes
that the number of imputations should be similar to the percentage of cases that contain
missing data. In other words, 20 imputed data sets should be created if 20% of the cases
in the dataset containing missing data on one or more variables in the model. These
studies show that when m is insufficiently small, there is a significant statistical power
drop-off. Conversely this is not the case when m is comparable to the proportion of
missing data.
This discussion produces fairly clear guidelines for how imputation should
proceed prior to analysis. First, I opt for an inclusive imputation model that includes all
of the variables in the full analysis model (see the following section), the dependent
variable, and auxiliary variables.
108
Second, all interaction terms will be constructed first
so as to be a part of the imputation model, prior to being imputed. Finally, given that 72

107
Incidentally, the “transform, then impute” method produces imputed values that are inconsistent with
one another. For example, imputed values for x
2
may not equal the square of x and may even be negative.
As mentioned above, however, the estimates are nevertheless unbiased. The “impute, then transform”
method however, will produce imputed values that are consistent with one another yet will be biased (von
Hippel 2009: 274).
108
The use of auxiliary variables is discussed in greater detail in the next chapter.

130
of the observations (countries) have missing data on at least one variable, a total of 72
imputed datasets will be created.
109
This will ensure a proper functional form for the
imputation model and prevent any significant power drop-off.

5. Method: Econometrics
Now that the issues of sample selection and missing data have been addressed, it
is possible to move forward toward considering the proper method of analysis. Recall
from previous chapters that the research question motivating this study is “what are the
political-institutional factors that determine fiscal policy outcomes?” And in particular,
“what influence do budget institutions have in determining fiscal policy outcomes?” As
was mentioned above, the best available indicators of budget institutions are qualitative
in nature, specifically responses to questionnaires regarding actual budget practices in 97
countries. Nevertheless, it was shown that it was possible to rank-order and numerically
scale the responses based on the theory developed in Chapter 3. Furthermore, indicators
of fiscal policy outcomes are all quantitative. The questions that arise, then, are what is
the best possible method for analyzing the existing data and how will this method provide
answers to the research question?
In particular, the objective of the study is two-fold. First, to test whether the
various budget institutions affect fiscal policy outcomes in the ways hypothesized and,
second, to compare the impact of budget institutions to other economic and political-
institutional variables argued to influence fiscal policy outcomes. Thus, given the nature

109
As discussed in the following chapter, only 83 of the 97 countries enter into the analyses. Of these 83
countries, only 11 have complete data. Thus, 72 countries are missing data on at least one variable and
explains why 72 imputations is necessary. This may seem like a dauntingly high figure, and to some degree
it is, but Stata is capable of handling such calculations (Acock 2010). What is more, the fact that only the
dataset only contains 11 complete cases is further evidence that multiple imputation is an absolute necessity.

131
of the data, the size of the sample, and the objectives of research, econometric analysis is
the appropriate method of analysis. Specifically, cross-sectional analysis given that the
data for political-institutions were either gathered at one particular point in time and/or
exhibit very little to no variation over time.
110

In order to examine the affects of the variables of interest on fiscal policy
outcomes as well as examine the relative influence of each “block” of variables (i.e.
economic, political-institutional, and budget institutional) I specify a series of nested
regression models. First, the analyses are run against a baseline “economic” model for
each of the fiscal policy outcomes:
ε β
β β β β β β
+ +
+ + + + + =
RESOURCES
TRADE LGDPPC POP POP LPOPTOT FISCAL
i
6
5 4 3 2 1 0
65 1564

where FISCAL denotes fiscal outcome i = {GOVSIZE, DEFICIT, DEBT, TAS, SSW,
PGOOD}.
Next, the analyses test a baseline “political institutional” model against each of the fiscal
outcomes:
111

ε β
β β β β β
β β β β β
+ +
+ + + + +
+ + + + =
FISCALCEN
FEDERAL PROPRES MAJPAR MAJPRES ENLP
LGSTPARTY GOVFRAC BICAMERAL TOTSEATS FISCAL
i
10
9 8 7 6 5
4 3 2 1 0

Subsequently, the two models are combined to run a “political-economic” model:

110
I return to this issue below.
111
In order test the influence of the relative size of the upper chamber (RELSUC) on the fiscal policy
outcomes, I separate the sample into bicameral and unicameral legislatures and run the regressions with
RELSUC on the bicameral subsample. This model is the same as the political institutional model describe
above, however with RELSUC in the place of BICAMERAL.

132

For each fiscal policy outcome, I examine the effect of budget institutions in a series of
three models. The first includes just the budget institutions variables describe above:
ε β β β β + + + + = CY TRANSPAREN PRORULES FISCRULES FISCAL
j i 3 2 1 0

where PRORULES and TRANSPARENCY refer, respectively, to the procedural rules and
transparency indices constructed according to equation 1, and FISCRULES refers to the
fiscal rules indices and j = {expenditure rule, revenue rule, balanced budget rule, debt
rule}.
112
The second is similar to the “political economy” model by combining the
budget institutions variables with the set of economic variables:
ε β β β β
β β β β β β
+ + + + +
+ + + + + =
CY TRANSPAREN PRORULES FISCRULES RESOURCES
TRADE LGDPPC POP POP LPOPTOT FISCAL
j
i
9 8 7 6
5 4 3 2 1 0
65 1564

Finally, the complete model including the economic, political, and budget institutions
variables is tested:
ε
β β β β
β β β β β
β β β β β
β β β β β β
+
+ + + +
+ + + + +
+ + + + +
+ + + + + =
CY TRANSPAREN PRORULES FISCRULES FISCALCEN
FEDERAL PROPRES MAJPAR MAJPRES ENLP
LGSTPARTY GOVFRAC BICAMERAL TOTSEATS RESOURCES
TRADE LGDPPC POP POP LPOPTOT FISCAL
j
i
19 18 17 16
15 14 13 12 11
10 9 8 7 6
5 4 3 2 1 0
65 1564
.
The models are “nested” in the sense that all of the predictors of the economic and
political-institutional baseline models are included in the political-economic model, and

112
I also examine whether the effect of fiscal rules is conditional to the level of transparency. This is done
by running models with just the fiscal rules and procedural rules variables and the level of transparency set
to the 25
th
, 50
th
, and 75
th
percentiles.
ε β β
β β β β β
β β β β
β β β β β β
+ + +
+ + + + +
+ + + +
+ + + + + =
FISCALCEN FEDERAL
PROPRES MAJPAR MAJPRES ENLP LGSTPARTY
GOVFRAC BICAMERAL TOTSEATS RESOURCES
TRADE LGDPPC POP POP LPOPTOT FISCAL
i
16 15
14 13 12 11 10
9 8 7 6
5 4 3 2 1 0
65 1564

133
all of the predictors in this equation are nested in the budget institutional model.
Importantly, the series of such models will provide some answers to the research
questions. First, they will allow for the testing of the individual impact of each variable
by showing whether a statistically significant relationship exists between each variable
and each fiscal policy outcome. Furthermore, they will show whether the relationship is
as expected based on theory and whether the relationship is substantively meaningful.
Second, the models will also permit a comparison of the relative influence of each
category of variables in explaining total variation of the fiscal policy outcomes (e.g. how
much variation is collectively explained by the economic variables compared to the
political variables, etc.) by showing the change in the variance of the outcome variables
explained by the inclusion of each additional category of variables.
Nevertheless, there are a number of shortcomings to the empirical models
presented above. First, given the cross-sectional nature of the data only differences
between groups (i.e. countries) can be measured and not change within them. Moreover,
one must rely on the differences that already exist between groups since one does not
observe countries prior to the “treatment” of budget institutions. In the absence of
observations of the fiscal policy outcome variables prior to and following the
establishment of budget institutions it is impossible to determine wither observed changes
in government size are preceded and caused by changes in budget institutions.
However, even if were able to add a time dimension to our analysis, the fact
remains that budget institutions, like most political institutions, are costly to overturn and
consequently exhibit a degree of “institutional inertia” or “path dependency” (Pierson

134
2000). For instance, as Alesina and Perotti (2008: 15, emphasis added) state in there
discussion of the endogeneity of budget institutions,
…to the extent that institutions are reasonably difficult to change, and therefore
are changed relatively infrequently, they can be considered predetermined, at least
in the short to medium run. In other words, since it is costly and complex to
change institutions, the existing ones have to be very unsatisfactory before it is
worth changing them; as a result, there is a strong “status quo” bias in
institutions reforms.
113

Thus, even when substantial reform to budget institutions does occur, it is sufficiently
infrequent to limit the variation required for large-N analysis, such as Time-Series-Cross-
Sectional (TSCS) analysis. In fact, the studies that collected data on budget institutions
over a period of time (e.g. Von Hagen 1992, Von Hagen and Harden 1995; Alesina et al.
1999, Gleich 2003; Dabla-Norris et al. 2010; Gollwitzer 2011) always average over the
period covered for this precise reason.
One solution to this problem would be to expand a previous study by collecting
information on the same set of countries and then pooling the data. However, this
approach presents a number of additional difficulties. First, previous studies are based on
surveys the authors created and sent directly to the budget directors of the respective
countries. One could, in principle, recreate the surveys and send them out for another
round of questioning. However, one would then have to wait weeks, possibly months (or
even years) for a response, if the directors were to respond at all. As an alternative, one
could examine the respective budget laws of the sample countries. Still, one would only
be able to examine budget procedures according to the letter of the law rather than budget
procedures as they are practiced. In addition, attempting to obtain information

113
See also Dabla-Norris et al. (2010: 22); Gollwitzer (2011: 148).

135
independently would severely limit the amount of information that one could collect
independently compared to surveys (Alesina et al. 1999).
114

Thus, given these limitations and the time constraints of this project, expanding a
previous study is not a feasible option. The cross-sectional analysis proposed above, on
the other hand, is feasible and is nevertheless consistent with the approaches found in the
existing literature (e.g. Von Hagen 1992; Von Hagen and Harden 1995; Alesina et al.
1999; Gleich 2003; Stein et al. 2008; Dabla-Norris et al. 2010; Gollwitzer 2011).

6. Conclusion
Chapter 2 began by discussing the importance of fiscal policy. The chapter then
introduced existing explanations to account for the differences in fiscal policy outcomes
observed across countries and within countries over time. That chapter made note of the
various weaknesses of these explanations and in Chapter 3 budget institutions were
introduced as a potential determinant of fiscal policy. Specifically, Chapter 3 theorized
the relationship between budget institutions and five fiscal policy outcomes and in the
process developed thirteen testable hypotheses.
The purpose of this chapter has been to explicate how these hypotheses will be
tested and how such an investigation will answer the overarching research questions
regarding political institutions and fiscal policy. Also discussed were the sources and
measures of the various economic and political control variables, and how the study will
be affected by case selection and missing data. In light of these issues, it was argued that
econometric analysis was an appropriate method of analysis. And while the method of

114
This is on account of the amount of time that would be required to collect such information as well as
the lack of the ability to gather information on actual practices.

136
analysis and some of the measures may be less than optimal, they are nonetheless
sufficient to provide a first approximation of the relationship between budget institutions
and fiscal policy outcomes. In addition, the aforementioned method and measures do
possess the desirable quality of being comparable to the measures and methods of similar
studies. They also possess the added advantage of feasibility.
In the chapters that follow the methods discussed in this chapter are employed to
test the hypotheses introduced in Chapter 3 against the data. Specifically, Chapter 5
examines the affect of budget institutions on the level and composition of government
expenditures. Chapter 6 examines the impact of these variables on deficits and levels of
debt. Finally, Chapter 7 summarizes the overall findings and discusses what contributions,
if any, have been made to advance the research agenda of the economic effects of
political institutions. Lastly, this chapter considers possible avenues of additional
research, given the limitations of this study and the remaining gaps in knowledge of this
topic.

137
CHAPTER 5: SIZE OF GOVERNMENT
AND THE COMPOSITION OF THE BUDGET

1. Introduction

In light of the theory developed in Chapter 3 and the methods discussed in the last
chapter, this chapter estimates the effects of budget institutions on fiscal policy outcomes.
I focus on the size of government and the composition of the budget. The chapter begins
with a brief discussion of the multiple imputation procedure. Specifically, I address the
extent of missing values among observations and variables, and the number of
imputations and the imputation model utilized to produce the data used in the empirical
estimations. The section also includes the results of the index construction; in particular
non-parametric tests of the perfect-substitutability assumption discussed in Chapter 4.
The results show these assumptions to be innocuous. I then turn to the various empirical
tests.
The theory developed in Chapter 3 put forth specific hypotheses about the effects
of budget institutions and a variety of fiscal policy outcomes. Fiscal rules that place limits
on expenditures and/or revenues are expected to lead to smaller governments. Similarly,
more transparent budget institutions and hierarchical procedural rules are expected to
lower expenditure levels. Regarding the composition of the budget, theory suggests that
fiscal rules and transparency should have no discernable effect on spending priorities.
115

Conversely, hierarchical procedures are expected to lower spending on transfers and
subsidies and increase expenditures on public goods.

115
Throughout this chapter, I use “composition of the budget” and “composition of expenditures”
interchangeably to refer to the levels of expenditure earmarked specifically for either transfers and
subsidies or for public goods.

138
The empirical findings are discussed in different sections based on the fiscal
policy outcome of interest. For each policy outcome I first estimate a baseline
“economic” model which will be compared against in latter parts of the chapter. I then
estimate a “political institutions” model comprised of variables theory and past studies
have suggested are determinants of fiscal policy outcomes. I then estimate the “budget
institutions” models that are the primary interest of this study. Each section concludes by
comparing the explanatory power of each model as well as a series of nested models
intended to provide a more comprehensive view of the determinants of size of
government and budget composition.
Section 3 considers the size of government. I find that procedural rules, and to a
lesser extent transparency, lower the size of government as expected. Fiscal rules,
however, do not affect government size, either directly, or conditioned on the level of
transparency, as expected. Section 4 evaluates the effect of budget institutions on budget
composition. Here, the results are more surprising. On the one hand, procedural rules are
shown to lower the level of spending on transfers and subsidies, as expected. However,
they do not increase the level of public goods expenditures as implied by theory.
Contrary to my expectation, fiscal rules reduce transfers and subsidies spending and
transparency lowers public good expenditures. Theory suggests that there should be no
relationship between these institutions and budget composition. Therefore, these
relationships require further explication.
Section 5 assesses the robustness of these findings to alternative specifications.
Section 5.1 examines the sensitivity of the results to alternative indicators of government
size (revenues) and the level of redistribution (social security and welfare spending).

139
Section 5.2 analyzes what effect, if any, alternative democracy indicators have on the
results. Finally, Section 5.3 divides the sample into OECD and Non-OECD sub-samples
and assesses whether the effects of budget institutions are conditioned by level of
development. Overall, the findings in sections 3 and 4 are generally robust to these
changes.

2. Preliminaries
As discussed in Chapter 4, the dataset used in this study suffers from a
considerable amount of missing data. Section 4 of that chapter briefly discussed the
potential causes and consequences of the problem and suggested multiple imputation as a
practical solution. The first half of this section delves into the specifics of the missing
data problem and presents the results of the imputations. This task is necessary since the
extent of missing data impacts the amount of inflation in the variance of the estimates,
thus potentially affecting the results of the subsequent econometric tests.
The second half of this section briefly discusses the results of the index
construction. The purpose of this section is to demonstrate the robustness of the indices to
alternative assumptions regarding the substitutability of the items comprising each index.
It will be shown that the assumption of “perfect substitutability” is plausible and thus the
use of the most parsimonious indices is permissible.

2.1 Missing Data and the Imputation Model
In general, missing data are a problem in this study, however one that is
manageable. Only 13% of the observations are complete; that is, they contain no missing

140
values on any of the variables. This means that 72 countries have missing data on at least
one variable.
116
Fortunately, most of the missing data is confined to a few variables and
most cases are missing only a few values. To illustrate, the majority of observations—
60%—are missing values for at most only two variables. Ninety percent of the cases are
missing values for five variables or less. Table 5.1 summarizes the extent of missing
values among observations.
Table 5.1
The Extent of Missing Values among Observations
# of Missing Frequency Percent Cumulative
Values % %
0 11 13.25 13.25
1 19 22.89 36.14
2 20 24.10 60.24
3 8 9.64 69.88
4 13 15.66 85.54
5 4 4.82 90.36
6 5 6.02 96.39
7 1 1.20 97.59
8 2 2.41 100.00
Total: 83 100.00

Regarding the variables themselves, only 13 are missing values. About half (six)
have less than 10 missing values. Two are missing at most 14 values. Four of the
remaining five variables are much more problematic. These variables are missing
between 23 and 37 values, or between 27% and 45% of values.
117
For future reference,
these variables and their number of missing values are summarized in Table 5.2.

116
These figures relate to the 83 countries that enter into the analyses. Thirteen countries have a
combination of Freedom House scores greater than or equal to 5, Polity IV scores of 5 or less, and are
considered “non-democracies” by both the Cheibub et al. (2010) and Boix et al. (forthcoming) indices.
Thus, these countries never enter into any of the econometric models and for the purposes of this study are
ignored. These countries are Cambodia, The Republic of Congo, Guinea, Haiti, Kyrgyzstan, Qatar, Rwanda,
Swaziland, Tajikistan, Tunisia, United Arab Emirates, Vietnam, and Zimbabwe. Hong Kong is not a
sovereign state and is therefore also ignored.
117
The fifth variable, for the relative size of the upper chamber (RELSUC), is technically missing 44 values,
however in reality these are just instances of unicameral legislatures and thus to which this measure should
not apply.

141
Attention should be drawn to the fact that all of the budget institutions indices
have values missing for observations as do all of the budget composition dependent
variables and the debt dependent variable. The transparency index and the budget
composition dependent variables are most problematic given their importance in this
study and the significant amount of missing values. It is important to keep in mind the
extent of missing values on these variables since the large number of imputations
required to complete these variables inflates the variance of the estimates in subsequent
analyses.
Table 5.2
Variables with Missing Observations
# Of Missing
Variable Observations % Missing
Debt 3 3.6
Fiscal Rules: Deb. 4 4.8
Fiscal Rules: Exp. 4 4.8
Polity IV 4 4.8
Fiscal Rules: Def. 5 6.0
Procedural Rules 7 8.4
Gini Coefficient 12 14.5
Fiscal Centralization 14 16.9
Transfers and
Subsidies 23 27.7
Public Goods 25 30.1
Social Security
and Welfare 26 31.3
Transparency 37 44.6
Relative Size of
Upper Chamber 44 53.0

Recall from Chapter 4 that multiple imputation (MI) was argued to be the best
among the methods available for handling missing data. Multiple imputation requires
that a specific “imputation model” be specified, which determines the source of
information as well as the method in which values are to be substituted for missing
observations. Furthermore, this process is to be repeated m number of times in order to

142
create m complete datasets from which to pool subsequent statistical estimates. Given
that 72% of the cases in the dataset contain missing data on one or more variables, m =
72 imputations were conducted.
In addition to the economic, political, budget institutions, and fiscal policy
dependent variables that will comprise the various analytical models, the imputation
model included nine auxiliary variables which are arguably correlates of variables with
missing values and/or correlates of missingness. The first auxiliary variable is a nominal
variable COLONY, which indicates whether a country was a former colony and, if so, the
name of the metropolitan state. Studies suggest that legal, economic, and political
institutions are to some degree inherited from a past colonial power and that these
institutions affect economic outcomes (Acemoglu et al. 2001). A second variable
indicates the geographic region (REGION) to which a country belongs since certain types
of institutions seem to be associated with some regions over others (e.g. presidentialism
within Latin America). In addition, the imputation model includes each country’s GINI
coefficient (GINI) as a measure of inequality and which correlates highly with the budget
composition variables.
The remaining six variables are “governance variables,” from the World Bank’s
Worldwide Governance Indicators database (Kaufmann et al. 2010). The purpose of
these variables is to capture the broader societal and institutional framework of each
country. These composite variables measure perceptions of: 1) the extent to which
citizens are able to participate politically within civil society (VOICE); 2) political
stability and the absence of violence/terrorism (STABILITY); 3) citizen perceptions of
government efficacy (EFFECTIVENESS); 4) regulatory quality and the promotion of

143
private sector development (REGULATORY); 5) rule of law, for example the quality of
contract enforcement, property rights, the likelihood of crime and violence, etc
(RULEOFLAW); and, 6) the degree to which political power is exercised for private gain
and/or captured by elite and private interests (CORRUPTION).
Inclusion of these variables in the imputation model is important for a number of
reasons. First, including these variables, whether or not they ever enter into an analytical
model, makes it less likely that correlates or causes of missingness will be omitted.
Second, if these variables are indeed correlated with variables with missing values, then
their inclusion may decrease the standard errors (Collins et al. 2001). This is particularly
important in cases of high missingness, since these instances lead to inflated errors
(Acock 2010). Thus, auxiliary variables may mitigate this bias.

2.2 Budget Institutions Indices: Spearman Rank Correlations
Prior to the analyses it is necessary to check the robustness of the budget
institution indices to the procedure of additively aggregating components. As mentioned
in the previous chapter, different indices of each budget institution were constructed with
different assumptions about the substitutability between components. Specifically,
indices for each institution were constructed assuming perfect substitutability (j = 1), low
substitutability (j = 0.5), and high substitutability (j = 2). Then, Spearman rank
correlations were conducted to test whether the indices were monotonically related.
118

118
Spearman rank correlation coefficient (ρ) is similar to Pearson’s correlation coefficient (r), however it is
more appropriate when the data are ordinal, or when the data do not satisfy the assumptions of normality,
homoskedasticity, and linearity. The procedure works by first converting the observations of each variable
into ranks and then calculating Pearson’s correlation coefficient on the ranked values of the data.
Coefficient values fall between -1 and +1 and depict whether the variables covary in a monotonic manner
(Acock 2010: 197).

144
The Spearman rank correlation coefficients for each of the indices under the various
substitutability assumptions are depicted in Table 5.3.
Table 5.3
Spearman Rank Correlations
j = 1 j = 0.5 j = 2
ρ
Fiscal Rules: Exp. 0.7775 0.6039
Fiscal Rules: Rev. 0.9575 0.9594
Fiscal Rules: Def. 0.8773 0.8264
Fiscal Rules: Deb. 0.8778 0.8376
Transparency 0.9725 0.9100
Procedural Rules 0.9440 0.9316

Overall, the rank correlations between each of the budget institutions are
relatively strong. The rank correlations of the transparency and procedural rules indices
and their alternative specifications are all over ρ > 0.90, indicating very strong and
positive relationships. The rank correlations for the fiscal rules indices for revenue,
deficits, and debt are also very strong, with coefficients well over ρ > 0.80. The index
with the lowest rank correlations is the fiscal rules index covering expenditures. The
strongest of the two correlations is between the “perfectly substitutable” index (j = 1) and
the “low substitutability” index (j =0.5), with a coefficient of ρ > 0.78. The coefficient
between the “perfectly substitutable” index and the “highly substitutable” index is the
lowest at ρ > 0.60. Nevertheless, these rank correlations are still “strong” (i.e. ρ ≥ 0.60)
and statistically significant. In fact, all of the rank correlations are strong, positive, and
significant at the 1% level (p < 0.01). Thus, the indices—even for fiscal rules covering
expenditures—are quite robust to the changes in specification. As a result, the index
corresponding to j = 1 is utilized throughout the rest of this study.

145
3. Size of Government
This section tests the theories described in the previous chapters to explain the
size of government, measured here as spending by the central government as a percentage
of GDP. It begins by first estimating a baseline “economic” model comprised of
covariates described in Chapter 4. Specifically, it examines the effects of population size
(LPOPTOT), level of economic development (LGDPPC), openness to trade (TRADE),
natural resources (RESOURCES and RESRENTS), as well as two demographic
characteristics, the proportion of the population between 15 and 64 years of age and the
proportion of the population over the age of 65 (POP1564 and POP65, respectively). The
results are reported in Table 5.4.
In the most parsimonious model, both demographic characteristics and openness
to trade are highly associated with government size; population total is weakly related to
government size. Increasing the proportion of citizens of working age reduces the size of
government by approximately 1% of GDP. Conversely, a 1% increase in the proportion
of the population over the age of 65 increases expenditures by about 1.23% of GDP.
Similarly, a 1% increase in the ratio of trade to GDP results in an increase in government
size of 0.071% of GDP. Larger populations also appear to have smaller governments,
which conforms to the theory outlined in the previous chapter (Alesina and Wacziarg
1998; Hallerberg et al. 2007). There, it was suggested that larger countries are associated
with economies of scale in public good provision and thus likely to spend less overall.
This appears to be the case; however, the evidence against the null hypothesis is weak (p
< 0.10). Lastly, the results suggest that government size is not meaningfully determined
by the level of development and natural resource abundance (both measures).

146
The next pair of models includes the regional and colonial indicator variables.
With these specifications, only two variables—the proportion of the population over 65
years and trade—remain statistically significant, with the estimates remaining relatively
the same. This time around, however, the proxy for the level of development becomes
significant in both models although it is not very precisely estimated. Curiously, the sign
of the estimate is the opposite of that expected. According to “Wagner’s Law,”
government expenditures are income-elastic and thus should increase with national
income (Cameron 1978). In any case this variable is highly unstable to model
specification.
119
Overall, the economic models find the strongest and most consistent
support for the arguments that the greater the proportion of the population above the age
of 65, and the greater the dependence on trade, the larger the size of government
(Wilensky 1976; Rodrik 1997).
Table 5.4
Size of Government: Economic Models
Model (1) (2) (3) (4)
LPOPTOT -1.01 -0.99 -0.79 -0.87
(0.59)* (0.59)* (0.69) (0.70)
POP1564 -0.54 -0.54 -0.46 -0.43
(0.22)** (0.22)** (0.30) (0.30)
POP65 1.23 1.23 1.28 1.31
(0.26)*** (0.24)*** (0.33)*** (0.32)***
LGDPPC -0.25 -0.28 -2.23 -2.26
(0.75) (0.76) (1.14)* (1.14)*
TRADE 0.07 0.07 0.07 0.07
(0.02)*** (0.02)*** (0.03)*** (0.03)***
RESOURCES -0.15 0.06
(2.04) (2.37)
RESRENTS -0.02 0.07
(0.11) (0.13)
Region No No Yes Yes
Colony No No Yes Yes
Number Of 83 83 83 83
Observations
Adjusted R
2
0.40 0.44 0.49 0.49
* significant at 10%; ** significant at 5%; *** significant at 1%.
Standard errors in parentheses.

119
Other empirical tests of “Wagner’s Law” have also failed to conform to expectations. See, for example,
Gupta (1967), and Bird (1971).

147
The next section tests the theories described in Chapter 2 regarding the effects of
political institutions on the size of government. The results are presented in Table 5.5.
There are eight models, which can be separated into three categories: models 1, 2, 5, and
6 estimate the effect of bicameral legislatures, whereas models 3, 4, 7, and 8 split the
sample into unicameral and bicameral legislatures and estimate the effect of the relative
size of the upper chamber among the latter. Secondly, all odd-numbered models control
for electoral systems by including indicator variables for countries that are: a)
majoritarian and presidential; b) majoritarian and parliamentary; and, c) proportional and
presidential, with a residual category for countries that are proportional and parliamentary.
All even-numbered models include indicator variables for only majoritarian and
presidential systems, with residual categories for proportional representation and
parliamentary systems, respectively. Finally, models 5 through 8 include the regional and
colonial controls from before, whereas models 1 through 4 do not.
First, note that the variables for the total number of legislative seats, the relative
size of the upper legislative chamber, the level of government fractionalization, the size
of the largest political party, and the indicator for federal systems are never statistically
significant at conventional levels in any of the models. Before moving forward, however,
a few statistical incongruities deserve to be addressed. Note first that the sign of
TOTSEATS is negative, when theory suggests that expenditures should increase with
each additional legislative representative (e.g. Weingast 1979). A simple correlation,
however, shows legislative size to be very weakly and insignificantly related to
government size (r = -0.064 p = 0.56), which further supports that its estimated effect is
indistinguishable from zero.

148

Table 5.5
Size of Government: Political Institutions Models
Model (1) (2) (3) (4) (5) (6) (7) (8)
TOTSEATS -0.002 -0.003 -0.004 -0.004 -0.002 -0.003 0.02 0.01
(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)
BICAMERAL 0.73 0.80 0.85 0.89
(1.95) (1.96) (1.98) (1.98)
RELSUC 2.14 2.05 8.83 15.42
(13.77) (13.29) (17.99) (16.48)
GOVFRAC -2.49 -3.40 -11.47 -11.57 -2.17 -3.96 -3.84 3.98
(6.03) (6.02) (12.01) (11.1) (6.74) (6.48) (13.67) (10.79)
LGSTPARTY -7.42 -9.14 -22.48 -22.6 4.92 1.74 -10.39 2.96
(8.15) (8.09) (17.48) (16.36) (10.01) (9.48) (30.04) (25.97)
ENLP 0.22 0.22 0.39 0.40 0.90 0.95 3.04 2.71
(0.75) (0.76) (1.15) (1.11) (0.77) (0.77) (1.32)** (1.26)*
MAJ -2.33 -2.06 -2.32 0.82
(1.81) (3.32) (1.93) (3.48)
PRES -10.06 -12.01 -7.32 -4.85
(1.73)*** (3.16)*** (2.50)*** (4.54)
MAPRES -11.53 -14.03 -9.07 -3.68
(2.42)*** (4.42)*** (3.01)*** (5.45)
MAJPAR -4.42 -2.16 -3.84 4.72
(2.38)* (4.77) (2.45) (5.37)
PROPRES -11.76 -12.07 -9.00 -0.72
(2.15)*** (3.73)*** (3.04)*** (6.48)
FEDERAL -0.50 -0.82 1.45 1.45 -0.96 -0.96 3.49 2.87
(2.88) (2.87) (4.35) (4.26) (2.84) (2.83) (4.25) (4.18)
FISCCEN 13.57 13.42 23.94 23.97 17.4 17.80 48.77 39.59
(8.59) (8.52) (14.86) (14.07) (8.59)** (8.53)** (17.05)** (13.91)**
Regions No No No No Yes Yes Yes Yes
Colonies No No No No Yes Yes Yes Yes
Number of 83 83 39 39 83 83 39 39
Observations
Adjusted R 0.34 0.33 0.28 0.31 0.44 0.44 0.59 0.59
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

Furthermore, the signs of the coefficients for GOVFRAC and LGSTPARTY are
also in the direction opposite of that expected in a number of models. Both are negative
when theory suggests that as political fractionalization within the executive and as the
number of legislative seats of the largest political party increases, so should expenditures
(Inman and Fitts 1990; Perotti and Kontopoulos 2002). Diagnostic tests reveal that there

149
may be an issue with multicollinearity between GOVFRAC and LGSTPARTY (r = -0.68),
GOVFRAC and ENLP (r = 0.71), and between LGSTPARTY and ENLP (r = -0.58).
120

Only when LGSTPARTY is removed from the estimation equation does the coefficient for
GOVFRAC become positive. Nevertheless, its effect still remains statistically
indistinguishable from zero. Regarding LGSTPARTY, regardless of whether GOVFRAC
and/or ENLP are removed from the estimation, it remains insignificant and has the
incorrect sign. According to Inman and Fitts (1990), the effect of the size of the largest
legislative party may be non-linear and there may be diminishing returns to size. When
the models are re-estimated with LGSTPARTY and a quadratic term, LGSTPARTY2, the
signs of both coefficients are in the wrong direction and also statistically insignificant.
Finally, an incremental F-test fails to reject the null hypotheses that the coefficients for
all three variables are zero (i.e. H
0
: β
GOVFRAC
= β
LGSTPARTY
= β
ENLP
= 0). Therefore, it is
safe to assume that multicollinearity (and in the case of LGSTPARTY, possible non-
linearity) is not responsible.
Lastly, observe that the sign for federal is incorrect in the subset of models
conditioned on bicameral legislatures. FEDERAL is moderately correlated with
FISCCEN (r = -0.59) but once more not at a level that would be cause for concern. In
addition, the effect of federal systems remains insignificant when we exclude the fiscal
centralization variable.
121
Here too, an F-test cannot reject the null hypothesis of a
coefficient indistinct from zero. The cause of this change is likely due to the small sample
size when the estimates are conditioned on bicameral legislatures (N = 39). Note that the

120
Multicollinearity is also suggested by the incorrect signs, small t-ratios, and high standard errors of these
variables.
121
The latter (FISCCEN) becomes significant when FEDERAL is excluded, however only at the 10% level.
Since their correlation coefficient is not considered dangerously high, this effect is likely inflated due to
omitted variable bias.

150
standard errors are substantially higher—in some cases nearly double the full sample—
thus likely causing highly unreliable estimates.
122

Based on the above, and given that the correlation coefficients between the
variables do not exceed the threshold of r = 0.80 suggested by Studenmund (2006) to be
problematic, I decide to do nothing and leave theses variables in subsequent models as
explanatory variables. To drop one or another would likely lead to omitted variable bias.
Doing so would lead to biased coefficient estimates and artificially small standard errors,
a problem much worse than multicollinearity.
The variables with robust effects statistically significant from zero are MAJPRES,
PROPRES, and PRES. These variables are highly significant (p < 0.01) in three of the
four models in which they appear. In the full-sample, full-control model (model 5), joint
majoritarian and presidential systems spend about 9% less than joint proportional and
parliamentary systems. A nearly identical effect is found for joint proportional and
presidential systems. When systems are more broadly separated into majoritarian vs.
proportional and presidential vs. parliamentary, we observe that presidential systems
spend less. Specifically, the size of government in presidential systems is 7.32% less than
in parliamentary ones.
123
The models do not distinguish, however, between the effects of
majoritarian and proportional systems. Not surprisingly, these results largely conform to
those reported in Persson and Tabellini (2003b).
Fiscal centralization is the only variable that demonstrably increases the size of
government. As the ratio of central government to general government spending

122
This may also explain the switching of signs for the coefficients for MAJPAR in model 7 and MAJ in
model 8. Small sample sizes also increase variability and thus higher effects. The latter may be evident in
the effect for ENLP. Note that the standard errors are larger, as are the coefficients, which leads to ENLP
becoming statistically significant in models 7 and 8.
123
This figure refers to the full-sample, full-control model 6.

151
increases by 1%, the size of government increases by 0.17% and 0.18%, respectively.
124

Put differently, if a completely fiscally-decentralized country became a fully fiscally-
centralized country, the size of its government would increase somewhere between
17.4% and 17.8%
125
Thus, this evidence of favors the “Leviathan” theory of Brennan and
Buchanan (1977, 1978, 1980) over the “fiscal federalism” theory of Wildavsky (1974),
Freemen (1975), and Tarschys (1975).
126

What effects do budget institutions have on the size of government? Table 5.6
reports the results of the empirical tests. Of the three budget institutions, only the effect
of procedural rules is statistically significant in the two specifications. In the
parsimonious model, an increase of one unit of the procedural rules index reduces the
size of government by 1.29% of GDP. The estimate is relatively stable when controlling
for regional location and colonial origin. In the second model, the estimate drops only
slightly to 1.15%. Furthermore, in both instances, the precision of the estimate is fairly
high (p < 0.001).
Conversely, the index for transparency is only statistically significant when
geographical location and colonial origin are controlled for. This is encouraging, given
that it is a more comprehensive model and thus more likely to satisfy the conditional
independence assumption (Wooldridge 2010). According to the results depicted in model
2, an increase in the transparency index by one unit lowers the size of government by
0.32 % of GDP. Nevertheless, the estimate is not precisely estimated (p < 0.10).

124
The variable FISCCEN is measured as a proportion. Thus, a coefficient of 17.4 (model 5) implies an
increase in expenditures of 0.174 (17.4/100) for a 1% increase in fiscal centralization.
125
Here too, the small sample sizes in models 7 and 8 are inflating the effects of fiscal centralization. For
this reason they are ignored.
126
See Chapter 2.

152
Compared to the other two budget institutions, neither test supports the hypothesis
that fiscal rules limiting expenditures reduces the size of government. However, in
Chapter 3 it was argued that strict fiscal rules may lead to greater effort on the part of
politicians to resort to “creative accounting” in order to circumvent such regulations (Von
Hagen 1991). Therefore, it was posited that fiscal rules would only be effective provided
the level of transparency was sufficiently high. Nevertheless, when we separate the
sample according to the level of transparency, there remains no distinguishable effect of
fiscal rules.
127
This is true whether transparency is set to the 25
th
, 50
th
, or 75
th

percentile.
128
Thus, hypothesis 3 cannot be confirmed.
How well do budget institutions compare with the alternative explanations for
government size? One way to compare theories, given that the dependent variable is
identical in all models and only the independent variables change, is to compare adjusted
R
2
. This enables a comparison of “goodness of fit,” where R
2
can be considered a
measure of “the proportional reduction in error from the null model (with no explanatory
variables) to the current model” (King 1986: 677).
129
By this criterion, budget institutions
fare pretty well. Comparing fully controlled models, one finds that the economic
baseline model (model 4, Table 5.4) explains 49% of the variance, the political
institutions model (model 5, Table 5.5) explains 44% of the variance, and budget
institutions (model 2, Table 5.6) explain 50% of the variance. As a first approximation,

127
Given that there is nothing substantial to report, I omit publishing these results.
128
Above this level, not only would fiscal rules be meaningless but there are also not enough cases in the
sample that satisfy this requirement.
129
I use adjusted R
2
because in small samples with several predictors R
2
often exaggerates the strength of
the relationship. This is because R
2
may increase with each additional variable simply by chance. Thus,
adjusted R
2
is the more appropriate statistic since it attempts to remove any chance effects (Acock 2010).
Furthermore, the evidence suggests that such an adjustment is warranted. For example, while the adjusted
R
2
for the budget institutions model is 0.50, the unadjusted R
2
is 0.61, an increase of 22%.

153
therefore, the motivation for emphasizing budget institutions over alternative
explanations—at least in terms of expenditures—appears justified.
Table 5.6
Size of Government: Budget Institutions Models
Model (1) (2)
FISCRULES_E -0.42 -0.23
(0.29) (0.30)
TRANSPARENCY -0.21 -0.32
(0.17) (0.16)*
PRORULES -1.29 -1.15
(0.26)*** (0.27)***
Region No Yes
Colony No Yes
Number of 83 83
Observations
Adjusted R
2
0.31 0.50
* significant at 10%; ** significant at 5%; *** significant at 1%.
Standard errors in parentheses.

As a further comparison it is possible to examine how much additional variance is
explained by budget institutions compared to other political institutions over the
economic baseline utilizing a series of nested models. Table 5.7 depicts these comparison
tests using economic model 4 from Table 5.4 as the base. Model 1 is the “political
economy” model which shows the unique effect of the political institutions variables
beyond what is explained by the economic variables. The results show that POP1564,
POP65, MAJPRES, PROPRES, and FISCCEN are all statistically significant at the 1%
level, TRADE, RESRENTS, and LGSTPARTY (correctly signed) are statistically
significant at the 5% level, and MAJPAR is statistically significant at the 10% level.
Furthermore, the model explains 70% of the variance in government size: adjusted R
2
=
0.70, F (31, 48.8) = 6.63, p < 0.001. Compared to the economic model, this represents a
21% increase in adjusted R
2
(0.70-0.49 = 0.21) when the political institutional variables
are added. Performing a Wald test shows that the set of political institutions variables

154
Table 5.7
Size of Government: Nested Models
Model (1) (2) (3)
LPOPTOT 0.26 -0.80 0.09
(1.04) (0.65) (1.03)
POP1564 -1.08 -0.08 -0.72
(0.27)*** (0.29) (0.32)**
POP65 1.46 1.05 1.34
(0.29)*** (0.30)*** (0.28)***
LGDPPC -1.14 -2.68 -1.40
(1.00) (1.06)** (0.98)
TRADE 0.06 0.04 0.04
(0.02)** (0.03)* (0.02)
RESRENTS 0.24 0.004 0.16
(0.12)** (0.12) (0.12)
TOTSEATS -0.003 -0.002
(0.01) (0.01)
BICAMERAL -0.16 -0.37
(1.58) (1.65)
GOVFRAC 1.03 1.33
(5.31) (5.23)
LGSTPARTY 18.16 15.08
(8.24)** (8.25)*
ENLP 0.76 0.74
(0.61) (0.61)
MAJPRES -10.1 -6.73
(2.66)*** (3.06)**
MAJPAR -3.21 -2.27
(1.89)* (1.90)
PROPRES -6.78 -4.89
(2.47)*** (2.62)*
FEDERAL -1.44 1.87
(2.39) (2.39)
FISCCEN 24.11 23.62
(7.05)*** (7.22)***
FISCRULES_E -0.33 -0.10
(0.27) (0.24)
TRANSPARENCY -0.23 -0.16
(0.15) (0.14)
PRORULES -0.89 -0.54
(0.26)*** (0.27)*
Average RVI 0.0842 0.1186 0.1601
Region Yes Yes Yes
Colony Yes Yes Yes
Number of 83 83 83
Observations
Adjusted R
2
0.70 0.62 0.73
* significant at 1%; ** significant at 5%; ***significant at 1%.
Standard errors in parentheses.

155
simultaneously are indeed statistically significant F(10, 48.8) = 4.91, p < 0.001. Thus, it
is possible to state that the effect of political institutions is moderately strong and
statistically significant.
The same tests are applied to budget institutions and are reported in the second
column of Table 5.7. Once more, only PRORULES is significant among the budget
institutions, again at the 1% level. Among the economic variables POP65 is statistically
significant (p < 0.01), as is LGDPPC (p < 0.05) and TRADE (p < 0.10). Comparatively,
the budget institutions model explains 62% of the variance in government size: adjusted
R
2
= 0.62, F(24, 55.6) = 5. 42, p < 0.001. This represents a 13% improvement in
adjusted R
2
(0.62 – 0.49 = .13), but it is much less than the effect of the political
institutions variables. Nevertheless, a Wald test shows that the effect of the set of budget
institutions, while moderate, is nonetheless statistically significant: F(3, 52.7) = 4.87, p <
0.01.
Finally, column 3 reports the results of the full model combining the effects of the
economic, political institutions, and budget institutions variables. The three sets of
variables explain 73% of the variance (adjusted R
2
= 0.73 F(34, 45.5) = 6.12, p < 0.001).
There are a few important things to note. First, PRORULES remains the only statistically
significant budget institution index; however, its precision has dropped substantially (p <
0.10). Second, adding budget institutions to the political economy model only improves
the adjusted R
2
by 3%. A Wald test reveals that removing the budget institutions
variables from the model would not adversely affect the model’s fit (F(3, 43.1) = 1.75, p
> 0.10).
130
In other words, once the economic and political institutions variables are

130
The economic and political institutions variables, on the other hand, are statistically significant (F(6,
45.4) = 4.73, p < 0.001; F(10, 45.4) = 2.88, p < 0.01, respectively).

156
controlled for, the budget institutions variables do not help very much in predicting the
size of government.
What may explain the conflicting results? Note that Table 5.7 reports a new
statistic, “Average RVI”. This figure is an estimate of the average relative inflation in
variance of the estimates as a result of missing values (Acock 2010). This inflation has
the effect of increasing standard errors, thus reducing t-values. Ideally, this estimate
should be as close to zero as possible. Indeed this is the case for the economic base model
(Average RVI = 0.00). Since FISCCEN had missing values for 16.9% of the observations,
the Average RVI for the political economy model is greater than zero but only 0.08.
Compare this figure to the Average RVI of the combined economic/budget institutions
model (Average RVI = 0.12) and the full model (Average RVI = 0.16).
131
In other words,
the standard errors of the budget institutions in all of the models may be higher than they
otherwise would be had there not been missing data. Recall that all of the budget
institutions are missing values (4.8% for expenditure fiscal rules, 8.4% for procedural
rules, and 44.6% for transparency). Thus, the insignificant results may be on account of
the inflated variance as a result of missing values. However, until it is possible to gather a
complete set of data, this conjecture cannot be confirmed.
To summarize, the empirical tests find strong support for hypothesis 11 in that
hierarchical budget procedures were demonstrated to lower the size of government. The
tests found some support for hypothesis 7, specifically model 2 (Table 5.6), which
suggested that higher levels of transparency are associated with lower levels of
expenditures. However, the evidence was not strong. Finally, the tests did not support the

131
While not reported, note that the Average RVI for the two budget institutions models in Table 5.6 is
0.42 and 0.17, respectively. The Average RVI for the political institutions model (model 5 Table 5.5),
conversely, is 0.03.

157
proposition that a fiscal rule limiting expenditures leads to smaller governments. No
effect was found directly, nor conditioned upon the level of transparency, as specified by
hypothesis 3. Nevertheless, comparison tests revealed that the effect and significance of
budget institutions is sensitive to the specific econometric specification. When
controlling for geographic location and colonial origin, budget institutions seem to
explain at least as much of the variation as the economic variables and more than the
political variables. However, when budget institutions are combined with the economic
variables and/or both economic and political institutions variables, their impact is
substantially reduced and they appear unable to add much more to the explanation of
government size. While this may be accounted for by the amount of missing data that
required imputation, at the present time this is just speculation and can be neither rejected
nor confirmed.

4. Budget Composition
Thus far, the evidence suggests that demographics, trade, the electoral system and
form of government, the degree of fiscal centralization, and hierarchical procedural rules
most strongly predict government expenditures. The question this section seeks to answer
is, “what variables best explain the composition of expenditures?” Two types of
expenditures, in particular, are examined here. The first is transfers and subsidies to
households and firms. As discussed in Chapters 3 and 4 this measure serves as a proxy
for the level of redistribution. The second type is expenditure on public goods provision.
Unlike transfers and pork-barrel projects (or local public goods) the benefits of public
goods are broadly dispersed and do not favor any particular narrow constituency

158
(Mukherjee 2003).
132
Thus, the objective of this section is to determine what variables are
most closely associated with targeted spending to special interests and specific
socioeconomic and/or demographic groups, and what variables are most closely
associated with expenditures that more or less benefit society as a whole.
Table 5.8 reports the results from the baseline economic model. Models 1 through
4 estimate the effects of the economic variables on transfers and subsidies. These models
also control for the level of income inequality (measured by the Gini coefficient). Theory
suggests that income inequality increases the level of redistributive transfer payments
(Meltzer and Richard 1981, 1983). The only covariate in the sample that has any effect
on transfers and subsidies is the proportion of the population over 65 years of age. This
result is similar to that found in Milesi-Ferretti et al. (2002). Based on these estimates, the
effect of a 1% increase within the retired population increases expenditures on transfers
and subsidies between 0.84% and 0.91% of GDP. Thus, with an average sample
population of approximately 10.8 million, as an extra 108 thousand people enter
retirement, governments, on average, spend an additional 0.87% of GDP on transfers and
subsidies, all else constant. Combined with the effect on expenditures, these results
confirm Wilensky’s (1976) insight on the strategic importance of the elderly when it
comes to public spending.
Unfortunately, the results for public goods spending (models 5 and 6) are much
less easy to interpret. On the one hand, only population total has any discernable effect,
however it is imprecisely estimated (p < 0.10). Furthermore, it is difficult to reconcile

132
I follow Lizzeri and Persico (2001) in making the distinction between public goods and pork-barrel
projects. The benefits of the latter are concentrated, whereas the benefits of the former are widely dispersed.
In fact, another way to think of pork is as a nonmonetary means of redistribution; one that is disguised as
public goods. Thus, for all intents and purposes, “public goods” are understood as producing benefits that
cannot be easily targeted to subsets of the population.

159
logically why larger populations spend less on public goods. In any case, the estimates
are not robust to the inclusion of the regional location and colonial origin controls. The
results (not reported) show that when these controls are included in the models the
variables predict public goods spending no better than the null model (with no
explanatory variables).
133
Also, note the size of the average RVI in each of the models.
134

Given this information, it is safe to assume that, with the exception of POP65, the
economic data utilized in this study do not predict public goods spending well.
Table 5.8
Composition of Expenditures: Economic Models
Dependent Variable: Transfers and Subsidies Public Goods
Model: (1) (2) (3) (4) (5) (6)
LPOPTOT -0.29 -0.24 -0.47 -0.45 -1.08 -1.12
(0.37) (0.37) (0.42) (0.43) (0.58)* (0.59)*
POP1564 -0.25 -0.24 -0.15 -0.16 0.06 0.06
(0.15) (0.15) (0.19) (0.20) (0.22) (0.22)
POP65 0.87 0.84 0.91 0.88 -0.03 -0.03
(0.17)*** (0.16)*** (0.19)*** (0.19)*** (0.24) (0.22)
LGDPPC 0.17 0.15 -0.51 -0.47 -1.07 -1.10
(0.44) (0.45) (0.72) (0.73) (0.66) (0.66)
TRADE 0.01 0.01 0.005 0.004 0.03 0.03
(0.01) (0.01) (0.02) (0.02) (0.02) (0.02)
RESOURCES 0.51 0.80 -1.31
(1.44) (1.67) (2.06)
RESRENTS -0.03 -0.01 0.01
(0.07) (0.09) (0.11)
Average RVI 0.6277 0.6842 0.625 0.6065 0.3717 0.3516
Region No No Yes Yes No No
Colony No No Yes Yes No No
Gini Yes Yes Yes Yes No No
Number of 83 83 83 83 83 83
Observations
Adjusted R
2
0.64 0.64 0.69 0.69 0.17 0.16
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

133
When the models include RESOURCES the F-statistic is F(21, 57.5) = 1.16, p = 0.3174. When instead
RESRENTS is substituted into the model, it fares worse, with an F-statistic of F(21, 57.5) = 1.10, p =
0.3766. This, and the fact that the addition of the controls does not substantively increase the R
2
, suggests
that the regional and colonial control variables may be extraneous when estimating the effect of economic
variables and spending on transfers and subsidies. Because no information is gleaned from these models, I
do not report the results.
134
Recall that the variable for public goods was initially missing values for 30% of the observations.

160
The regressions with political institutions perform noticeably better. The results of
these tests are depicted in Table 5.9. As before, models 1 and 2 control for bicameral
legislatures while models 3 and 4 examine the effect of the relative size of the upper
chamber to the lower chamber within bicameral legislatures. The only variables that enter
the transfers and subsidies models at statistically significant levels are the electoral
variables.
135
Majoritarian and presidential systems, proportional and presidential systems,
and to a lesser degree, majoritarian and parliamentary systems are found to spend less on
transfers and subsidies than proportional and parliamentary systems. In the case of
MAJPRES and PROPRES the effect is to spend approximately 4.5% less than
proportional and parliamentary regimes. Majoritarian and parliamentary systems spend
about 3% less. Models 2 and 4, which break down the electoral variables into simply
majoritarian vs. proportional and presidential vs. parliamentary comparisons, suggest that
these effects are largely driven by presidential regimes. These forms of government
spend about 3% less than parliamentary governments (in the full sample). Nevertheless,
the effect does appear additive since combining presidentialism with either electoral rule
increases the net effect. These results largely conform to the findings in Milesi-Ferretti et
al. (2002) and Persson and Tabellini (2003b). Yet, the results are highly dependent upon
specification. When the regional and colonial controls are introduced (model 5), only the

135
Observe that the coefficients for GOVFRAC and ENLP are in the direction opposite of expected. Once
more, these two variables correlate with one another. Removing one variable or the other from the equation
reverses the effect; however, in either case the variables remain insignificant. The fact that ENLP is
insignificant is somewhat surprising given the findings of Mukherjee (2003), whose results indicate that
ENLP is significantly associated with increased transfer and subsidies spending. One possible explanation
is the poor precision of the estimates as indicated by the standard errors of the variable and average RVI of
the models. The null result of GOVFRAC however, is consistent with Mukherjee (2003).

161
indicator for proportional and presidential systems remains statistically significant, but it
is not estimated very precisely (p < 0.10).
136

Table 5.9
Composition of Expenditures: Political Institutions Models
Dependent Variable: Transfers and Subsidies Public Goods
Model: (1) (2) (3) (4) (5) (6) (7)
TOTSEATS 0.00 0.00 0.00 0.00 0.00 -0.01 -0.01
(0.003) (0.004) (0.005) (0.005) (0.004) (0.01)* (0.004)*
BICAMERAL 0.14 0.24 0.92 1.53 1.51
(1.35) (1.36) (1.37) (1.73) (1.72)
RELSUC -6.50 -7.72
(8.01) (7.80)
GOVFRAC -1.42 -2.09 -0.10 -1.98 -0.12 1.20 1.42
(4.24) (4.25) (7.14) (6.84) (4.52) (5.57) (5.59)
LGSTPARTY -5.19 -6.10 -8.38 -10.29 -0.74 -2.80 -2.39
(6.15) (6.17) (11.73) (11.48) (7.66) (7.58) (7.58)
ENLP -0.10 -0.09 -0.22 -0.13 0.02 -0.23 -0.23
(0.56) 0.56 (0.72) (0.72) (0.55) (0.75) (0.75)
MAJ -1.69 -1.62 1.30
(1.28) (2.19) (1.72)
PRES -3.33 -4.59 -2.36
(1.39)** (2.23)* (1.56)
MAJPRES -4.62 -5.65 -2.85 -1.27
(1.94)** (2.85)* (1.99) (2.44)
MAJPAR -2.84 -3.12 -2.10 1.78
(1.56)* (2.77) (1.59) (2.10)
PROPRES -4.40 -5.64 -4.05 -1.97
(1.61)*** (2.55)** (2.07)* (1.86)
FEDERAL -0.42 -0.56 -1.93 -1.77 -1.51 -0.02 0.04
(1.91) (1.91) (2.51) (2.51) (1.86) (2.44) (2.43)
FISCCEN -6.31 -6.26 -10.68 -9.02 -3.59 15.14 15.12
(5.51) (5.49) (8.48) (8.23) (5.32) (7.27)** (7.25)**
Average RVI 0.4348 0.4364 0.2359 0.2429 0.4906 0.3627 0.3654
Region No No No No Yes No No
Colony No No No No Yes No No
Gini Yes Yes Yes Yes Yes No No
Number of 83 83 39 39 83 83 83
Observations
Adjusted R
2
0.47 0.46 0.40 0.40 0.58 0.16 0.17
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

With regard to public goods spending, the results show that more fiscally
centralized systems spend more on public goods and relatively large legislatures spend

136
When the additional controls are included in models 2 through 4 there are no statistically significant
results. Therefore, in the interest of space I do not report these figures.

162
less. However, their economic effect is quite small: one fifteenth of one percent in the
case of fiscal centralization and one hundredth of one percent in the case of total seats in
the legislature. Overall, the small proportion of variance explained in both models
suggests that the political institutions variables do not predict public good spending well.
Moreover, the effects not only disappear but the fit of the models become insignificant
when the geographic location and colonial origin controls are added (not shown). This
result is not entirely unexpected since other studies have had trouble predicting levels of
public goods spending with political institutional variables as well (e.g. Milesi-Ferretti et
al. 2002).
137
Nevertheless, to my knowledge, finding an effect between fiscal
centralization and public goods spending is new, and thus supports the suggestion made
in Chapter 2 for testing the effects of political institutions on an exhaustive list of fiscal
policy outcomes. This is an avenue for future research.
Do the effects of budget institutions extend to expenditure composition? Table
5.10 reports on the linear regression estimates. The results show mixed support for the
hypotheses put forward in Chapter 3. With regard to fiscal rules, hypothesis 4 posited that
there would not be a direct effect on budget composition. This was based on the idea that
fiscal rules place limits on aggregates and not spending priorities. Therefore, it was
argued that budget composition is determined by other factors. This appears to be the
case in terms of public goods spending. Neither model (3 or 4) found an affect
statistically different from zero. However, note that the average RVI are high, suggesting
inflated standard errors. Thus, it may be that fiscal rules do affect spending priorities; yet,
given the data, the null hypothesis cannot be rejected. This is not necessarily the case,
however, in terms of spending on transfers and subsidies. Model 1 found an effect of

137
However, see Mukherjee (2003).

163
fiscal rules on transfers and subsidies. According to the results, for an increase in the
budget institutions index of one unit, spending on transfers and subsidies falls by 0.35%
of GDP, all else constant. Nevertheless, the approximation is imprecisely estimated (p <
0.10) and is not robust to specification since the effect disappears once additional controls
are accounted for.
138
To my knowledge, the relationship between fiscal rules and budget
composition has not been thoroughly explored. This remains an open area for further
investigation.
The results are mixed for the other two budget institutions as well. In terms of
transparency, the regressions partially confirm hypothesis 8. It appears that transparency
has no effect on transfers and subsidies spending but decreases the level of expenditures
on public goods. One of the postulates developed in Chapter 3 posited that transparency
would not directly impact expenditure composition. Again, because expenditure
composition is largely determined by the spending priorities of the various policymakers
involved in drafting and approving the budget, there is no obvious reason to expect
transparency to influence the decision to spend more or less on transfers and subsidies or
more or less on public goods. And yet, at least in this study’s sample, more transparent
budget institutions are associated with less public goods spending. Specifically, ceteris
paribus, for a unit increase in budget transparency, public goods spending declines
between 0.45% and 0.47% of GDP. Moreover, the effect is robust and precisely
estimated (p < 0.01). This finding is even more impressive given that the average RVI of
the models are high since nearly 30% of both composition dependent variables were
missing observations. However, once more there is little theoretical guidance to explain
this phenomenon.

138
Note, however, the high average RVI of the models.

164
Lastly, an imprecise but nonetheless statistically significant effect was found in
one model (1) between procedural rules and transfers and subsidies (p < 0.10).
Hypotheses 13 posited that the more hierarchical the procedural rules, the less that would
be spent on transfers and subsidies. According ot the findings, this may be the case.
However, the result is not robust to the inclusion of additional controls. In terms of public
goods spending, hypothesis 12 argued that hierarchical procedural rules would be
associated with increased public goods spending. However, in no specification does this
appear to be the case. Perhaps a relationship will be found with better data, but at the
present time there is no support for hypothesis 12.

Table 5.10
Composition of Expenditures: Budget Institutions Models
Dependent Variable: Transfers and Subsidies Public Goods
Model: (1) (2) (3) (4)
FISCRULES_E -0.35 -0.18 -0.17 -0.23
(0.20)* (0.22) (0.27) (0.30)
TRANSPARENCY 0.11 0.04 -0.47 -0.45
(0.10) (0.09) (0.13)*** (0.14)***
PRORULES -0.31 -0.28 -0.14 -0.11
(0.18)* (0.17) (0.24) (0.26)
Average RVI 0.6913 0.5541 1.1408 0.6278
Region No Yes No Yes
Colony No Yes No Yes
Gini Yes Yes No No
Number of 83 83 83 83
Observations
Adjusted R
2
0.47 0.59 0.30 0.35
* significant at 10%; ** significant at 5%; *** significant at 1%.
Standard errors in parentheses.

165
How well do budget institutions compare to economic and political institutional
explanations for budget composition? On their own, budget institutions do not predict
transfers and subsidies spending any better than other political institutional explanations
(adjusted R
2
= 0.59 compared to an adjusted R
2
= 0.58, respectively). And, they perform
worse than economic explanations (adjusted R
2
= 0.69).
139
This poor performance is
exacerbated when comparing nested models. As reported in columns 1 through 3 in Table
5.11, the “political economy,” joint budget institutions and economic, and full, models do
not add to the variance explained by the economic model alone. In all of these models,
only the proportion of the population over the age of 65 has any statistical significance. In
fact, it is this variable that is doing all of the work. When controlling for regional location,
colonial origin, and level of inequality, POP65 alone explains 67% of the variance. Thus,
the low additional variance explained suggests that the political institutions and budget
institutional variables are largely irrelevant for predicting transfers and subsidies
spending once these other factors have been controlled for. As further evidence, Wald
tests for all models fail to reject the null hypotheses of no effects of the political
institutions and budget institutions variables.
140
However, given the high average RVI of
these models, any interpretation of these results warrants caution. Like the previous
comparisons, more work needs to be done with better data so that more accurate
estimations can be made.

139
These figures refer to model 4 of Table 5.8, model 5 of Table 5.9, and model 2 of Table 5.10.
140
For example, the results of the Wald tests of the full model are for the economic variables: F(6, 57.7) =
3.55, p < 0.01; for the political institutions variables: F(10, 57.7) = 0.20, p > 0.10; and for the budget
institutions: F(3, 54.9) = 0.81, p > 0.10. The results are qualitatively the same for the other models.

166
Table 5.11
Composition of Expenditures: Nested Models
Dependent Variable: Transfers and Subsidies Public Goods
Model: (1) (2) (3) (4) (5) (6)
LPOPTOT -0.55 -0.57 -0.55 -1.30 -0.81 -0.90
(.83) (.44) (.70) (1.10) (0.57) (1.03)
POP1564 -0.18 -0.12 -0.18 -0.02 0.06 0.00
(.23) (.20) (.16) (0.23) (0.22) (0.23)
POP65 0.87 0.85 0.78 0.06 -0.03 0.02
(.24)*** (.20)*** (.19)*** (0.27) (0.22) (0.25)
LGDPPC -0.24 -0.76 -0.04 -0.94 -0.45 -0.35
(.84) (.77) (.51) (0.74) (0.67) (0.74)
TRADE 0.00 0.00 0.00 0.01 0.02 0.01
(.017) (.02) (.02) (0.02) (0.02) (0.03)
RESRENTS -0.01 0.00 -0.04 0.07 -0.02 -0.01
(.10) (.09) (.08) (0.12) (0.10) (0.11)
TOTSEATS 0.00 0.00 0.00 0.00
(.01) (.01) (0.01) (0.01)
BICAMERAL -0.29 -0.43 2.18 0.52
(1.37) (1.20) (1.76) (1.72)
GOVFRAC -0.54 -2.40 1.61 -0.25
(4.34) (4.08) (5.85) (5.43)
LGSTPARTY 0.85 -3.34 -5.20 -6.86
(7.73) (5.72) (8.01) (7.18)
ENLP 0.23 0.32 -0.47 -0.46
(.55) (.54) (0.79) (0.75)
MAJPRES -0.76 0.49 -1.70 -1.45
(2.18) (2.08) (3.12) (2.89)
MAJPAR -1.22 -0.10 1.20 1.26
(1.47) (1.42) (2.23) (2.03)
PROPRES -2.19 -0.73 -1.66 -1.14
(2.10) (1.60) (2.35) (2.20)
FEDERAL 1.05 1.61 0.17 1.43
(1.85) (1.73) (2.62) (2.33)
FISCCEN 3.29 3.80 11.60 10.07
(5.31) (5.36) (8.65) (7.95)
FISCRULES_E -0.13 -0.15 -0.28 -0.18
(.20) (.19) (0.26) (0.27)
TRANSPARENCY 0.07 0.05 -0.41 -0.43
(.09) (.09) (0.15)** (0.16)**
PRORULES -0.16 -0.20 -0.04 -0.01
(.19) (.17) (0.25) (0.26)
Average RVI 0.5666 0.6417 0.5848 0.3425 0.8500 0.6876
Region Yes Yes Yes No No No
Colony Yes Yes Yes No No No
Gini Yes Yes Yes No No No
Number of 83 83 83 83 83 83
Observations
Adjusted R2 0.68 0.70 0.69 0.18 0.39 0.41
* significant at 1%; ** significant at 5%; ***significant at 1%. Standard errors in parentheses.

167
Budget institutions are better able to explain public goods spending. As shown in
column 4 of Table 5.11, the “political economy” model is unable to explain much
variance in public goods expenditures (adjusted R
2
= 0.18).
141
No variables enter the
model at conventional levels of statistical significance and, in fact, the overall model is a
poor fit: F(16, 62.4) = 1.35, p > 0.10. When the budget institutions variables are added,
however, things begin to markedly improve. In the case of the full model, adding the
budget institutions variables increases the amount of variance explained by an impressive
23% (0.41 – 0.18 = 0.23). Nevertheless, only transparency is statistically significant and
as mentioned above there is no clear explanation for this relationship. As it stands,
however, the models are not very robust. When the standard controls are added the
models become a poor fit, even though transparency remains statistically significant at
conventional levels.
142
For this reason, Table 5.11 reports the results without the regional
and colonial controls. Nonetheless, keep in mind that, once again, the average RVI for the
models is extremely high.
In sum, tests for the effects of budget institutions on expenditure composition
produced decidedly mixed results. Hypotheses 4, 8, and 13 were partially, albeit not very
strongly, confirmed. Fiscal rules were shown to have no effect on public goods spending
as expected, however, a relationship did emerge between fiscal rules and transfers and
subsidies spending. Likewise, transparency had no effect on transfers and subsidies
spending but unexpectedly affected the level of spending on public goods. With regard to
procedural rules, hypothesis 13 predicted that hierarchical procedural rules would be

141
Note that this is only a 2% improvement over the explanatory power of the economic variables and the
political institutions variables on their own, each with an adjusted R
2
of 0.16.
142
The goodness-of-fit statistics for the fully-controlled models are as follows: for the “political economy”
model: F(31, 47.6) = 1.00, p > 0.10; for the joint budget institutions and economic model: F(24, 53.4) =
1.63, p > 0.05; and for the full model F(34, 43.8) = 1.31, p > 0.10.

168
associated with lower levels of transfers and subsidies spending. The evidence supported
this expectation; however, the relationship was not very robust. Finally, hypothesis 12,
which argued that hierarchical procedural rules would be associated with greater levels of
public goods spending, was not supported by the data. However, to my knowledge, this
study is one of the first to explore the relationship between budget institutions and the
composition of spending. In light of these results, additional work is required to explain
and further test the robustness of these relationships. The following section is a first step
in this direction.

5. Robustness Tests
Here, I test the sensitivity of the findings on budget institutions using alternative
indicators and samples. The results of these robustness tests are presented in Tables 5.12a
and 5.12b. In models 1 and 2 I test the sensitivity of the effects found between
transparency, procedural rules, and smaller government, and between fiscal rules,
procedural rules, and transfers and subsidies spending. Models 3 through 14 assess the
robustness of the results to alternative democratic samples. Specifically, models 3
through 6 look at size of government, models 7 though 10 examine transfers and
subsidies, and models 11 though 14 analyze public goods spending. Here, samples are
conditioned based on the Freedom House, Polity IV, Cheibub et al. (2010), and Boix
(forthcoming), democracy indices discussed in the previous chapter. Finally, in models
15 though 17 I consider the sensitivity of the results when limited to members of the
Organization for Economic Cooperation and Development (OECD), and conversely to
non-members in models 18 to 20. As will be shown, the results are surprisingly robust.

169

Table 5.12a
Robustness Tests
Dependent
Variable:
Social Security
Revenues & Welfare Expenditures Transfers and Subsidies
Model: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
FISCRULES_E[R] -1.16 -0.38 -0.18 -0.18 -0.10 -0.42 -0.30 -0.22 -0.30 -0.28
(0.92) (0.22)* (0.35) (0.38) (0.35) (.30) (0.21) (0.21) (0.22) (0.22)
TRANSPARENCY -0.34 0.05 -0.42 -0.38 -0.31 -0.33 0.06 0.10 0.06 0.06
(0.16)** (0.11) (0.19)** (0.20)* (0.19) (0.15)** (0.09) (0.11) (0.11) (0.10)
PRORULES -1.24 -0.21 -1.18 -1.16 -1.05 -1.30 -0.25 -0.33 -0.33 -0.33
(0.29)*** (0.19) (0.3)*** (0.31)*** (0.28)*** (0.24)*** (0.17) (0.17)* (0.17)* (0.17)*
Average RVI 0.1623 0.7348 0.1970 0.2003 0.1458 0.4154 0.4875 0.5020 0.4487 0.4598
Regions Yes No Yes Yes Yes Yes No No No No
Colony Yes No Yes Yes Yes Yes No No No No
Gini No Yes No No No No Yes Yes Yes Yes
Number of 83 83 66 66 68 70 66 66 68 70
Observations
Democracy Index
GASTIL <
5
GASTIL <
5
GASTIL <
3.5
POLITY >
5 CGV = 1 BMR = 1
GASTIL <
3.5
POLITY >
5
CGV =
1
BMR =
1
Adjusted R2 0.48 0.34 0.54 0.52 0.51 0.52 0.52 0.50 0.46 0.48
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors is parentheses.

170
5.1 Alternative Indicators
Section 3 demonstrated that transparency and procedural rules were associated
with reduced levels of central government spending. When central government revenues
are substituted for expenditures as a measure of the size of government, the results from
Table 5.6 not only hold, but they are estimated more precisely and their effect is stronger.
Whereas transparency was shown to reduce expenditures by 0.32% of GDP, transparency
reduces revenues by 0.34% of GDP, all else constant. Similarly, procedural rules reduce
expenditures by 1.15% of GDP but decrease revenues by 1.24% of GDP, all else equal.
Note, however, that fiscal rules limiting revenues have no effect on revenues.
The general results are less robust when spending on social security and welfare
is substituted for spending on transfers and subsidies as a measure of the level of
redistribution (model 2). Whereas before there was a small effect between fiscal rules and
procedural rules and transfers and subsidy spending, when the indicator is changed to
social security and welfare spending only the effect of fiscal rules remains. However, the
effect on social security and welfare spending is larger (0.38 versus 0.35) but, as before,
it is imprecisely estimated. Nevertheless, this specific effect is not sensitive to the
specification, although, as mentioned above, the explanation is not immediately obvious.

5.2 Alternative Democratic Samples
Chapter 3 argued that the theory underlying the relationship between budget
institutions and fiscal policy outcomes only applied to democratic regimes. However,
Chapter 4 discussed and cited evidence of the possibility of sensitive results based on the
specific measure of democracy utilized. For this reason each of the models is re-estimated

171
using a Freedom House score less than 3.5, a Polity IV greater than 5, and scores of 1 on
both the Cheibub et al. (2010), and Boix (forthcoming) indices as sampling thresholds.
In this study, sensitivity to democratic classification appears only to be an issue for
transfers and subsidies spending. The other budget institutions hold remarkably well.
Models 3 through 6 show that the effect of transparency and procedural rules on
size of government is stable across democratic indices. The effect of procedural rules is
precisely estimated (p < 0.01) and the effect is stronger than the initial estimate of 1.15%
in three out of the four specifications. In terms of transparency, the precision of the
estimate is as good as the initial estimate (p < 0.10), or better (p < 0.05), in three of the
four models. Furthermore, the estimated effect is stronger than the initial estimate of 0.32
in the same models. Thus, it is fairly safe to state that the effects of transparency and
procedural rules are robust to the democracy index used here.
As mentioned above, the effect of budget institutions on transfers and subsidies
spending is not robust to different democratic samples. Whereas the initial estimates
suggested that fiscal rules and procedural rules diminish transfers and subsidies spending
(by 0.35% and 0.31% of GDP, respectively), in models 7 through 10 only the effect of
procedural rules remains. While the estimated effect is slightly larger, (0.33 versus 0.31),
it is imprecisely estimated (p < 0.10) and statistically significant in only three of the four
models. Adding to the confusion is the fact that fiscal rules, which were initially shown
to affect transfers and subsidies spending as well as social security and welfare spending,
is now found to have no impact on the former. Further research in required to explore the
conditions under which this relationship holds.

172

Table 5.12b
Robustness Tests (cont.)
OECD Non-OECD
Dependent
Variable:
Transfers Public Transfers Public
Public Goods Expenditures
&
Subsidies Goods Expenditures
&
Subsidies Goods
Model: (11) (12) (13) (14) (15) (16) (17) (18) (19) (20)
FISCRULES_E -0.17 -0.24 -0.10 -0.16 -0.01 0.11 -0.12 -0.46 -0.44 -0.22
(0.29) (0.30) (0.25) (0.27) (0.56) (0.27) (0.38) (0.31) (0.25)* (0.33)
TRANSPARENCY -0.49 -0.49 -0.41 -0.49 -0.37 -0.01 -0.39 -0.30 0.07 -0.48
(0.14)*** (0.15)*** (0.13)*** (0.13)*** (0.29) (0.12) (0.23) (0.18) (0.12) (0.16)***
PRORULES -0.09 -0.11 -0.08 -0.12 -1.49 -0.39 -0.03 -0.86 -0.12 -0.30
(0.26) (0.24) (0.20) (0.24) (0.36)*** (0.19)* (0.26) (0.36)** (0.26) (0.35)
Average RVI 0.8969 0.9028 0.8502 0.9357 0.2963 0.1643 0.6393 0.2953 0.7243 0.9773
Regions No No No No No No No No No No
Colony No No No No No No No No No No
Gini No No No No No Yes No No Yes No
Number of 66 66 68 70 33 33 33 50 50 50
Observations
Democracy Index
GASTIL <
3.5
POLITY >
5 CGV = 1 BMR = 1 GASTIL < 5
GASTIL <
5
GASTIL <
5 GASTIL < 5
GASTIL <
5
GASTIL <
5
Adjusted R2 0.31 0.31 0.26 0.31 0.41 0.47 0.21 0.21 0.29 0.31
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

173
Finally, columns 11 through 14 depict the results of public goods spending. As
before, greater budget transparency is related to lower public goods spending. The effects
are precisely estimated in each model (p < 0.01) and slightly larger than the initial
estimate in three of the four specifications (0.49 versus 0.47). Nevertheless, as with the
relationship between fiscal rules and transfers and subsidies spending, the association
between budget transparency and reduced expenditure on public goods is not only
unexpected, but is inexplicable, given the theory developed in Chapter 3. Based on the
evidence presented here the theory behind the relationship between budget institutions
and the composition of spending is an area of research that needs to be further developed.

5.3 Level of Development
As a final sensitivity test, I separate the sample roughly into two groups: those
countries considered “developed” based on membership in the Organization of Economic
Cooperation and Development and those considered to be “less developed” and “under
developed.” The results are depicted in models 15 though 20 in Table 5.12b. Among
OECD countries, only procedural rules appear to affect the fiscal policy outcomes
hypothesized in this study. Procedural rules reduce government size and transfers and
subsidies spending. Moreover, the effects are slightly larger than in the full sample: a
1.49% reduction in the size of government among OECD countries versus 1.29% for the
full sample; and, 0.39% lower spending on subsidies and transfers among OECD
countries compared with 0.31 in the full sample. There is no discernable affect of budget
institutions on public goods spending, as before.

174
Conversely, the effects of budget institutions appear to be more effective among
non-OECD countries. Interestingly, each budget institution appears to influence one, and
only one, fiscal policy outcome. Fiscal rules reduce the level of spending on transfers and
subsidies and transparent budget institutions reduce public goods spending. Each of these
estimates is slightly larger among non-OECD countries than for the sample as a whole.
For instance, spending on transfers and subsidies is 0.44% of GDP less among non-
OECD countries with expenditure rules compared to 0.35% of GDP among the full
sample. Similarly, public goods spending is 0.48% of GDP less among non-OECD
countries with highly transparent budget institutions compared with 0.47% of GDP
among the full sample. Overall, expenditure levels are lower among countries with more
hierarchical procedural rules, although the effect is substantially smaller among non-
OECD countries (0.86) than it is for the full sample (1.29).
Thus, with the exception of the overall level of expenditures, budget institutions
appear to be most effective among the less developed and underdeveloped countries in
the study sample. This is encouraging and suggests that international efforts such as the
IMF’s “Code of Good Practices and Fiscal Transparency” and the “International Budget
Partnership” initiative at the Center on Budget and Policy Priorities may be having
positive effects.

6. Conclusion
This chapter examined the effects of budget institutions on size of government
and budget (expenditure) composition. By and large, the results conformed to the
expectations outlined in Chapter 3; however, the strength and robustness of the results do

175
vary slightly according to the controls and samples utilized. In general, however, it is safe
to say that budget institutions, specifically procedural rules and budget transparency, do
tend to lower the size of government. This was found to hold regardless of whether
government size is measured as the level of expenditures or revenues as a percent of GDP.
Furthermore, budget institutions explained government size comparatively well when
juxtaposed against alternative economic and other political institutional explanations.
Perhaps the most surprising results were the findings that fiscal expenditure rules
lower the amount of spending on transfers and subsidies and transparent budget
institutions lower the level of expenditures on public goods. In both instances, the theory
developed in Chapter 3 predicted that there would be no discernable relationship between
these budget institutions and the composition of spending. Nevertheless, these results are
surprisingly robust to econometric specifications. This suggests that since this study is
one of the first to explore these relationships there is solid ground on which to build
further theoretical frameworks and hypothesis testing. In the concluding chapter I discuss
possible explanations for these seemingly anomalous results.
One comment is in order regarding the effect of missing data on this study. It
should be obvious by now that multiple imputation, while helpful, is not a panacea for
problems with poor data. Various models were shown to have high, and in some cases,
extremely high, average variance inflations. The significance of this effect is that
standard errors are typically higher (although to what degree cannot be known) than they
otherwise would be had the data been complete. This implies that t-scores are lower than
they should be and thus effects may be understated. This applies to the effects of budget
institutions but also the instances in which no effect was found among the economic and

176
other political institutions variables. Therefore, the results, and especially the results of
the comparison tests, should be read with caution. Until the theory and comparisons are
tested against complete data, the results reported here, while suggestive, must be
considered tentative.

177
CHAPTER 6: BUDGET DEFICITS AND GOVERNMENT DEBT
1. Introduction
The previous chapter examined the effects of budget institutions on size of
government and budget composition. In this chapter examine the relationship between
budget institutions and budget deficits and government debt. According to the theory
developed in Chapter 3, fiscal rules limiting the size of budget deficits or requiring a
balanced budget, greater transparency, and hierarchical procedural rules are expected to
result in better budget balances. Likewise, fiscal rules that impose ceilings on borrowing,
greater budget transparency, and hierarchical procedural rules are predicted to be related
to lower levels of government debt.
Section 2 of this chapter tests the effect of budget institutions on the budget
deficit. I find that fiscal rules improve the budget balance, provided that the level of
transparency is sufficiently high. This conforms to theoretical expectation. Unexpectedly,
I find that there is also a direct effect of fiscal rules on lower deficit levels, although it is
of lesser magnitude. I find no evidence that the degree of transparency or hierarchical
procedural rules is associated with lower budget deficits, contrary to expectation. Overall,
neither budget institutions nor the economic and other political institutional variables
explain the variation in budget deficits particularly well.
Section 3 evaluates the effect of budget institutions on government debt. Here, the
results are more in line with my theoretical expectations. Assuming a sufficiently high
level of budget transparency, fiscal debt rules are related to smaller government debt
levels. Similarly, hierarchical procedural rules are inversely related to government debt.
There is some evidence that budget transparency is also related to lower levels of

178
government debt, although the evidence is not a strong. Nevertheless, as with deficits, the
variables considered here do a poor job of explaining government debt.
Finally, and similar to the format of Chapter 5, I evaluate the robustness of the
positive findings to alternative specifications. This is done in Section 4. Section 4.1
examines the sensitivity of the results to the alternative democracy indices. The results
show the findings to be insensitive to the particular definition of democracy utilized.
Section 4.2 analyzes the effect on the results of dividing the sample according to OECD
membership. It turns out that the results of sections 2 and 3 are most effective among the
OECD countries. This stands in stark contrast to the results of the previous chapter where
budget institutions where demonstrated to be more effective among non-OECD countries.
The implication of these results is that budget institutions may have differential effects
depending upon a country’s level of development and the fiscal policy outcome of
interest. Section 5 concludes.

2. Budget Deficits
This section tests the theories described in the previous chapters to explain the
size of the central government deficit (if negative) or surplus (if positive) as a percentage
of GDP. As before, it begins by first estimating a baseline “economic” model with which
to compare against the budget institutions model and alternative political institutional
explanations. The economic models utilize the same set of variables as those used to
explain the size of government and the composition of the budget, since they have also
been shown to affect the size of the budget deficit and the level of debt (e.g. Von Hagen

179
and Harden 1995; Persson and Tabellini 2003b; Hallerberg et al. 2007). The results of
these estimations are depicted in Table 6.1.
Table 6.1
Budget Deficits: Economic Models
Model: (1) (2) (3) (4)
LPOPTOT -0.10 -0.17 -0.18 -0.44
(0.22) (0.22) (0.29) (0.28)
POP1564 -0.04 -0.03 -0.04 0.04
(0.08) (0.08) (0.13) (0.12)
POP65 0.08 0.02 0.06 0.05
(0.10) (0.09) (0.14) (0.13)
LGDPPC 0.52 0.71 1.04 1.04
(0.28)* (0.28)** (0.48)** (0.46)**
TRADE 0.01 0.01 0.01 0.00
(0.01) (0.01) (0.01) (0.01)
RESOURCES 2.64 3.10
(0.76)*** (1.00)***
RESRENTS 0.16 0.22
(0.04)*** (0.05)***
F-statistic 3.38 4.16 1.28 1.73
p-Value 0.01 0.00 0.22 0.05
Region No No Yes Yes
Colony No No Yes Yes
Number of 83 83 83 83
Observations
Adjusted R
2
0.15 0.19 0.07 0.16
* significant at 10%; ** significant at 5%; ***significant at 1%.
Standard errors in parentheses.

First, note that across models only per capita GDP (LGDPPC) and the pair of
measures for natural resource abundance (RESOURCES and RESRENTS) are statistically
significant at conventional levels. According to model 1, the so-called advanced and
resource abundant countries are more likely to run budget surpluses. Specifically, for a
10% increase in per capita GDP, the budget balance of a country is expected to increase
by 0.05% of GDP, all else constant.
143
Furthermore, the budget balances of hydrocarbon

143
Because GDP per capita is naturally logged, the effect on the dependent variable for a p% increase in the
independent variable is calculated using the following formula: p% increase in Y =
( ) 100 100 ln p + × β
)
(Wooldridge 2008).

180
and/or mineral-rich countries are estimated to be 2.64% larger than countries that are not
resource rich, holding all other predictors constant. When total natural resource rents are
substituted into the model (model 2), the effect of GDP per capita increases to 0.07% and
a 1% increase in resource abundance is estimated to increase the budget balance by
0.16% of GDP, ceteris paribus. In models 3 and 4, the standard regional and colonial
origin controls are added to the specifications. With the possible effects of these
characteristics and the other variables controlled for, the effect of per capita GDP once
again increases, to 0.10% (assuming a 10% increase in GDP per capita), as does the
indicator for resource abundance, to 3.10%, and the measure of total resource rents, to
0.22%.
The important thing to note here is the moderately poor performance of these
models in terms of variance explained. Without the regional and colonial controls, the
economic baseline models capture between 15% and 19% of the variance explained.
While this is comparable to the explanatory power of these models to public goods
spending, it is much less than their performance in terms of explaining size of
government (≈ 40%-45%) and the transfer and subsidy spending (≈ 65%-70%) observed
in Chapter 5. When the geographic and colonial history controls are added, the variance
explained actually decreases, explaining only between 7% and 16% of the variance. The
fact that these models have higher standard errors and lower adjusted R
2
suggests that the
geographic and colonial controls are irrelevant for estimating budget deficits
(Studenmund 2006). Further evidence of this is that the F-statistics of these models are
insignificant at conventional levels, indicating that there is a “lack of fit” and that the
models do not explain budget deficits any better than the null (without independent

181
variables) model (F(21, 59.1) = 1.28, p = 0.22 and F(21, 59.1) = 1.73, p = 0.052,
respectively). Nevertheless, as will be shown below, political institutions and budget
institutions also poorly explain budget deficits, and in most cases perform worse than the
baseline model.
Table 6.2 presents the results on the effects of political institutions on the budget
balance. Models 1 and 2 include the full sample of countries. In these specifications, only
the size of the legislature (TOTSEATS) and degree of fiscal centralization (FISCCEN)
appear to affect the budget balance; and both do negatively. For instance, the effect of
adding an additional seat in the legislature (lower house) is to reduce the budget balance
by 0.01% of GDP, all else equal. The effect of increasing the share of central government
expenditures to total government expenditures by 1% is to reduce the budget balance by
0.0564% of GDP. Put another way, moving from one extreme of a fully decentralized
fiscal system to a fully centralized one would lead to an increase in the budget deficit of
5.64% of GDP. These results are in line with the theoretical arguments and empirical
findings regarding legislature size and fiscal centralization found elsewhere. Chapter 2
presented arguments which posited that increasing the size of the legislature would
exacerbate the common-pool problem (Weingast 1979, Shepsle and Weingast 1981;
Weingast et al. 1981) and that fiscal centralization would increase the likelihood of fiscal
irresponsibility since it eliminates tax and revenue competition from lower levels of
government (Brennan and Buchanan 1977, 1978, 1980). However, the theorized effects
of legislative structure, political parties, and electoral systems, were not supported by the
results.

182

Table 6.2
Budget Deficits: Political Institutions Models
Model: (1) (2) (3) (4) (5) (6)
TOTSEATS -0.01 -0.01 -0.005 -0.005 -0.01 -0.01
(0.002)*** (0.002)*** (0.002)** (0.002)** (0.003)* (0.003)*
BICAMERAL 0.32 0.32
(0.73) (0.72)
RELSUC 1.12 1.15 5.48 6.41
(3.38) (3.26) (5.45) (4.87)
GOVFRAC -1.34 -1.38 -5.79 -5.74 -7.44 -6.42
(2.23) (2.21) (2.90)* (2.66)** (4.18) (3.10)*
LGSTPARTY -1.87 -1.95 -3.46 -3.39 -2.36 -0.65
(3.03) (2.98) (4.21) (3.94) (8.69) (7.11)
ENLP 0.18 0.18 0.39 0.39 0.70 0.66
(0.28) (0.28) (0.28) (0.27) (0.40) (0.37)*
MAJ 0.42 0.08 0.16
(0.66) (0.81) (0.97)
PRES -0.45 0.18 1.15
(0.64) (0.77) (1.33)
MAJPRES 0.01 0.24 1.36
(0.90) (1.08) (1.64)
MAJPAR 0.33 0.12 0.69
(0.88) (1.17) (1.66)
PROPRES -0.53 0.20 1.72
(0.80) (0.91) (2.00)
FEDERAL 0.49 0.48 2.04 2.03 3.15 3.07
(1.07) (1.05) (1.05)* (1.03)* (1.32)** (1.26)*
FISCCEN -5.64 -5.63 0.69 0.63 6.44 5.19
(3.03)* (3.01)* (3.51) (3.32) (5.65) (4.33)
F-statistic 1.49 1.67 1.53 1.76 1.45 1.60
p-Value 0.16 0.11 0.19 0.12 0.25 0.19
Region No No No No Yes Yes
Colony No No No No Yes Yes
Number of
Observations 83 83 39 39 39 39
Adjusted R
2
0.06 0.08 0.13 0.16 0.25 0.29
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

Models 3 through 6 separate bicameral legislatures from unicameral legislatures
to examine the effect of size disparities between chambers. As with the other measures of
fiscal policy outcomes, the effect of this variable is not statistically different from zero.
Note that among bicameral legislatures government fractionalization and federalism work

183
to decrease, and increase, the budget balance, respectively. Thus, as the probability that
two deputies picked at random from among the government parties will be from different
parties increases, so too will budget deficits. Nevertheless, given the extremely small
sample sizes of these models the estimated effects are quite large and highly likely to be
inflated. For this reason I do not attempt to interpret the significance of these effects until
such tests can be run on a larger sample and therefore the results should be read
cautiously. Observe, however, that the size of the legislature remains statistically
significant and its effect remains constant regardless of specification. In fact, as the
results of the nested comparisons will demonstrate, this is the only robust result from the
political institutions explanations for budget deficits found in this study.
Despite the handful of statistically significant results it must be kept in mind that
the set of political institutions examined here does not explain budget deficits well.
144

Observe that in every model presented the F-statistic measuring “goodness of fit” is low
and in each and every case the null hypotheses of no effects of the independent variables
cannot be rejected.
145
This explains the lack of variance explained in these models—6%
and 8%, respectively, in the first two models of primary interest. What should be taken
away from this exercise is that the size of the legislature affects the budget balance, and
this effect is quite robust.
146

The results of the effects of budget institutions on deficits are shown in Table 6.3.
These models test three hypotheses from Chapter 3. Hypothesis 1 posited that that fiscal

144
That is, among the sample and period under observation and given the data quality of this study.
145
It is for this reason that I do not present the results of the full sample political institutions models with
full controls. The F-statistic and significance tests for models 1 and 2 with full controls, respectively, are
F(25, 55) = 0.91, p = 0.5890 and F(24, 56) = 0.97, p = 0.5216.
146
It is for this reason that the bicameral models (models 5 and 6) include the set of full controls. They
demonstrate that the effect of size of the legislature on deficits is consistent across specifications.

184
rules limiting the size of the deficit or requiring a balanced budget would lead to lower
deficit levels. This hypothesis was expected to hold provided that the level of budget
transparency was sufficiently high. Surprisingly, the results indicate that fiscal rules
lower deficits regardless of the level of transparency (i.e. there is a direct effect), yet the
effect is stronger among more transparent budget institutions (i.e. there is a conditional
effect). Models 1 and 2 show that increasing the restrictions on deficit levels can improve
the budget balance. The direct effect of an increase in the fiscal rule index by one unit is
shown to improve the balance by between 0.29% and 0.34% of GDP. Among countries
with budget institutions above the 50
th
percentile in transparency, however, the effect is
greater, 0.45% of GDP, according to Model 4. Model 3 depicts the results of countries
where budget transparency is below the 50
th
percentile. Here, the effect disappears.
According to these results, there may be some truth to the fear that low transparency
encourages “creative accounting” practices and thus dilutes the effect of fiscal rules (e.g.
Von Hagen 1991; Alesina and Perotti 1996, 2008). These results support hypothesis 1.
Table 6.3
Budget Deficits: Budget Institutions Models
Model: (1) (2) (3) (4)
FISCRULES_B 0.29 0.34 0.08 0.45
(0.15)* (0.17)** (0.22) (0.22)**
TRANSPARENCY 0.05 0.03
(0.05) (0.06)
PRORULES 0.01 0.02 0.08 0.04
(0.09) (0.12) (0.12) (0.14)
Average RVI 0.1782 0.0437 N/A N/A
F-statistic 1.39 0.59 0.30 2.55
p-Value 0.25 0.89 0.74 0.09
Region No Yes No No
Colony No Yes No No
Number of 83 83 35 48
Observations
Adjusted R
2
0.03 N/A N/A 0.06
* significant at 10%; ** significant at 5%; ** significant at 1%.
Standard errors in parentheses.

185
Conversely, the tests do not find effects on the budget balance for transparency
(direct), or procedural rules. Hypothesis 5 suggested that more transparent budget
institutions would be associated with lower deficits. While the sign of the transparency
index coefficient is in the hypothesized direction, its effect is not statistically different
from zero. The same is true for procedural rules. Hypothesis 9 argued that more
hierarchical budget institutions would be related to budget surpluses. Here, too, the
direction of the effect is as expected but it is not statistically or economically impactful.
Therefore, the findings do not support hypotheses 5 and 9.
These last results possibly are to be expected given the low proportion of variance
explained by any of the models and their overall lack of fit. In fact, the lack of fit is so
poor in the cases of models 2 and 3 that no Adjusted-R
2
information is output from the
statistical software. It should come as no surprise that three variables alone explain very
little of something as important as the budget balance. In fact, as will be shown below,
even the full model is capable of explaining only one fifth of the variance in deficits
across counties. However, the average rate of variance of inflation is high, so once again
these results are tentative. It is encouraging, nonetheless, that the effect of fiscal rules
appears to be quite robust to specification.
Given that the budget institutions and alternative political institutions models do
not explain deficits particularly well it does not make much sense to evaluate competing
explanations by comparing the amount of variance in deficit levels each set of variables
are capable of explaining on their own. The more meaningful comparisons are between
nested models. Table 6.4 depicts these results.

186
Table 6.4
Budget Deficits: Nested Models
Model: (1) (2) (3)
LPOPTOT 0.40 -0.19 0.30
(0.41) (0.22) (0.42)
POP1564 0.00 -0.03 0.02
(0.09) (0.09) (0.10)
POP65 0.05 0.02 0.04
(0.10) (0.09) (0.11)
LGDPPC 0.73 0.66 0.63
(0.30)** (0.29)** (0.32)*
TRADE 0.01 0.01 0.01
(0.01) (0.01) (0.01)
RESRENTS 0.14 0.15 0.14
(0.05)*** (0.04)*** (0.05)***
TOTSEATS -0.01 -0.01
(0.003)** (0.003)**
BICAMERAL 0.12 0.27
(0.70) (0.74)
GOVFRAC 0.90 1.24
(2.19) (2.21)
LGSTPARTY 2.76 1.71
(3.01) (3.11)
ENLP 0.08 0.00
(0.27) (0.28)
MAJPRES 1.02 1.30
(1.19) (1.25)
MAJPAR 0.32 0.62
(0.89) (0.91)
PROPRES 0.12 0.19
(0.97) (0.98)
FEDERAL -0.39 -0.44
(1.06) (1.08)
FISCCEN -3.01 -2.75
(3.29) (3.29)
FISCRULES_B 0.25 0.25
(0.14)* (0.15)
TRANSPARENCY 0.04 0.05
(0.05) (0.06)
PRORULES 0.04 -0.02
(0.10) (0.10)
Average RVI 0.0146 0.0893 0.0599
F-Statistic 2.15 3.00 1.96
p-Value 0.02 0.00 0.03
Region No No No
Colony No No No
Number of 83 83 83
Observations
Adjusted R
2
0.19 0.21 0.21
* significant at 10%; ** significant at 5%; * significant at 1%.
Standard errors in parentheses.

187
Recall that the set of variables that comprise the baseline economic models are
capable of explaining 19% of the variance in budget deficits.
147
Column 1 of Table 6.4
shows the result when the set of political institutions variables is added to the baseline
model. First, observe that per capita GDP, resource rents, and the size of the legislature
are statistically significant and of the same magnitude as in previous models. Note,
however, that the adjusted-R
2
of the “political economic” model is no different than the
baseline model (adjusted-R
2
= 0.19). This is despite the fact that an additional ten
variables have been added which themselves explained 6% of the variance. The adjusted-
R
2
did not go down, indicating that the additional variables are not wholly irrelevant, but
neither did they contribute to the proportion of the explained variance. Clearly, the effect
of TOTSEATS is preventing the adjusted-R
2
from being reduced by irrelevant variables
but it is conversely being prevented from adding to the explained variance by the rest of
the political institutions variables.
148

By comparison, adding the budget institutions variables to the baseline does
improve the model’s explanatory power, albeit weakly. Accounting for budget
institutions improves the proportion of explained variance by 2% (0.21 - 0.19 = 0.02).
The variables for GDP per capita, resource rents, and fiscal rules are statistically
significant and relatively of the same magnitude as in previous models. Furthermore, the
model is a decent fit (F(9, 70.6) = 3.00, p < 0.01), unlike the budget institutions models
in Table 6.3.

147
This figure refers to model 2 of Table 6.1.
148
When all of the political institutions variables, with the exception of TOTSEATS, are dropped from the
model, the adjusted-R
2
improves to 0.25. Nevertheless, for the moment the purpose is to compare
explanations rather than find the model that “best” explains deficits. Furthermore, there is no theoretical
reason to exclude these variables.

188
Column 3 presents the results of the full economic-institutional model. Once more,
adding the political institutions variables—this time to the economic and budget
institutions variables—neither improves, nor diminishes, the proportion of the variance
explained. Therefore, the model does not represent an improvement over the joint
economic and budget institutions model. In fact, the argument can be made that of the
three nested models the full model is statistically the “worst.” This is because not only
does the full model increase the adjusted-R
2
over model 2, but it also has the lowest of the
three F-statistics and thus is the poorest fit. Note however, that the fiscal rules measure is
no longer significant. This can be explained by the fact that one of the effects of adding
non-influential variables is to increase the variance and lower the t-scores of the other
variables (Studenmund 2006). As can be seen, the standard errors are indeed larger in
model 3 than they are in either of the other two models.
In summary, only hypothesis 1 found support among the statistical tests. Provided
that the level of budget transparency is sufficiently high, fiscal rules restricting deficit
levels are associated with greater fiscal discipline. Interestingly, the results demonstrated
that there is also a direct effect of fiscal rules, regardless of the level of budget
transparency. However, the effect is not as strong. Conversely, the tests did conform to
expectations regarding budget transparency and procedural rules. While their estimated
effects were in the correct directions according to hypotheses 5 and 9, the effects were
not statistically distinguishable from zero. Nevertheless, budget institutions performed
relatively well when compared to the set of alternative political institutions.
Unfortunately, none of the models explored here were capable of explaining much
of the cross-country variation in the budget balance. At most the best performing models

189
are able to account for a mere fifth of said variation. Therefore, other important
determinants of budget deficits are unaccounted for by the variables and set of controls
considered in this study. All the same, there remain strong theoretical reasons for
expecting the political and budget institutions variables to affect fiscal policy outcomes
such as deficits.
149

3. Government Debt
This section explores the determinants of central government debt as a proportion
of GDP. Compared to the previously studied fiscal policy outcomes, debt possesses a
number of outliers which could distort the estimates in subsequent regression analyses.
Diagnostic tests identify Japan, Luxembourg, and Malawi as potentially influential
cases.
150
Their debt to GDP ratios of 128%, 2.45%, and 187.27% , respectively, deviate
substantially from the average sample debt level of 56.72% of GDP. To account for the
possibility that these cases may distort the findings, each of the models is run using
“robust regression,” in addition to ordinary least squares. Robust regression is an iterative
weighting procedure where each observation is assigned a weight according to its level of
influence. The more influential an observation (the more extreme of an outlier) the less
heavily it is weighted in the regression calculations. This way the models are able to

149
For the size of the legislature see Weingast et al. (1981) and Gilligan and Matsusaka (2001). For
Bicameral legislatures see Heller (1997). For government fractionalization see Perotti and Kontopoulos
(2002) and Ricciuti (2004). For political parties see Inman and Fitts (1990), Mukherjee (2003), and Bawn
and Rosenbluth (2006). For electoral systems see Persson and Tabellini (2003b). For fiscal centralization
see Rodden (2003, 2004). For budget institutions see Poterba (1994), Alesina et al. (1996), Alesina and
Perotti (1996, 1999), and Hameed (2005), among others.
150
The diagnostic test used to identify influential cases was “Cook’s Distance.” This statistic measures the
overall change in the estimated coefficients when each observation is excluded from the model. If an
observation has a Cook’s distance larger than 4/N (in this instance 4/83 = 0.05), then estimates differ
significantly when the observation is included/excluded and thus it is considered influential (Hamilton
2008). The Cook’s distances for the three influential cases in this section are 0.09 for Japan, and 0.16 for
both Luxembourg and Malawi.

190
“resist the pull” of the outliers (Hamilton 2008: 239). Unfortunately, the robust
regression procedure does not calculate statistics of interest such as adjusted-R
2
.
Therefore, in order to continue with the comparisons of interest, OLS methods are also
utilized. Running both procedures will also act as a test of the robustness of the results to
the inclusion/exclusion of the influential cases.
The effects of the economic variables on government debt are reported in Table
6.5. Generally, the demographic measures exhibit opposite, yet statistically and
economically significant, effects on debt levels. Increasing the proportion of the working
age population by 1% reduces the level of debt between 3% and 3.5% of GDP.
Conversely, an equal change in the population of retirement age increases the level of
government debt between 2.75% and 4.25% of GDP. Furthermore, the economic
variables are able to explain 19% of the variance.
The upper bounds of these estimates, depicted in models 3 and 4, however, should
be read cautiously. Note that the inclusion of the geographic and colonial controls leads
to higher standard errors of the economic variables and a miniscule increase in variance
explained. In addition, the F-statistics fall substantially and their level of significance is
only high enough to reject the null hypotheses of no effects of the independent variables.
Overall, this suggests that the controls are largely irrelevant to determining debt and their
inclusion is deleterious to the estimates of the economic variables.
Models 5 and 6 depict the results of the analyses using the aforementioned robust
regression procedure. Observe that accounting for the potential effects of the three
outliers has reduced both the standard errors and the coefficient estimates of the variables.
This suggests that the three countries in fact had influenced the previous results.

191
Nevertheless, the proportion of the population between the ages of 15 and 64 remains
statistically significant and precisely estimated (p < 0.05) in both models. However, this
variable’s estimated effect is now to reduce debt by approximately 2.30% of GDP, an
effect about 1% less than originally estimated. The proportion of the population above the
age of 65 is only significant to the inclusion of resource rents, although it is not estimated
very well (p < 0.10). Moreover, its effect has also fallen to about 1.65% of GDP, also
about 1% than the initial estimate. Nevertheless, the analyses provide further
confirmation of the importance of a given country’s demographic composition in
determining fiscal outcomes (Wilensky 1976).
Table 6.5
Government Debt: Economic Models
Model: (1) (2) (3) (4) (5) (6)
LPOPTOT 0.60 0.84 -1.00 -0.58 0.08 0.10
(2.49) (2.51) (3.11) (3.21) (2.24) (2.26)
POP1564 -3.20 -3.21 -3.33 -3.45 -2.27 -2.30
(0.95)*** (0.95)*** (1.35)** (1.38)** (0.85)** (0.86)***
POP65 2.75 2.72 4.18 4.24 1.64 1.67
(1.09)** (1.03)** (1.49)*** (1.47)*** (1.00) (0.94)*
LGDPPC -3.44 -3.69 -11.18 -11.22 -2.84 -2.93
(3.31) (3.30) (5.18)** (5.17)** (2.99) (2.99)
TRADE -0.05 -0.04 -0.11 -0.10 0.00 0.00
(0.10) (0.10) (0.12) (0.12) (0.09) (0.09)
RESOURCES -1.30 -6.27 -1.06
(8.81) (11.02) (7.96)
RESRENTS -0.28 -0.35 -0.07
(0.46) (0.60) (0.41)
Average RVI 0.0608 0.0549 0.0584 0.0566 0.0877 0.0717
F-Statistic 3.94 4.04 1.79 1.79 2.67 2.76
p-Value 0.0018 0.0015 0.0415 0.0410 0.0211 0.0179
Region No No Yes Yes No No
Colony No No Yes Yes No No
Number 83 83 83 83 83 83
of Observations
Adjusted R
2
0.19 0.19 0.20 0.20 N/A N/A
* significant at 10%; significant at 5%; significant at 1%. Standard errors in parentheses.

192
The fact that the demographic variables affect debt levels may be somewhat
surprising, given that the previous section was unable to detect an impact of these
variables on the deficit and government debt is essentially accumulated deficits. One
possible explanation for this seemingly anomalous result is to recognize that the deficit is
a flow variable (i.e. measured over time), whereas debt is a stock variable (i.e. measured
at a given point in time). Thus, it may be that the influence of the demographic variables
on the deficit is manifested over time, which explains why it is captured in the debt
analyses. Unfortunately, without longitudinal data this hypothesis cannot be tested. I
therefore leave it as a conjecture for future exploration.
Table 6.6 presents the estimates of the political institutions variables. Similarly to
deficit levels, these variables are poor predictors of government debt. First, note that in
every specification the null hypothesis of zero effect of the independent variables cannot
be rejected.
151
This is related to the fact that no political institutional variable is
consistently significant across models. Second, observe that the proportion of variance
explained by these variables is extremely low; only between 3% and 4% of the total
variance. Next, notice that the standard errors are very large, as are some of the
coefficient estimates. There are likely causes for the biased coefficient estimates and
inflated standard errors With regard to the coefficient estimates, these are larger than
reasonably expected and in some cases (e.g. GOVFRAC) of the opposite sign.
152
This
suggests that omitted variable bias may be present. Conversely, the large standard errors
may be a consequence of the inclusion of irrelevant variables (Studenmund 2006).

151
These conditions only worsen when the geographic and colonial controls are included. For this reason I
do not report these results.
152
Multicollinearity is not a problem. The variance inflation factor (VIF) for each of the coefficients is well
below standard thresholds of 10 and 5.

193

Table 6.6
Government Debt: Political Institutions Models
Model: (1) (2) (3) (4) (5) (6)
TOTSEATS 0.02 0.02 0.01 0.01 0.01 0.01
(0.03) (0.03) (0.03) (0.03) (0.02) (0.02)
BICAMERAL -7.51 -7.25 -2.28 -1.90
(8.53) (8.55) (7.74) (7.78)
RELSUC -8.12 -7.61
(47.26) (45.58)
GOVFRAC -37.99 -41.32 -15.14 -14.42 -31.27 -29.03
(27.5) (27.41) (42.41) (39.66) (25.50) (25.75)
LGSTPARTY -2.76 -9.04 -9.99 -9.17 5.25 -0.92
(36.22) (35.55) (59.87) (55.97) (33.07) (32.78)
ENLP 5.06 5.07 5.31 5.27 6.21 5.10
(3.83) (3.83) (4.69) (4.61) (3.78) (3.83)
MAJ 9.43 14.33 5.53
(7.94) (11.36) (7.27)
PRES 8.17 -1.33 4.51
(7.57) (10.80) (6.91)
MAJPRES 20.70 12.74 14.53
(10.98)* (15.23) (10.38)
MAJPAR 1.83 14.95 -0.24
(10.37) (16.27) (9.44)
PROPRES 1.98 -0.98 -0.43
(9.31) (12.74) (8.59)
FEDERAL -5.92 -7.09 -11.78 -11.82 -0.04 -1.76
(12.39) (12.39) (14.75) (14.44) (12.41) (12.36)
FISCCEN -7.16 -7.64 -35.90 -36.41 17.82 11.43
(35.70) (35.62) (49.16) (46.59) (38.61) (37.24)
Average RVI 0.0831 0.0838 0.0798 0.0899 0.1945 0.1795
F-Statistic 1.00 0.98 0.56 0.64 0.73 0.49
p-Value 0.4487 0.4617 0.8282 0.7525 0.6936 0.8781
Region No No No No No No
Colony No No No No No No
Number 83 83 39 39 83 83
of Observations
Adjusted R
2
0.04 0.03 0.06 0.06 N/A N/A
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

194
This conclusion is further supported by the low absolute magnitude of the t-scores (near
zero in most cases) and the low adjusted-R
2
of the models.
153
Finally, these problems
persist even after using the robust regression procedure, suggesting that the influential
cases are not responsible. As will be shown below, examining the nested models further
supports these conclusions. All in all, one may reasonably infer that the political
institutions examined among the sample and period covered here simply cannot explain
government debt levels.
What effects do budget institutions have on government debt levels? Recall from
Chapter 3 that, according to hypothesis 2, fiscal rules that impose borrowing constraints
should be associated with lower levels of debt. Similarly, according to hypothesis 6 and
hypothesis 10, higher levels of budget transparency and more hierarchical budget
procedures, respectively, are expected to be related to smaller debt. Table 6.7 depicts the
results of these tests. The first column presents the results of the most parsimonious
model, which includes only the budget institutions variables. All of the variables are in
the expected direction, indicating that higher levels of each budget institution index are
associated with lower debt levels. Overall, budget institutions are capable of accounting
for 9% of the variance. However, only the procedural rules index is statistically
significant, albeit not precisely estimated (p < 0.10). The estimated effect of increasing
the degree of hierarchy of procedural rules in the budget process is to lower debt by
1.79% of GDP. This supports hypothesis 10.
Nonetheless, the budget institutions models are not particularly well-fit, as
indicated by the low F-statistics. This problem is exacerbated in model 2, which includes
the standard set of controls. Once more, the higher standard errors between models 1 and

153
Only the latter are reported in Table 6.6.

195
2 and the reduction in adjusted-R
2
from 0.09 to 0.03 imply that the controls are irrelevant
for determining debt. One encouraging result is that the estimated effect of procedural
rules remains statistically significant from zero even in this poorly fit model.
In order to test hypothesis 2, regressions were run holding transparency levels at
the 25
th
, 50
th
, and 75
th
percentiles. Positive results became manifest only once
transparency was held at the 75
th
percentile. These results are presented in column 3. The
results suggest that once budget transparency is above the 75
th
percentile, fiscal rules
imposing borrowing constraints result in a statistically significant reduction in the level of
debt. In other words, among countries with sufficiently transparent budget institutions
(above the 75
th
percentile), increasing the strength of fiscal borrowing rules results in a
decrease of debt by 7.90% of GDP. While this finding supports hypothesis 2, it cannot be
ruled out that the estimate is an artifact of the small sample size (N = 22). Therefore, this
result is tentative. Model 4 presents the results holding transparency below the 75
th

percentile. The effect of fiscal borrowing rules disappears, yet procedural rules remain
significant and the precision of the estimate is improved (p < 0.05).
Table 6.7
Government Debt: Budget Institutions Models
Model: (1) (2) (3) (4) (5) (6) (7)
FISCRULES_D -2.61 -2.62 -7.90 -1.08 -2.00 -7.12 -0.84
(1.70) (2.02) (3.53)** (2.00) (1.35) (4.00)* (1.53)
TRANSPARENCY -0.92 -0.64 -1.06
(0.69) (0.79) (0.53)*
PRORULES -1.79 -2.14 1.18 -2.82 -1.95 0.56 -2.80
(1.01)* (1.20)* (1.85) (1.11)** (0.84)** (2.10) (0.85)***
Average RVI 0.3729 0.1183 N/A N/A 0.4800 N/A N/A
F-Statistic 2.20 0.91 2.52 3.33 3.68 1.59 5.53
p-Value 0.0951 0.5675 0.1074 0.0427 0.0159 0.229 0.0063
Region No Yes No No No No No
Colony No Yes No No No No No
Number 83 83 22 61 83 22 61
of Observations
Adjusted R
2
0.09 0.03 0.13 0.07 N/A N/A N/A
* significant at 1%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

196
Columns 5 through 7 re-estimate the models in columns 1, 3, and 4 using the
robust regression procedure. First, note that the models are now better fit in two of the
three cases. Second, once the influence of the outliers is accounted for, the effect of
transparency becomes statistically significant. Therefore, increasing transparency is
associated with reduced debt levels; according to the estimates, an increase of one unit in
the transparency index is associated with 1.06% reduction in government debt. Thus,
there is mixed support for hypothesis 6.
Models 6 and 7 once more hold transparency above and below the 75
th
percentile,
respectively. The results show that fiscal rules and procedural rules are insensitive to the
robust regression procedure, although in the former instance the precision of the estimate
and the magnitude of the effect have decreased. Overall, the results of the robust
regressions provide support for hypotheses 2, 6, and 10, and demonstrate the robustness
of hypothesis 2 and hypothesis 10 to various specifications. Based on these results,
budget institutions appear to have a statistically and economically meaningful impact of
government debt levels.
How well do budget institutions compare with the alternative explanations for
government debt? Comparing “goodness of fit,” budget institutions explain a greater
proportion of the variance in government debt than does the set of other political
institutions (adjusted-R
2
= 0.09 vs. adjusted-R
2
= 0.04). However, budget institutions are
unable to explain as much variance in government debt as the set of economic variables
(adjusted-R
2
= 0.19).
154
Nevertheless, none of the sets of variables on their own explains
debt levels particularly well and other important determinants of government debt are
unaccounted for in this study.

154
These figures refer to model 1 of Table 6.5, model 1 of Table 6.6, and model 1 of Table 6.7.

197
Table 6.8
Government Debt: Nested Models
Model: (1) (2) (3) (4) (5) (6)
LPOPTOT -1.20 1.19 -1.20 -4.61 -0.33 -3.67
(4.93) (2.49) (4.93) (4.48) (2.08) (4.67)
POP1564 -2.82 -2.53 -1.92 -2.34 -1.19 -1.18
(1.04)*** (0.98)** (1.11)* (1.01)** (0.82) (1.00)
POP65 2.18 2.09 1.69 1.73 0.40 0.96
(1.24)* (1.04)* (1.22) (1.13) (0.92) (1.08)
LGDPPC -4.30 -3.64 -4.49 -2.85 -2.43 -2.63
(3.66) (3.28) (3.64) (3.43) (2.75) (3.26)
TRADE -0.05 -0.10 -0.11 0.00 -0.09 -0.11
(0.12) (0.11) (0.12) (0.11) (0.09) (0.12)
RESRENTS -0.61 -0.31 -0.69 -0.27 -0.04 -0.28
(0.54) (0.45) (0.54) (0.51) (0.37) (0.49)
TOTSEATS 0.02 0.02 0.04 0.03
(0.04) (0.04) (0.04) (0.04)
BICAMERAL -3.57 -3.36 -0.79 -1.79
(8.32) (8.66) (7.71) (7.91)
GOVFRAC -48.04 -52.45 -46.12 -45.99
(27.27)* (27.02)* (26.02)* (25.51)*
LGSTPARTY -34.13 -38.72 -33.19 -25.68
(35.96) (34.81) (41.19) (35.69)
ENLP 4.49 4.38 6.41 5.92
(3.77) (3.63) (4.52) (4.31)
MAJPRES 2.29 7.66 8.06 10.30
(14.29) (14.75) (13.49) (13.25)
MAJPAR -0.65 -0.09 -4.37 -2.06
(10.51) (10.43) (9.46) (9.39)
PROPRES -3.80 2.07 -2.48 2.14
(11.48) (11.40) (10.49) (10.15)
FEDERAL 1.02 1.11 12.58 13.44
(12.57) (12.51) (12.30) (12.03)
FISCCEN -7.65 -11.94 37.53 38.96
(39.45) (38.72) (39.80) (40.17)
FISCRULES_D -2.20 -2.23 -1.63 -1.22
(1.57) (1.72) (1.34) (1.59)
TRANSPARENCY -0.61 -0.78 -0.73 -0.87
(0.70) (0.77) (0.63) (0.73)
PRORULES -1.85 -2.11 -2.39 -2.66
(1.06)* (1.18)* (0.93)** (1.18)**
Average RVI 0.0729 0.1859 0.1524 0.2856 0.3375 0.4411
F-Statistic 1.70 3.20 1.82 1.40 2.83 1.52
p-Value 0.0688 0.0027 0.0413 0.1730 0.0067 0.113
Region No No No No No No
Colony No No No No No No
Number 83 83 83 83 83 83
of Observations
Adjusted R
2
0.15 0.25 0.23 N/A N/A N/A
* significant at 1%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

198
Table 6.8 depicts the nested model comparisons. The first column presents the
results of the “political economy” model. The demographic variables are robust to the
inclusion of the political institutions variables. However, as discussed above, the effects
of the latter variables are largely immaterial for determining government debt. Compared
to the baseline economic model (model 1 Table 6.5), the political economy model
explains 4% less of the variance in government debt (0.19 – 0.15 = 0.04). Furthermore,
the F-statistic is low (F(16, 63.8) = 1.70) and by most standards does not represent an
improvement over the null model (p = 0.0688).
The second column of Table 6.8 shows the effect on government debt of adding
the budget institutions variables to the baseline economic model. POP1564, POP65, and
PRORULES remain statistically significant in this specification. Importantly, the
inclusion of the budget institutions variables raises the proportion of variance in
government debt explained by 6% (0.25-0.19 = 0.06). Nevertheless, Wald tests reveal
that, collectively, the set of economic variables is significant (F(6, 70.7) = 3.50, p <
0.01), whereas the set of budget institutions variables is not (F(3, 66.8) = 1.70, p > 0.10).
The latter result should not be surprising given that, of the budget institutions, only the
effect of procedural rules is statistically different from zero.
Column 3 presents the results of the full model. The results largely conform to the
pattern of findings thus far, however, POP65 is no long statistically significant and
POP1564 is less precisely estimated (p < 0.10). This is likely due to the inclusion of
irrelevant political institutions variables, which causes an increase in the variance of the
estimates (as evidenced by the larger standard errors). As before, the inclusion of these
variables reduces the proportion of variance explained between the joint budget

199
institution and economic model and the full model by 2% (0.25 – 0.23 = 0.02).
Importantly, the effect of procedural rules remains robust to the inclusion of the other
political institutions variables. This result adds further confirmation in favor of
hypothesis 10.
Columns 4 though 6 depict the results of the robust regressions. Since this
procedure does not report adjusted-R
2
when pooling over the imputed datasets, “variance
explained” comparisons cannot be made. Instead, the comparison is whether the
significant variables from the first three models remain so after correcting for the outliers,
and whether the coefficients are negatively affected by this adjustment. By these criteria,
budget institutions perform fairly well. Across models, POP1564 is only significant in
the robust political economy model (model 4) and POP65 is no longer significant in any
of the models. Furthermore, the magnitude of the effects for POP1564 (when it is
significant) and GOVFRAC are lower after adjusting for outliers. It is only for procedural
rules that the magnitude of the effect actually improves after controlling for the outliers,
as does the precision of the estimate (p < 0.05). In the end, there is robust support for
hypothesis 10 and one may confidently state that hierarchical procedural rules reduce
government debt levels.
In sum, the findings on the effects of budget institutions on government debt
levels vary according to model specification. Hypothesis 10, which posited that more
hierarchical procedural rules would be associated with lower levels of debt, found the
strongest and most robust support; statistically and economically significant in nine out of
eleven estimations. Hypothesis 2 argued that fiscal constraints on government borrowing
would result in lower levels of government debt, provided that budget transparency is

200
sufficiently high. The tests conducted here discovered that this hypothesis is only
supported at levels of budgetary transparency above the 75
th
percentile. At this level of
transparency fiscal debt rules are associated with nearly an 8% reduction in the ratio of
debt to GDP. Although, this effect may be driven by the small sample size (N = 22),
since a small sample size increases the likelihood of Type-I error (false positive) (Agresti
and Finlay 1997). Therefore, this finding should be considered tentative until it can be
tested against a larger sample of countries. Hypothesis 6 relating greater budget
transparency to lower debt levels received the least amount of support. While
transparency was found to be statistically significant in the robust estimation of the effect
on debt of budget institutions (model 5 Table 6.7), it was not precisely estimated (p <
0.10) and the effect was not robust to specification since its effect never became
statistically distinguishable from zero once the economic controls were accounted for.
Overall, the results of this section are encouraging. Each of the hypotheses tested
here received at least some empirical support. Furthermore, budget institutions fared
pretty well in comparison to the set of other political institutions variables. The next
section examines in more depth the robustness of these findings and those of the previous
section.

4. Robustness Tests
4.1 Alternative Democratic Samples
The results in Section 2 supported the expectation that, provided a sufficiently
high level of budget transparency, fiscal rules limiting budget deficits are associated with
greater budget discipline. Interestingly, these results also demonstrated that there was a

201
direct, albeit weaker, effect of such fiscal rules irrespective of the level of transparency.
In a similar fashion, the results in Section 3 more or less conformed to expectations.
Hierarchical procedural rules were found to be related to lower levels of government debt
and fiscal rules placing limits on government borrowing were found to be associated with
smaller debt. However, the latter effect was only present among a small sample of
countries with budget transparency levels above the 75
th
percentile. With regard to
transparency, the budget index was related to lower debt levels in only one specification;
however, the effect disappeared once additional covariates were taken into account.
Each of the models from Table 6.3 and Table 6.7 were re-estimated in order to
examine whether the results are somehow driven by the very loose definition of
democracy used to estimate the models.
155
As before, the sensitivity of the results to
alternative democratic samplings are tested using a Freedom House score less than 3.5, a
Polity IV score greater than 5, and scores of 1 on both Cheibub et al. (2010) and Boix
(forthcoming) indices, as conditions for entering into the estimation sample. The results
of these robustness tests are reported in Tables 6.9a and 6.9b.

155
Recall that the standard criterion used to sample from the set of countries is a Freedom House
democracy score below 5.

202

Table 6.9a
Robustness Tests
Dependent Variable: Budget Deficits
Model: (1) (2) (3) (4) (5) (6) (7) (8)
FISCRULES_B 0.34 0.39 0.36 0.37 0.43 0.51 0.40 0.45
(0.16)** (0.17)** (0.17)** (0.16)** (0.21)* (0.23)** (0.23)* (0.22)**
TRANSPARENCY 0.06 0.05 0.05 0.06
(0.05) (0.06) (0.06) (0.06)
PRORULES 0.01 0.00 0.00 -0.02 0.00 -0.01 -0.03 -0.05
(0.09) (0.10) (0.10) (0.09) (0.14) (0.15) (0.14) (0.14)
Average RVI 0.1596 0.1684 0.1726 0.1696 N/A N/A N/A N/A
F-Statistic 1.79 1.99 1.61 2.10 2.23 2.76 1.70 2.59
p-Value 0.1583 0.1251 0.1957 0.1087 0.1206 0.0754 0.1944 0.0862
Region No No No No No No No No
Colony No No No No No No No No
Number of 66 66 68 70 43 43 45 47
Observations
Democracy Index
GASTIL
<3.5
POLITY >
5 CGV = 1 BMR = 1
GASTIL
<3.5
POLITY >
5
CGV =
1 BMR = 1
Adjusted R2 0.06 0.07 0.05 0.07 0.06 0.08 0.03 0.06
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.

203

Table 6.9b
Robustness Tests
Dependent Variable: Government Debt
Model: (9) (10) (11) (12) (13) (14) (15) (16)
FISCRULES_D -2.19 -1.50 -1.86 -1.74 -7.79 -7.79 -6.14 -7.12
(1.57) (1.51) (1.55) (1.57) (4.30)* (4.30)* (3.54) (4.00)*
TRANSPARENCY -0.92 -0.95 -0.99 -1.06
(0.57) (0.58) (0.66) (0.60)*
PRORULES -2.45 -2.59 -2.30 -2.01 0.75 0.75 -1.62 0.56
(0.97)** (0.88)*** (0.89)** (0.91)** (2.33) (2.33) (1.83) (2.10)
Average RVI 0.3823 0.3424 0.5553 0.439 N/A N/A N/A N/A
F-Statistic 3.72 4.19 3.24 3.27 1.65 1.65 2.33 1.59
p-Value 0.0165 0.0095 0.0287 0.0273 0.2221 0.2221 0.1273 0.2290
Region No No No No No No No No
Colony No No No No No No No No
Number of 66 66 68 70 20 20 20 22
Observations
Democracy Index
GASTIL
<3.5
POLITY >
5 CGV = 1 BMR = 1
GASTIL
<3.5
POLITY >
5
CGV =
1
BMR =
1
Adjusted R2 0.07± 0.08± 0.09± 0.07± 0.12± 0.12± 0.22± 0.13±
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.
± Non-robust adjusted R2.

204

Table 6.9c
Robustness Tests
OECD Non-OECD

Budget Deficits Government Debt Budget Deficits Government Debt Dependent Variable:
Model: (17) (18) (19) (20) (21) (22) (23) (24)
FISCRULES_B[D] 0.61 0.59 -2.54 -8.80 0.15 0.33 -2.80 -8.63
(0.30)* (0.36) (2.93) (6.98) (0.17) (0.26) (1.60)* (5.17)
TRANSPARENCY 0.04 -0.75 0.02 -0.80
(0.12) (1.44) (0.06) (0.60)
PRORULES 0.09 -0.01 -3.45 -2.35 0.02 0.04 -0.77 3.92
(0.15) (0.20) (1.44)** (3.45) (0.13) (0.19) (1.17) (2.89)
Average RVI 0.1894 N/A 0.4367 N/A 0.2051 N/A 0.3894 N/A
F-Statistic 1.25 1.69 1.57 0.97 0.28 0.86 1.88 1.62
p-Value 0.3131 0.2052 0.2218 0.4311 0.8369 0.4396 0.1478 0.2459
Region No No No No No No No No
Colony No No No No No No No No
Number of 33 28 33 9 50 20 50 13
Observations
Democracy Index
GASTIL
< 5
GASTIL <
5
GASTIL <
5
GASTIL <
5
GASTIL <
5
GASTIL <
5
GASTIL <
5
GASTIL <
5
Adjusted R2 0.07 0.05 0.16± 0.10± 0.01 0.00 0.05± 0.16±
* significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors in parentheses.
± Non-robust adjusted R2.

205
Table 6.9a tests the robustness of the effects of budget institutions on budget
deficits. Recall that the estimated direct effect of fiscal rules was to increase the budget
balance by 0.29% of GDP. As columns 1 through 4 show, this effect is indeed robust to
the different democratic samples. Note, however, that the magnitude of the estimated
effect is larger in each of the samples than before (between 0.34% and 0.39% of GDP)
and more precisely estimated (p < 0.05 in each case as opposed to p < 0.10 before).
Thus, when more restrictive definitions of democracy are utilized, the effect of fiscal
rules in lowering budget deficits becomes even more pronounced.
The second half of Table 6.9a presents the four models which test the robustness
of fiscal rules according to alternative democracy indices conditioned on the level of
budget transparency. As the models demonstrate, the effect is quite robust. All of the
estimates possess magnitudes near or above the original finding of 0.45% lower deficits.
Furthermore, all but the estimate utilizing the narrow Freedom House threshold is as
precisely estimated as before (p < 0.05). Overall, then, the findings of this chapter
provide strong support in favor of hypothesis 1. Fiscal rules limiting deficit levels do in
fact positively impact the budget balance and this effect is even stronger among countries
with transparent budgets above the 50
th
percentile.
Table 6.9b depicts the results of the robustness tests with regard to government
debt.
156
In Section 3 it was observed that hierarchical procedural rules, fiscal debt rules,
and to a much lesser extent, greater budget transparency, were related to smaller

156
All models in Table 6.9b were estimated using the robust regression procedure previously described,
given that the results were shown to be sensitive to the inclusion of outliers. The adjusted-R
2
reported in
Table 6.9b refer to the non-robust regression models (not reported) since the robust regression procedure
(when used with multiple imputation) does not report this statistic. The regressions themselves, however,
are robust regression results. Thus, the adjusted-R
2
do not directly apply to each model and are only meant
for illustrative purposes to show what proportion of variance “might” be explained by these models.

206
government debt levels. With the exception of budget transparency, these effects are
insensitive to democratic specification. Hierarchical procedural rules continue to reduce
government debt. Although, among the more restrictive samples the effect has improved,
from lowering debt by 1.95% of GDP to lowering debt by between 2.01% and 2.59% of
GDP. Overall, these effects are estimated more precisely than before (p < 0.05 and p <
0.01). Based on the findings of this chapter, I can confidently state that hierarchical
procedural rules are associated with smaller government debt.
Conversely, transparency is shown to lower government debt at a statistically
significant level in only one model specification (model 12). The effect of transparency
among this sample is similar to the full sample (model 5 Table 6.7) of lowering debt by
an estimated 1.06% of GDP, but the effect is not precisely estimated (p < 0.10). While
the effect of transparency may not be robust in this study, it could benefit from an
increased sample size. Note that this finding is statistically significant in the largest two
samples.
157
Therefore, it may be that the sample size of this study is too small and there is
simply not enough power to detect the effect. This is another area that warrants future
exploration.
Models 13 though 16 examine the robustness of the finding that fiscal rules
limiting the amount of borrowing the government can undertake are related to lower debt
levels, provided the budget process is sufficiently transparent. According to the figures
reported in Table 6.9b, this result is fairly stable. In three of the four models fiscal debt
rules are negatively related to debt at low, but nevertheless statistically significant levels
(p < 0.10). Furthermore, between these models the magnitude of the effect is as great as

157
Observe that because I use robust regression in these estimations, the results are unlikely to be due to an
outlier entering the sample as the number of observations increases.

207
or greater than that previously estimated; among countries above the 75
th
percentile in
budget transparency, fiscal debt rules lower debt between 7.12% and 7.79% of GDP.
Thus, the results further support hypothesis 2. Nevertheless, bear in mind that the samples
used to estimate this effect were especially small and the impact of fiscal debt rules may
be overstated in this study.
All in all, the results of the robustness test largely confirm the findings of the
previous sections. Hypotheses 1, 2, and 10 do not demonstrate much, if any, sensitivity to
how democracy is defined and how the sample of countries is selected according to this
criterion. With regard to budget transparency, the present findings suggest that the
relationship between transparency and debt could benefit from further testing based on a
larger sample of countries. At present, however, the result that greater budget
transparency may be related to lower government debt, while suggestive, remains
inconclusive.

4.2 Level of Development
Similar to Chapter 5, as a final sensitivity test I divide the sample in two based on
OECD membership to crudely distinguish between “developed” and “less- or under-
developed” countries. The results are reported in Table 6.9c. Among OECD countries
fiscal deficit rules have a direct and larger-than-usual effect on budget deficit levels,
although the effect disappears when the sample is restricted to OECD members with
budget processes above the 50
th
percentile in transparency. In terms of government debt,
there is no effect of fiscal debt rules among the OECD countries, even among those

208
members with budget transparency above the 75
th
percentile.
158
On the contrary, the
effect of hierarchical procedural rules is even greater among OECD members. Compared
to the full sample (where hierarchical procedural rules were estimated to lower
government debt by 1.95% of GDP), among the developed countries the effect is to lower
debt by 3.45% of GDP.
These results are stark in contrast to the effects of budget institutions on budget
deficits and debt in the sample of non-OECD members. The only budget institution found
to have any statistically significant effect is fiscal debt rules. Nevertheless, the effect is
imprecisely estimated (p < 0.10) and is only direct; the effect disappears among the
sample of countries with budget transparency above the 75
th
percentile.
159
Budget
institutions do not appear to have much of an effect among the less-developed and
underdeveloped countries in this study.
Thus, the results of this chapter appear to be driven by the advanced economies
in the sample. This is evidenced by larger effects of the budget institutions variables
among the OECD members. It may be that non-members are responsible for a downward
bias in the estimates. This finding is interesting, considering that budget institutions were
more effective among non-OECD members when it came to the level and composition of
expenditures. Therefore, the effectiveness of budget institutions varies according to the
specific combination of development level and fiscal policy outcome of interest. To my
knowledge, no other study on the relationship between budget institutions and fiscal
policy outcomes has uncovered a similar result. Exploring this relationship further is a
promising avenue for future inquiry.

158
It is important to note that there are only 9 countries in this sample, so a null-result is hardly out of the
ordinary.
159
Here, there are only 13 countries in the sample and a null-result is once again not surprising.

209
5. Conclusion
This chapter has examined the effects of budget institutions on budget deficits and
government debt. When confronted with the data, four of the six hypotheses tested
conformed to my theoretical expectations. The results provided strong evidence that
hierarchical procedural rules are associated with lower government debt. There is also
evidence that fiscal rules which limit the ability of governments to run deficits and limit
the depth and scope of government borrowing are effective at lowering both budget
deficits and debt, although the support is not as strong. Finally, there is weak evidence
that greater budget transparency may be related to less debt. Nevertheless, none of the
explanatory variables considered here have done a satisfactory job of explaining much
variance in either budget deficits or government debt. However, comparatively budget
institutions do appear to perform better than the political institutions that have been
emphasized in previous studies. The fact remains that important determinants of budget
deficits and government debt were left unaccounted for in this study.
Perhaps the most unexpected result was the finding that budget institutions are
more effective at reducing budget deficits and lowering government debt among the
OECD members. This is in contrast to the finding in the previous chapter that budget
institutions are more effective at reducing the size and composition of government among
non-OECD members. These findings suggest the possibility that budget institutions have
differential effects depending upon a country’s level of development and the particular
fiscal policy outcome to be explained. This finding, and the unexpected result from
Chapter 5 that fiscal rules and transparency affect the composition of spending, drives

210
home the argument made in Chapter 2 that studies of fiscal policy need to examine the
effects of political institutions according to an exhaustive list of fiscal policy outcomes.
In the concluding chapter I briefly summarize the findings of this study. I judge
the success of the results according to the theoretical expectations laid out in Chapter 3
and how well this study has overcome the list of shortcomings identified in other studies
discussed in Chapter 2. I then suggest potential areas for further research given this
study’s limitations, some of the surprising results, and the questions that remain
unanswered. By this time, I hope to have demonstrated the value of studying budget
institutions and fiscal policy outcomes.

211
CHAPTER 7: CONCLUSION

1. Introduction

In this chapter I evaluate the empirical findings of this study. Section 2 reviews
the results discovered in Chapters 5 and 6, as well as the lessons to be learned from these
findings. In particular, it identifies those theoretical ideas that were supported by the
empirical results and those which require more development. Also discussed are the
theoretical puzzles that remain. On the basis of this discussion, Section 3 closes the
chapter and the dissertation by suggesting some directions for future research.

2. Budget Institutions and Fiscal Policy Outcomes: Theory and Empirical Results

Chapter 3 developed a theory which predicted the effects of budget institutions on
a number of fiscal policy outcomes. From this theory, I deduced thirteen hypotheses and
predicted a series of causal links between three budget institutions—fiscal rules, budget
transparency, and procedural rules—and four fiscal policy outcomes—size of
government, budget composition, budget deficits, and government debt. Table 7.1
recapitulates these theoretical priors, which provide a checklist for comparing the results
of the empirical investigations. The columns in Table 7.1 labeled “Theory” summarize
the predicted effects of each budget institution in relation to each of the fiscal policy
outcomes of interest. The columns labeled “Data” depict the results of the empirical tests.
Entries with an “L” indicate that the expected relationship between the given budget
institution and a particular fiscal policy outcome is inverse: higher levels or a greater
degree of the budget institution is expected to be associated with lower levels the
corresponding fiscal policy outcome. Entries with an “H” indicate that greater levels of

212
the budget institution are associated with higher levels of the corresponding fiscal
outcome. “N/A” signifies that the theory in Chapter 3 did not make a theoretical
prediction between the given budget institution and a specific policy outcome. “0”
signifies that despite a theoretical prediction having been made, the empirical findings
failed to uncover a statistically significant result. Of course, the multiple theoretical
findings of the previous chapters cannot be completely summarized in a table.
Nonetheless, Table 7.1 is useful as a qualitative checklist on which theoretical priors are
matched to empirical posteriors. As such, it serves as a guide for the following
subsections, which discuss the results of each budget institution in detail.

Table 7.1
Budget Institutions and Fiscal Policy: Theory and Results
Policy
Outcome
Fiscal Rules Transparency Procedural Rules
Theory Data Theory Data Theory Data
Size of
Government L 0 L L L L
Budget
Composition
Transfers &
Subsidies N/A L N/A N/A L L
Public
Goods N/A N/A N/A L H 0
Budget
Deficits L L L 0 L 0
Government
Debt L L L 0 L L
Note: An "L" in the Theory column indicates that the budget institution in the column is expected to lead
to a smaller degree or lower level of the fiscal policy outcome in the adjoining row. Conversely, an "H"
indicates that a higher degree or higher level of the fiscal policy outcome is expected. N/A denotes that
that there is no theoretical prior related the budget institution to a particular fiscal policy outcome.
The letters in the Data columns indicate the direction of the empirically estimated effect. A "0" indicates
an inconclusive empirical result.

2.1 Fiscal Rules

In Chapter 3 a fiscal rule was defined as a policy rule that placed a permanent
constraint on fiscal policy. The most common type of fiscal rule is a numerical target

213
imposing constraints on certain budgetary aggregates. This study examined fiscal rules
that restricted the level of expenditures and revenues, limited the size of the deficit, and
rules that place ceilings on the level of public debt. It was argued that such policy rules
lend credibility to the budgeting process by making discretionary intervention less likely
and ensure policy predictability by guaranteeing that the policy will continue to be
followed regardless of the government in power. Most importantly, because fiscal rules
limit the extent to which policymakers are able to exploit fiscal policy instruments, the
common-pool resource problem may be mitigated.
Consequently, the expectation was for countries with strong fiscal expenditure,
revenue, deficit, and debt rules to be associated with lower levels of these fiscal policy
outcomes. This was provided, of course, that the level of budget transparency was
sufficiently high since previous work suggested that such rules increase the likelihood
that policymakers resort to “creative accounting” in order to circumvent such stringent
policy requirements. The only fiscal policy outcome fiscal rules were not expected to
affect was the composition of the budget, particularly the level of spending toward
transfers and subsidies and public goods. The reasoning behind this expectation is that
fiscal rules pertain to budgetary aggregates and not spending priorities. Thus, there is no
reason to suspect that fiscal rules determine in any way what the spending priorities will
be.
In most instances, the estimated effects of the various fiscal rules were in line
with my expectations. More stringent fiscal deficit and debt rules are in fact related to
better fiscal discipline. Deficit rules were estimated to reduce deficit levels by around
0.30% of GDP and debt rules were estimated to reduce debt levels by approximately 7%

214
of GDP. These effects were even more pronounced once the level of budgetary
transparency was taken into account. Furthermore, fiscal rules did not affect the level of
spending on public good provision, as expected.
There were a few instances where the results did not conform to my expectations.
In particular, no statistically significant relationship was found between fiscal
expenditure and revenue rules and the size of government. Perhaps most surprisingly, the
data uncovered a relationship between fiscal rules and transfer and subsidies spending.
Fiscal expenditure rules were found to reduce the level of expenditure toward transfers
and subsidies by 0.35% of GDP (about 4% of program size in the average country in the
sample). Furthermore, the effect is robust to the particular measure used; fiscal rules were
shown to reduce spending on social security and welfare by 0.38% of GDP. At present,
there is no theoretical explanation for this unexpected result.
Another remaining puzzle is the conditions under which these findings hold. The
effect of fiscal rules in lowering transfers and subsidies held among non-OECD members
but effectively disappeared once samples were selected from alternative democracy
indices. Conversely, the effects of fiscal rules in lowering budget deficits and government
debt were robust to the definition of democracy; however, the effect on debt vanishes
among OECD members and the effect on budget deficits fades among non-OECD
members. Thus, the efficacy of budget institutions appears to be related to level of
development. Yet, why this is so remains unclear and requires further thought and
exploration.

215
2.2 Fiscal Transparency

Fiscal transparency refers to the level of openness during the process of
formulating and implementing the budget. It provides the public at large with information
regarding government structure and functions, fiscal policy intentions, the state of public
sector accounts, and financial projections. Fiscal transparency is a desirable feature and a
requirement of good governance because politicians possess informational advantages
over voters with regard to their policy preferences and/or the ability to observe their
actions. Thus, there is the incentive to misuse fiscal policy for private and/or political
gain. The purpose of fiscal transparency is to temper this problem by making
policymakers’ actions visible and raising awareness of policymakers’ spending priorities
and revenue sources. This way, economic policy can be set with productive, rather than
just political, goals in mind.
Therefore, there was an expected direct effect of fiscal transparency on fiscal
policy outcomes, in addition to the indirect effect of preventing creative accounting in the
presence of fiscal rules. Because transparent budget institutions provide the public with a
clear sense of the government’s fiscal priorities, the state of public sector accounts, and
performance projections, citizens are able to hold public officials accountable for deviant
behavior. Fiscal restraint is more likely when responsibility is clearly established. The
expectation, then, was for expenditures, deficits, and debt to be lower under more
transparent institutions since politicians will be politically unable and/or unwilling to
excessively spend and borrow. Similarly to fiscal rules, however, fiscal transparency was
not expected to impact the composition of government spending. Transparency may

216
expose the government’s spending priorities but it is not expected to determine these
policy preferences in any way.
As expected, greater budget transparency is associated with smaller government
and lower debt. Higher levels of transparency are estimated to reduce overall spending by
0.32% of GDP and revenue by 0.34% of GDP, and reduce debt by 1.06% of GDP.
However, the latter effect is sensitive to sample size.
160
Transparency was not found to
influence budget deficits nor spending on transfers and subsidies. The lack of effect on
the latter policy outcome was expected, whereas the lack of effect on the former was not.
Nevertheless, budget transparency is still a valuable institution and may be effective in
reducing deficits (and debt), if not directly, then through its affect on fiscal rules.
Surprisingly, the data revealed a discernable effect of transparency on spending toward
public goods. Greater fiscal transparency is estimated to reduce public good provision by
approximately 0.45% of GDP. This result was not only unexpected but also contradicts
the theory set forth in Chapter 3.
Similarly to fiscal rules, the conditions under which these relationships hold
present a puzzle for future research. Whereas the effect of transparency on government
size and public good spending was stable over the different democratic samples, the
effect on debt was only present in one quarter of the models. Likewise, among OECD
countries there was no effect of transparency found on expenditure levels, public good
provision, budget deficits, or government debt. Among non-members, the only effect of
transparency was to reduce public good spending. Once more, level of development
appears to be a determining factor in the efficacy of budget institutions.

160
Because this effect is present only in the largest two samples I would say that support for this hypothesis
is “mixed” or “inconclusive;” I thus assign a “0” to the appropriate entry in Table 7.1.

217
2.3 Procedural Rules

Procedural rules are the set of protocols to be followed during the preparation,
negotiation, and execution stages of the budget. They encompass both the formal and
informal rules regulating budgetary decisions within the various branches of government.
In particular, procedural rules consist of the rules determining the formulation of the
budget within the executive, its submission to and passage through the legislature, and its
implementation by the bureaucracy. Procedural rules are important in determining fiscal
outcomes because they stipulate the relative distribution of roles and responsibilities
among the various actors at each step of the budget process, thereby distributing strategic
influence over fiscal policy. The potential to exacerbate the common-pool resource
problem exists at each stage of the budget process. Thus, the purpose of procedural rules
is to alleviate the strains on public finances caused by the common-pool resource
problem by raising awareness among policymakers of the effects of their fiscal decisions
on the government’s budget constraints.
The theory in Chapter 3 distinguished between two types of procedures based on
those actors endowed with agenda-setting power within the executive and between the
executive and the legislature. Under hierarchical procedures the central budget authority
possesses agenda-setting control over the various ministries when preparing the budget
and the executive branch possesses agenda-setting power over the legislature during the
budget approval phase. Collegial procedures possess the opposite features. As such,
hierarchical procedures were expected to lead to smaller governments, smaller budget
deficits, and less debt. This is because they grant budgetary power to the actors with an
incentive to internalize the entire costs and benefits of policy and limit the autonomy of

218
spending ministers and legislature that are less likely to consider (or be concerned with)
the potential externalities. In addition, hierarchical rules were expected to lead to greater
spending toward public goods and less spending toward transfers and subsidies. This is
because one would expect the spending priorities of the executive/CBA to prevail under
hierarchical procedures and those of the spending ministers/legislature under collegial
rules. The benefits of public goods are widely dispersed and do not favor any particular
narrow constituency. Transfers and subsidies, on the other hand, can be targeted to
special interests and specific socioeconomic and/or demographic groups. Since the CBA
and executive are more likely to be concerned with the interests of the average taxpayer
they are more likely to prefer public good provision. Conversely, spending ministers and
legislators are more likely to prefer budgets that spend more on transfers and subsidies so
as to maximize political support among their particular constituency groups.
As with the other budget institutions, more often than not procedural rules
performed as expected in the empirical tests. Procedural rules are associated with smaller
governments, less public debt, and spend less on transfers and subsidies. Specifically,
more hierarchical procedural rules reduce expenditures by 1.15% of GDP, revenues by
1.24% of GDP, lower government debt by approximately 2% of GDP, and reduce
spending on transfers and subsidies by 0.31% of GDP. Importantly, these effects are
robust to the particular definition of democracy used to sample cases. Contrary to what
was expected, however, there was no discernable effect of procedural rules with regard to
lowering deficits and increasing public good provision.
Perhaps the most unanticipated finding was that procedural rules are most
efficacious among the more advanced economies. For instance, among the OECD

219
countries the effect of procedural rules on the size of government is 30% greater than the
effect for the sample as a whole; expenditures are reduced by 1.49% of GDP among
OECD countries compared to 1.15% in the full sample. Similarly, the effect of procedural
rules on transfers and subsidy spending and public debt is 26% and 77% greater,
respectively, among the advanced countries compared to the full sample. Whereas the
effect of more hierarchical procedural rules for the average country in the sample is to
reduce transfers and subsidies spending by 0.31% of GDP and to lower debt by 1.95% of
GDP, among OECD countries the effect is to lower these respective outcomes by 0.39%
of GDP and 3.45% of GDP. Among non-members, the only effect of procedural rules is
to reduce the size of government by 0.86% of GDP; however, this effect is 25% smaller
than the effect for the average country in the sample.

2.4 Preliminary Policy Prescriptions

Overall, there is a fairly robust relationship between budget institutions and the
fiscal policy outcomes of interest in this study. There is also an equally robust
relationship between the efficacy of budget institutions and level of development. Form a
policy standpoint it is often costly and complicated to change institutions. Nevertheless,
provided that policymakers are sufficiently dissatisfied with the outcomes of their budget
institutions, a number of possible reforms may be suggested.
If excessive spending is a recurring problem, an institutional solution would be to
increase the level of budget transparency and/or increase the degree of procedural
centralization. Increasing transparency is probably the more feasible of the two options,
although its effect on reducing spending will likely be less than if procedural rules are

220
reformed. If transparency is increased to reduce spending it will also reduce public good
spending. If procedural rules are reformed, a concomitant effect will be to lower spending
toward transfers and subsidies. Policymakers should be aware of these effects and
calibrate possible reforms accordingly. If persistent deficits are a problem, the evidence
suggests that instituting a deficit rule is sufficient to improve the budget balance.
Combining a deficit rule with increased transparency, however, would probably augment
the effect of the rule. Excessive debt may be reined in by greater centralization, increased
transparency, and/or a debt rule. Once more, increasing the level of budget transparency
is potentially the more feasible option, followed by implementation of a debt rule, and
then increased centralization. Countries with debt larger than the size of the economy (e.g.
Greece, Italy, Japan, Malawi, etc.) may require reform of all three institutions.
These policy prescriptions are preliminary and should be refined as new evidence
becomes available. The next section discusses some ways in which this research agenda
may be advanced.

3. Limitations and Future Research

Chapter 2 leveled a critique against existing theories of political institutions and
fiscal policy outcomes, and pointed out four shortcomings in particular which needed
rectification. First, I argued that there was a lack of theory which related the effects of
political institutions to an exhaustive list of fiscal policy outcomes. The existing literature
was critiqued for its inability to explain a host of effects and for its narrow focus on two
fiscal outcomes at most. Second, many theoretical explanations for fiscal policy
outcomes were shown to lack evidentiary support. Thus, the current literature attempts to

221
explain very little and by and large does a poor job at that. Lastly, two conceptual
critiques were made. On the one hand, the current literature does not focus on the proper
set of actors and their roles in determining fiscal policy; on the other hand, nor does the
literature focus on the most consequential institutions for determining fiscal policy. I
argued that all actual fiscal policy decisions are made within the confines of budget
institutions and that these institutions determine the relevant set of actors and their roles
and responsibilities in formulating and executing fiscal policy.
This study has attempted to rectify these issues by theorizing the effects of budget
institutions on five fiscal policy outcomes and testing for effects among a sample of 83
countries. As such, it represents one of the first studies of its kind in terms of scope and
country coverage.
161
Overall, the empirical results supported the importance of budget
institutions. In general, budget institutions were demonstrated to affect these fiscal
outcomes in an economically meaningful way and they faired well in comparison to
alternative institutional explanations. This study has demonstrated the necessity of
considering budget institutions when studying the effects of political institutions on fiscal
policy outcomes. In addition, this study will raise awareness of possible reforms and their
consequences among policymakers looking to improve the fiscal performance of their
countries. After all, reforming budget institutions and processes is likely to be far more
feasible than altering the legislative structure or the electoral system. Still, there are some

161
As previously discussed, most empirical work on political institutions and fiscal policy examines at most
three policy outcomes. Furthermore, most studies on budget institutions focus on a single region. For
examples focusing on the OECD see Perotti and Kontopoulos (2002); for the EU see Von Hagen and
Harden (1995) and Hallerberg et al. (2007); for Latin America see Alesina et al. (1999), Filc and
Scartascini (2004) and Hallerberg and Marier (2004); for Central and Eastern Europe see Gleich (2003); for
Africa see Gollwitzer (2011); for low-income countries more generally, see Dabla-Norris et al. (2010).

222
serious limitations to this study and hence a number of avenues remain open for this
research agenda.
The main priority is to improve the quality of the data. There are four ways in
which this may be accomplished. First, complete data needs to be gathered among the
sample of countries for which data currently exists. In most instances this means re-
surveying a handful of countries and asking for answers to only a few questions. The next
step would be to increase the sample size by including additional countries in further
rounds of surveys. This point is especially poignant with regard to the 2008 Open Budget
Survey since the sample overlaps with only 50 of the 97 countries of the OECD
survey.
162
With regard to this study, larger sample sizes may be better able to evaluate the
effect of budget transparency and government debt, since the effect only appeared in the
largest two samples. The effect of transparency may be small, and thus only detectable
with the greater statistical power that accompanies larger sample sizes. Next, budget
institutions must eventually be measured over time. This way the change within countries
may be evaluated. This task is particularly daunting given the “institutional inertia”
discussed in Chapter 4. Thus, at a minimum, budget institutions would need to be
measured over a 20 year period. Most importantly, the measures of budget institutions
need to be improved. The vast majority of existing studies develop measures of budget
institutions using different surveys that contain different questions. One step toward
improving these measures would be to standardize the questionnaires utilized by scholars.
The questionnaires developed by the OECD and the International Budget Partnership are
by far the most comprehensive in terms of coverage of countries and extensiveness of

162
To their credit the International Budget Partnership has been increasing the size of the sample for some
time now. Their original 2006 survey contained 59 countries whereas the most recent 2012 survey contains
100 countries.

223
questions posed. In my opinion, the OECD survey is lacking in questions necessary for
capturing the degree of budget transparency. Thus, one improvement would be for the
OECD survey to include the same questions as the Open Budget Survey. If combined
with extended country coverage such a survey could serve as the industry standard.
With or without improved measures, there remain a number of fruitful avenues
for further research. Two in particular present themselves as natural extensions of this
study. First, additional research needs to be conducted that explores the conditions under
which the results of this study hold. In this study, level of development was especially
influential in determining the results. Budget institutions appear to be most effective in
reducing expenditure and affecting spending composition among non-OECD members
but seem more effective in reducing deficits and debt among OECD members. Why this
is the case requires explanation. Perhaps the latter may be explained by the number of
OECD members who are also Eurozone members and therefore bound by the Maastricht
Treaty to reduce deficits to 3% of GDP and debt to 60% of GDP. Nevertheless, the
measure for development is particularly crude and it should be examined whether the
result remains with other measures.
A second area for research is to develop a better theory to explain the findings
between fiscal rules, transparency, and the composition of spending. In Chapter 5 it was
shown that fiscal rules reduce transfers and subsidies spending and transparency reduces
spending toward public goods. As discussed above, these budget institutions were not
expected to affect budget composition. Thus, the theory needs to be refined in order to
account for these findings. Note that since the early 1980s entitlement programs have
grown across the globe and encompass increasingly larger shares of government budgets

224
(Tabellini 2000). One possible explanation for the relationship between fiscal rules and
reduced transfers and subsidy spending is that since fiscal rules are intended to place
ceilings on budgetary aggregates it only makes sense for such rules to affect the spending
category with the highest growth rate and/or comprises (part of) the largest share of the
budget. With regard to transparency and public good spending, there is no immediate
theoretical explanation. There is however, one possible methodological explanation. The
conceptual distinction was made between public goods and pork-barrel projects. Yet, part
of the measure for public goods used in this study—government investment—may not
make such a distinction and may include expenditures toward pork. Thus, the statistical
tests may be capturing the effect of transparency in reducing the amount of pork-barrel
projects allocated in the budget.
163
Perhaps the distinction may be made with alternative
measures of public goods.
164
This is not to say, however, that there is no possible
theoretical explanation for this relationship. Clearly, the association between budget
institutions and budget composition could benefit from some theoretical reformulation.
More generally, empirical research on the effects of budget institutions on fiscal
policy can benefit from alternative methodologies. The vast majority of studies on budget
institutions are based on moderately-sized samples using cross-sectional ordinary least
squares regression. Improved data will surely open up research to longitudinal studies.
Yet, the research agenda could also be improved through more qualitative work. For
example, the effects discovered in the empirical chapters may be considered “black-box

163
Recall that public good spending was defined as the sum of current and capital spending on goods and
services and was measured as the sum of government wage and non-wage consumption, plus government
investment, net of depreciation. If government investment were to be taken out of the measure of public
goods then the measure would simply become government consumption and thus an entirely different
concept.
164
The measure for public goods used in this study and elsewhere (Milesi-Ferretti et al. 2002; Mukherjee
2003), however, is the only one of which I am aware.

225
explanations” since the link between the independent and dependent variables were not
given a defined structure (Hedstrom and Swedberg 1998). In other words, the empirical
results do not describe through what processes the relationships were brought about.
Through qualitative research it may be possible to gain a better understanding of the
causal mechanisms through which a country’s budget institutions influence its fiscal
policy. This could be accomplished through process-tracing or pathway type case studies
(Gerring 2007; Mahoney 2010; Collier 2011). Other types of cases studies, such as
diverse, extreme-case, deviant, most-similar-, and most-different cases, may also be
useful in generating and refining theory (Gerring 2007).
To illustrate how such a study might proceed, suppose the objective is to explore
causal mechanisms, and that the relationship of interest is between procedural rules and
the size of government. Having established that hierarchical procedural rules reduce the
size of government (see Chapter 5), suppose one wanted to examine whether the cause of
the lower level of spending is attributable to the distribution of agenda-setting power and
the favoring of the policy prerogatives of the CBA and the Executive during the budget
process (see Chapter 3).
The difficulty, of course, would be in isolating the effect of procedural rules from
alternative influences. This could be accomplished with a “pathway case” (Gerring 2007).
The objective of a pathway case is to identify representative cases where the outcome (i.e.
size of government) can be demonstrated to be strongly influenced by the theoretical
variable of interest (here, procedural rules), even while taking all other factors into
account. This way, the causal effect of the variable of interest can be isolated from other
confounding factors and the causal mechanisms can be explored (Gerring 2007: 122).

226
In the present example, to determine which cases are most strongly influenced by
procedural rules, given the effects of the economic and alternative political institutions
variables, one must compare the size of the residuals for each case in an econometric
model that excludes procedural rules (the reduced-form model), to the size of the
residuals for each case in a full model with procedural rules included. A pathway case is
the one which shows the greatest difference between the residuals from the reduced-form
model and the full model. This procedure identifies cases where the addition of
hierarchical procedural rules significantly helps “explain” the size of government for the
cases. The effect of procedural rules on the size of government could then be explored
using one or more of these cases since the level of spending could not be explained well
by the economic and alternative political institutional factors. In other words,
hierarchical procedural rules in these countries would appear to have a strong
independent effect of the size of their governments (Gerring 3007: 130).
Table 7.2 illustrates how this procedure could be used to identify pathway cases,
using a sample of 66 countries for which complete data is available on all the variables of
interest. In this example, the full model includes the procedural rules index and all of the
economic variables and political institutional variables from the previous empirical
chapters. The reduced-form model is the same, except it excludes procedural rules. The
size of government is regressed on both of these models and the residuals are found for
each country and compared. The table shows the six cases with the highest absolute
change in residuals. These are all candidates for selection since the absolute values of the
residuals from the reduced-form model are all larger than the absolute value of the
residuals from the full-model. In other words, size of government in these countries is

227
better explained once hierarchical procedural rules are included. One could, for example,
select either New Zealand and/or South Korea for further exploration, since they both
possess high differences in residuals and their full-model residuals are close to zero,
indicating that they are representative cases. It would therefore be possible to examine
these cases in more depth for the hypothesized causal mechanisms linking hierarchical
procedural rules to reduced government spending since there is strong evidence that other
factors are much less important in determining the outcome of interest.
Table 7.2
Pathway Case Example: Size of Government
Reduced-Form Full-Model
Country Residual Residual ΔResidual
New Zealand -3.20 0.19 -3.38
Kenya -4.87 -1.31 -3.56
Germany -5.25 -1.61 -3.65
South Korea -4.84 -0.96 -3.88
Cyprus 7.40 3.49 3.90
Brazil 2.87 -1.28 4.15
Reduced-form residual is the standardized residual from the reduced
model (without procedural rules).Full-model residual is the standardized
residual for a case obtained from the full model (with procedural rules).
ΔResidual = reduced-from residual minus full-model residual, listed in
order of absolute value.

Finally, the research agenda on budget institutions can be expanded to examine
their effects on other economic policy outcomes. This study only focused on fiscal policy
since budget institutions are more likely to have a direct impact on these types of
economic outcomes. However, it is possible that budget institutions affect other types of
economic policy outcomes as well, through their effect on public policy more generally.
Recent studies, for example, have examined the effects of political institutions and rent

228
extraction (corruption), and economic productivity and development.
165
The study of
budget institutions could easily be expanded to include these outcomes, as well as other
policy instruments, such as the structure of taxation, and perhaps even regulatory policy
and trade policy.
In the end, further exploration into the effects of budget institutions on economic
policy is a promising, albeit challenging, undertaking. Moving forward requires a trifecta
of rigorous theory development, meticulous data collection, and thorough empirical work.
The current state of knowledge is in the early stages and progress on this research agenda
will undoubtedly contribute to the advancement of the study of comparative political
economy.

165
For studies on political institutions and corruption see Tanzi (1998), Treisman (2000), Fisman and Gatti
(2002), Persson et al. (2003), and the sources cited therein. For studies on political institutions and
economic development see Hall and Jones (1997, 1999), Parente and Prescott (2000), and Acemoglou et al.
(2001), among others.

229
BIBLIOGRAPHY

Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2001. “The Colonial Origins
of Comparative Development.” American Economic Review 91: 1369-1401.

Acock, Alan C. 2010. A Gentle Introduction to Stata, 3
rd
Edition. College Station, TX:
Stata Press.

Adsera, Alicia and Carles Boix. “Trade, Democracy, and the Size of the Public Sector:
The Political Underpinnings of Openness.” International Organization 2002:
229-262.

Afonso, Antonio, Ludger Schuknecht and Vito Tanzi. 2005. “Public Sector Efficiency:
An International Comparison.” Public Choice 123: 321-347.

Agresti, Alan and Barbara Finlay. 1997. Statistical Methods for the Social Sciences, 3
rd

Ed. Upper Saddle River, NJ: Prentice Hall.

Alesina, Alberto. 1987. “Macroeconomic Policy in a Two-Party System as a Repeated
Game.” Quarterly Journal of Economics 102: 651-678.

_____. 1989. “Politics and Business Cycles in Industrial Democracies.” Economic
Policy 4: 55-98.

Alesina, Alberto and Silvia Ardagna. 2009. “Large Changes in Fiscal Policy: Taxes
Versus Spending.” National Bureau of Economic Research Working Paper 15438.

Alesina, Alberto, Dorian Carloni and Giampaolo Lecce. 2011. “The Electoral
Consequences of Large Fiscal Adjustments.” National Bureau of Economic
Research Working Paper No. 17655.

Alesina, Alberto, Gerald D. Cohen and Nouriel Roubini. 1992. “Macroeconomic Policy
and Elections in OECD Democracies.” Economics and Politics 5: 1-30.

Alesina, Alberto and Alex Cukierman. 1990. “The Politics of Ambiguity.” Quarterly
Journal of Economics 105: 829-850.

Alesina, Alberto and Francesco Giavazzi. Forthcoming. “Fiscal Policy After the
Financial Crisis: Introduction.” in Alberto Alesina and Francesco Giavazzi (eds.)
Fiscal Policy After the Financial Crisis. Chicago, IL: University of Chicago Press,
1-24.

Alesina, Alberto, Ricardo Hausmann, Rudolf Hommes, and Ernesto Stein. 1999. “Budget
Institutions and Fiscal Performance in Latin America.” Journal of Development
Economics 59: 253-273.

230
Alesina, Alberto and Roberto Perotti. 1995. “The Political Economy of Budget Deficits.”
IMF Staff Papers 42: 1-31.

_____. 1996. “Fiscal Discipline and the Budget Process.” American Economic Review
86: 401-407.

_____. 2008. “Budget Deficits and Budget Institutions.” In James M. Poterba and Jurgen
Von Hagen (eds.) Fiscal Institutions and Fiscal Performance. Chicago, IL:
University of Chicago Press, 13-36.

Alesina, Alberto, Roberto Perotti and Jose Tavares. 1998. “The Political Economy of
Fiscal Adjustments.” Brookings Papers on Economic Activity 1: 197-266.

Alesina, Alberto and Howard Rosenthal. 1995. Partisan Politics, Divided Government,
and the Economy. New York, NY: Cambridge University Press.

Alesina, Alberto and Nouriel Roubini. 1992. “Political Cycles in OECD Economies.”
Review of Economic Studies 59: 663-688.

Alesina, Alberto and Guido Tabellini. 1989. “External Debt, Capital Flight and Political
Risk.” Journal of International Economics 27: 199-220.

_____. 1990. “A Positive Theory of Fiscal Deficits and Government Debt.” Review of
Economic Studies 57: 403-414.

Alesina, Alberto and Romain Wacziarg. 1998. “Openness, Country Size and
Government.” Journal of Public Economics 69: 305-321.

Allison, Paul D. 2000. “Multiple Imputation for Missing Data: A Cautionary Tale.”
Sociological Methods and Research 28: 301-309.

_____. 2002. Missing Data. Thousand Oaks, CA: Sage Publications, Inc.

Alt, James E. 1985. “Political Parties, World Demand, and Unemployment: Domestic and
International Sources of Economic Activity.” American Political Science Review
79: 1016-1040.

Alt, James E. and K. Alec Chrystal. 1983. Political Economics. Berkeley, CA: University
of California Press.

Alt, James E. and David Dreyer Lassen. 2006a. “Transparency, Political Polarization, and
Political Budget Cycles in OECD Countries.” American Journal of Political
Science 50: 530-550.

_____. 2006b. “Fiscal Transparency, Political Parties, and Debt in OECD Countries.”
European Economic Review 50: 1403-1439.

231
Alt, James E., David Dreyer Lassen, and Shanna Rose. 2006. “The Causes of Fiscal
Transparency: Evidence from the U.S. States.” IMF Staff Papers 53: 30-57.

Alt, James E., David Dreyer Lassen, and David Skilling. 2002. “Fiscal Transparency,
Gubernatorial Approval, and the Scale of Government: Evidence from the States.”
State Politics and Policy Quarterly 2: 230-250.

Alt, James E. and Robert C. Lowry. 1994. “Divided Government and Budget Deficits:
Evidence from the States.” American Political Science Review 88: 811-828.

_____. 2010. “Transparency and Accountability: Empirical Results from US States.”
Journal of Theoretical Politics 22: 379-406.

Alvarez, Mike, Jose Antonio Cheibub, Fernando Limongi, and Adam Przewroski. 1996.
“Classifying Political Regimes.” Studies in Comparative International
Development 31: 3-36.

Banks, Jeffrey S. 1989. “Agency Budgets, Cost Information, and Auditing.” American
Journal of Political Science 33: 670-99.

Banks, Jeffrey S., and Barry R. Weingast. 1992. “The Political Control of Bureaucracies
under Asymmetric Information.” American Journal of Political Science 36:
509-524.

Bannock, Graham, Ron Baxter, and Even Davis. 2003. Dictionary of Economics, 4
th

Edition. Princeton, NJ: Bloomberg Press.

Barnett, Steven and Rolando Ossowski. 2002. “Operational Aspects of Fiscal Policy in
Oil-Producing Countries.” International Monetary Fund Working Paper
WP/02/177.

Baron, David P. 1989. “A Noncooperative Theory of Legislative Coalitions.” American
Journal of Political Science 33: 1048-1084.

_____. 1991. “Majoritarian Incentives, Pork Barrel Programs, and Procedural Control.”
American Journal of Political Science 35: 57-90.

Baron, David P. and John Ferejohn. 1989. “Bargaining In Legislatures.” American
Political Science Review 83: 1181-1206.

Barro, Robert J. 1974. “Are Government Bonds Net Wealth?” Journal of Political
Economy82: 1095-1117.

_____. 1979. “On the Determination of the Public Debt.” Journal of Political Economy
87: 940-971.

232
Barro, Robert J. and David B. Gordon. 1983. “A Positive Theory of Monetary Policy in a
Natural Rate Model.” Journal of Political Economy 91: 589-610.

Bawn, Kathleen and Frances Rosenbluth. 2006. “Short versus Long Coalitions: Electoral
Accountability and the Size of the Public Sector.” American Journal of Political
Science 50: 251-265.

Bawn Kathleen and Frances Rosenbluth. 2003. “Coalition Parties versus Coalitions of
Parties: How Electoral Agency Shapes the Political Logic of Costs and Benefits.”
Presented at the 2003 Annual Meeting of the American Political Science
Association.

Beck, Morris. 1976. “The Expanding Public Sector: Some Contrary Evidence.” National
Tax Journal 29: 15-21.

_____. 1979. “Public Sector Growth: A Real Perspective.” Public Finance 34: 313-356.

Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh. 2001.
“New Tools in Comparative Political Economy: The Database of Political
Institutions.” World Bank Economic Review 15: 165-176.

Bendor, Jonathan, Serge Taylor and Roland Van Gaalen. 1985. “Bureaucratic Expertise
versus Legislative Authority: A Model of Deception and Monitoring in Budgeting.
American Political Science Review 79: 1041-1060.

_____. 1987a. “Politicians, Bureaucrats, and Asymmetric Information.” American
Journal of Political Science 31: 796-828.

_____. 1987b. “Stacking the Deck: Bureaucratic Missions and Policy Design.” American
Political Science Review 81: 873-896.

Berry, William D. and David Lowery. 1984a. “The Growing Cost of Government: A Test
of Two Explanations.” Social Science Quarterly 65: 735-749.

_____. 1984b. “The Measurement of Government Size: Implications for the Study of
Government Growth.” The Journal of Politics 46: 1193-1206.

_____. 1987. “Explaining the Size of the Public Sector: Responsive and Excessive
Government Interpretations.” The Journal of Politics 49: 401-440.

Besley, Timothy. 2006. Principled Agents? The Political Economy of Good Government.
New York, NY: Oxford University Press.

Besley, Timothy and Anne Case. 1995a. “Incumbent Behavior: Vote-Seeking, Tax
Setting, and Yardstick Competition.” American Economic Review 85: 25-45.

233
_____. 1995b. “Does Electoral Accountability Affect Economic Policy Choices?
Evidence from Gubernatorial Term Limits.” Quarterly Journal of Economics 110:
769-798.

Besley, Timothy and Michael Smart. 2007. “Fiscal Restraints and Voter Welfare.”
Journal of Public Economics 91: 755-773.

Bird, Richard M. 1971. “’Wagner’s Law’ of Expanding State Activity.” Public Finance
26: 1-26.

Blinder, Alan S. 2004 “The Case Against the Case Against Discretionary Fiscal Policy.”
Center for Economic Policy Studies Working Paper No. 100, Princeton University.

Blinder, Alan S. and Robert M. Solow. 1972. “Does Fiscal Policy Matter?” Econometric
Research Program Research Memorandum No. 144, Princeton University.

Bodner, Todd E. (2008) “What Improves With Increased Missing Data Imputations?”
Structural Equation Modeling: A Multidisciplinary Journal 15: 651-675.

Boix, Carles, Michael Miller, and Sebastian Rosato. Forthcoming. “A Complete Data Set
of Political Regimes, 1800-2007.” Comparative Political Studies: 1-32.

Bollen, Kenneth A. 1980. “Issues in the Comparative Measurement of Political
Democracy.” American Sociological Review 45: 370-390.

Bollen, Kenneth A. and Robert W. Jackman. 1989. “Democracy, Stability, and
Dichotomies.” American Sociological Review 54: 612-621.

Bond, Trevor G. and Christine M. Fox. 2012. Applying the Rasch Model: Fundamental
Measurement in the Human Sciences, 2
nd
edition. New York, NY: Routledge.

Borcherding, Thomas E. 1985. “The Causes of Government Expenditure Growth: A
Survey of the U.S. Evidence.” Journal of Public Economics 28: 359-382.

Bradbury, John Charles and W. Mark Crain. 2001. “Legislative Organization and
Government Spending: Cross-Country Evidence.” Journal of Public Economics
82: 309-325.

Brender, Adi. 2003. “The Effect of Fiscal Performance on Local Government Election
Results in Israel: 1989-1998.” Journal of Public Economics 87: 2187-2205.

Brender, Adi and Allan Drazen. 2005. “Political Budget Cycles in New Versus
Established Democracies.” Journal of Monetary Economics 52 2171-1295.

234
_____. 2008. “How do Budget Deficits and Economic Growth Affect Reelection
Prospects? Evidence from a Large Panel of Countries.” American Economic
Review 98: 2203-2220.

Brennan, Geoffrey, and James M. Buchanan. 1977. “Towards a Tax Constitution for
Leviathan.” Journal of Public Economics 8: 255-273.

_____. 1978. “Tax Instruments and Constraints on the Disposition of Public Revenues.”
Journal of Public Economics 9: 301-318.

_____. 1980. The Power to Tax: Analytical Foundations of a Fiscal Constitution. New
York, NY: Cambridge University Press.

Breton, Albert and Ronald Wintrobe. 1975. “The Equilibrium Size of a Budget
Maximizing Bureau: A Note on Niskanen’s Theory of Bureaucracy.” Journal of
Political Economy 83: 195-207.

Brown, C. V. and P. M. Jackson. 1991. Public Sector Economics, 4
th
Edition. Oxford,
UK: Blackwell Publishers.

Buchanan, James M. and Lyn R. Flowers. 1980. Public Finances, 4
th
Edition. Homewood,
IL: Irwin Inc.

Buchanan, James M. and Richard E. Wagner. 1977. Democracy in Deficit: The Political
Legacy of Lord Keynes. New York: NY: Academic Press.

Bunch, Beverly S. 1991. “The Effects of Constitutional Debt Limits on State
Governments’ Use of Public Authorities.” Public Choice 68: 57-69.

Butler, Eamonn. 2012. Public Choice—A Primer. London, UK: The Institute of
Economic Affairs.

Cameron, David R. 1978. “The Expansion of the Public Economy: A Comparative
Analysis.” American Political Science Review 1243-1261.

Campos, J. Edgardo and Sanjay Pradhan. 2008. “Budgetary Institutions and the Levels of
Expenditure Outcomes in Australia and New Zealand.” In James M. Poterba and
Jurgen Von Hagen (eds.) Fiscal Institutions and Fiscal Performance. Chicago,
IL: University of Chicago Press, 233-263.

Carroll, Kathleen A. 1989. “Industrial Structure and Monopoly Power in the Federal
Bureaucracy: An empirical Analysis.” Economic Inquiry 27: 683-703.

_____. 1990. “Bureaucratic Competition and Inefficiency: A Review of the Evidence.”
Journal of Economic Behavior and Organization 13: 21-40.

235
Casper, Gretchen and Claudiu Tufis. 2003. “Correlation Versus Interchangeability: The
Limited Robustness of Empirical Findings on Democracy Using Highly
Correlated Data Sets.” Political Analysis 11: 196-203.

Cattell, Raymond B. 1965. “Factor Analysis: An Introduction to Essentials I. The
Purpose and Underlying Models.” Biometrics 21:190-215.

Cheibub, Jose Antonio. 2007. Presidentialism, Parliamentarism, and Democracy. New
York, NY: Cambridge University Press.

Cheibub, Jose Antonio, Jennifer Gandhi, and James Raymond Vreeland. 2010.
“Democracy and Dictatorship Revisited.” Public Choice 143: 67-101.

Chen, Jowei and Neil Malhotra. 2007. “The Law of k/n: The Effect of Chamber Size on
Government Spending in Bicameral Legislatures.” American Political Science
Review 657-676.

Collier, David. 2011. “Understanding Process Tracing.” PS: Political Science and
Politics 44: 823-830.

Collier, David and Robert Adcock. 1999. “Democracy and Dichotomies: A Pragmatic
Approach to Choices about Concepts.” Annual Review of Political Science 2:
537-565.

Collier, David and Steven Levitsky. 1997. “Democracy with Adjectives: Conceptual
Innovation in comparative research.” World Politics 49: 430-451.

Collins, Linda M., Joseph L. Schafer, and Chi-Ming Kam. 2001. “A Comparison of
Inclusive and Restrictive Strategies in Modern Missing Data Procedures.”
Psychological Methods 6: 330-351.

Cordes, Joseph J. “Reconciling Normative and Positive Theories of Government.”
American Economic Review 87: 169-172,

Coutinho, Leonor. 2011. “The Resource Curse and Fiscal Policy.” Cyprus Economic
Policy Review 5: 43-70.

Cox, Gary. 1990. “Centripetal and Centrifugal Incentives in Electoral Systems.”
American Journal of Political Science 34: 903-935.

Cukierman, Alex and Allan H. Meltzer. 1986a.“A Positive Theory of Discretionary
Policy, The Cost of Democratic Government and the Benefits of a Constitution.”
Economic Inquiry 24: 367-388.

_____. 1986b “A Theory of Ambiguity, Credibility, and Inflation Under Discretion and
Asymmetric Information.” Econometrica 54: 1099-1128.

236
Curristine, Teresa and Maria Bas. 2007. “Budgeting at a Glance in Latin America: The
Results of the 2006 OECD Survey on Budget Procedures and Practices for Latin
American Countries.” Organization of Economic Cooperation and Development
Report GOV/PGC/SBO (2007) 3.

Dabla-Norris, Era, Richard Allen, Luis-Felipe Zanna, Tej Prakash, Eteri Kvintradze,
Victor Lledo, Irene Yackovlev, and Sophia Gollwitzer. 2010. “Budget Institutions
and Fiscal Performance in Low-Income Countries.” International Monetary Fund
Working Paper WP/10/80.

Dahl, Robert A. 1971. Polyarchy: Paticiaption and Opposition. New Haven, CT: Yale
University Press.

de Figueiredo Jr., Rui J.P. 2002. “Electoral Competition, Political Uncertainty, and Policy
Insulation.” American Political Science Review 96: 321-333.

_____. 2003. “Budget Institutions and Political Insulation: Why States Adopt the Item
Veto.” Journal of Public Economics 87: 2677-2701.

De Haan, Jakob and Jan-Egbert. Sturm. 1994. “Political and Institutional Determinants of
Fiscal Policy in the European Community.” Public Choice 80: 157-172.

_____. 1997. “Political and Economic Determinants of OECD Budget Deficits and
Government Expenditures: A Reinvestigation.” European Journal of Political
Economy 13: 739-750.

Diermeier, Daniel and Timothy J. Feddersen. 1998. “Cohesion in Legislatures and the
Vote of Confidence Procedure.” American Political Science Review 92: 611-621.

Diermeier, Daniel and Keith Krehbiel. 2003. “Institutionalism as a Methodology.”
Journal of Theoretical Politics 15: 123-144.

Dixit, Avinash K. 1999. The Making of Economic Policy: A Transaction-Cost Politics
Perspective. Cambridge, MA: MIT Press.

Doorenspleet, Renske. 2000. “Reassessing the Three Waves of Democratization.” World
Politics 52: 384-406.

_____. 2005. Exploring the Structural Sources of the Fourth Wave. Boulder, CO: Lynne
Rienner Publishers, Inc.

Downs, Anthony. 1967. Inside Bureaucracy. Boston, MA: Little, Brown and Co.

Drazen, Allen. 2000. Political Economy in Macroeconomics. Princeton, NJ: Princeton
University Press.

237
_____. 2002. “Fiscal Rules from a Political Economy Perspective.” Paper presented at
IMF-World Bank Conference on Rules-Based Fiscal Policy in Emerging Market
Economies, Oaxaca, Mexico, February 2002.

Drazen, Allan and Marcela Eslava. 2006. “Pork Barrel Cycles.” National Bureau of
Economic Research Working Paper 12190.

_____. 2010. “Electoral Manipulation Via Voter-Friendly Spending: Theory and
Evidence.” Journal of Development Economics 92: 39-52.

Easterly, William. 1999. “When is Fiscal Adjustment an Illusion?” Economic Policy 14:
57-76.

Easterly, William and Sergio Rebelo. 1993. “Fiscal Policy and Economic Growth.”
Journal of Monetary Economics 32: 417-458.

Edin, Per-Anders. and Henry Ohlsson. 1991. “Political Determinants of Budget Deficits:
Coalition Effects versus Minority Effects.” European Economic Review 35:
1597-1603.

Edwards, Sebastian and Guido Tabellini. 1991. “Explaining Fiscal Policies and Inflation
in Developing Countries,” Journal of International Money and Finance 10:
S16-S48.

Eichengreen, Barry. 1992. Should the Maastricht Treaty Be Saved? Princeton Studies in
International Finance, no. 74. Princeton, NJ: Princeton University Press.

Fearon, James D. and David D. Laitin. 2003. “Ethnicity, Insurgency, and Civil War.”
American Political Science Review 97: 75-90.

Feldstein, Martin S. 2009. “Rethinking the Role of Fiscal Policy.” National Bureau of
Economic Research Working Paper 14684.

Ferejohn, John. 1999. “Accountability and Authority: Toward a Theory of Political
Accountability.” In Adam Przeworski, Susan C. Stokes, and Bernard Manin (eds.)
Democracy, Accountability, and Representation. New York, NY: Cambridge
University Press, 131-153.

Ferejohn, John and Keith Krehbiel. 1987. “The Budget Process and the Size of the
Budget.” American Journal of Political Science 31: 297-320.

Filc, Gabriel and Carlos Scartascini. 2004. “Budget Institutions and Fiscal Outcomes:
Ten Years of Inquiry on Fiscal Matters at the Research Department.” Paper
prepared for presentation at the Research Department of the Inter-American
Development Bank Ten Year Anniversary Conference, Washington, D.C. 17
September.

238
_____. 2006. “Budgetary Institutions.” In Eduardo Lora (ed.) State of State Reforms in
Latin America. Herndon, VA: World Bank Publications, 157-184.

Fisman, Raymond and Roberta Gatti. 2002. “Decentralization and Corruption: Evidence
Across Countries.” Journal of Public Economics 83: 325-345.

Frankel, Jeffrey A. 2011. “A Solution to Fiscal Procyclicality: The Structural Budget
Institutions Pioneered by Chile.” National Bureau of Economic Research
Working Paper No. 16945.

Freeman, Roger A. 1975. The Growth of American Government: A Morphology of the
Welfare State. Stanford, CA: Hoover Institution Press.

Gallagher, Michael and Paul Mitchell. 2008. The Politics of Electoral Systems. New York,
NY: Oxford University Press.

Gasiorowski, Mark J. 1996. “An Overview of the Political Regime Change Dataset.”
Comparative Political Studies 29: 469-483.

Gerring, Jonh. 2007. Case Study Research: Principles and Practices. New York, NY:
Cambridge University Press.

Gilligan, Thomas W. and John G. Matsusaka. 2001. “Fiscal Policy, Legislature Size, and
Political Parties Evidence from State and Local Governments in the First Half of
the 20
th
Century.” National Tax Journal 54: 57-81.

Gleditsch, Kristian S. and Michael D. Ward. 1997. “Double Take: A Reexamination of
Democracy and Autocracy in Modern Polities.” Journal of Conflict Resolution
41: 361-383.

Gleich, Holger. 2003. “Budget Institutions and Fiscal Performance in Central and Eastern
European Countries.” European Central Bank Working Paper no. 215.

Goodman, John B. 1991. “The Politics of Central Bank Independence.” Comparative
Politics 23: 329-349.

Gonzalez, Maria de los Angeles. 2002. “Do changes in Democracy Affect the Political
Budget Cycle? Evidence from Mexico.” Review of Development Economics 6:
204-224.

Gollwitzer, Sophia. 2011. “Budget Institutions and Fiscal Performance in Africa.”
Journal of African Economies 20: 111- 152.

Gorsuch, Richard L. 1983. Factor Analysis. Hillsdale, NJ: Lawrence Erlbaum Associates,
Publishers.

239
Graham, John W., Allison E. Olchowski, and Tamika D. Gilreath. 2007. “How Many
Imputations are Really Needed? Some Practical Clarifications of Multiple
Imputation Theory.” Prevention Science 8: 206-213.

Gramlich, Edward M. 1989. “Budget Deficits and National Saving: Are Politicians
Exogenous?” Journal of Economic Perspectives 3: 23-35.

_____. 1995. “The Politics and Economics of Budget Deficit Control: Policy Questions
and Research Questions.” In Jeffrey S. Banks and Eric A. Hanushek (eds.)
Modern Political Economy. New York, NY: Cambridge University Press,
171-190.

Grilli, Vittori, Donato Maschiandaro, and Guido Tabellini. 1991. “Political and Monetary
Institutions and Public Financial Policies in the Industrial Countries.” Economic
Policy 6: 341-392.

Gupta, Shibshankar P. 1967. “Public Expenditure and Economic Growth: A Time Series
Analysis.” Public Finance 22: 423-461.

Hall, Robert E. and Charles I. Jones. 1997. “Levels of Economic Activity Across
Countries.” American Economic Review 87: 173-177.

_____. 1999. “Why Do Some Countries Produce So Much More Output Per Worker than
Others?” Quarterly Journal of Economics 114: 83-116.

Hallerberg, Mark. 2003. “Fiscal Rules and Fiscal Policy.” In B. Guy Peters and Jon
Pierre (eds.) Handbook of Public Administration. London, UK: Sage Publications
Ltd., 393-401.

Hallerberg, Mark and Patrik Marier. 2004. “Executive Authority, the Personal Vote, and
Budget Discipline in Latin American and Caribbean Countries.” American
Journal of Political Science 48: 571-587.

Hallerberg, Mark, Rolf Strauch, and Jurgen Von Hagen. 2007. “The Design of Fiscal
Rules and Forms of Governance in European Union Countries.” European
Journal of Political Economy 23: 338-359.

Hallerberg, Mark and Jurgen Von Hagen. 2008. “Electoral Institutions, Cabinet
Negotiations, and Budget Deficits in the European Union.” In James Poterba and
Jurgen Von Hagen (eds.) Fiscal Institutions and Fiscal Performance. Chicago,
IL: University of Chicago Press, 209-32.

Hameed, Farhan. 2005. “Fiscal Transparency and Economic Outcomes.” International
Monetary Fund Working Paper WP/05/225.

Hamilton, Lawrence C. 2008. Statistics with Stata. Belmont, CA: Brooks/Cole.

240
Hardouin, Jean-Benoit. 2007. “Rasch Analysis: Estimation and Tests with raschtest.” The
Stata Journal 7: 22-44.

Harel, Ofer. 2009. “The Estimation of R
2
and Adjusted R
2
in Incomplete Data Sets Using
Multiple Imputation.” Journal of Applied Statistics 36: 1109-1118.

Heckman, James J. 1979. “Sample Selection Bias as a Specification Error.”
Econometrica 47: 153-161.

Hedstrom, Peter and Richard Swedberg. 1998. “Social Mechanisms: An Introductory
Essay.” In Peter Hedstrom and Richard Swedberg (eds.) Social Mechanisms: An
Analytical Approach to Social Theory. New York, NY: Cambridge University
Press, 1-31.

Hegre, Havard, Tanja Ellingsen, Scott Gates, and Nils Petter Gleditsch. 2001. “Toward a
Democratic Civil Peace? Democracy, Political Change, and Civil War, 1816-
1992.” American Political Science Review 95: 33-48.

Heller, William B. 1997. “Bicameralism and Budget Deficits: The Effect of
Parliamentary Structure on Government Spending.” Legislative Studies Quarterly
22: 485-516.

Hemming, Richard, Michael Kell, and Selma Mahfouz. 2002. “The Effectiveness of
Fiscal Policy in Stimulating Economic Activity—A Review of the Literature.”
International Monetary Fund Working paper WP/02/208.

Herber, Bernard P. 1975. Modern Public Finance: The Study of Public Sector Economics.
Homewood, IL: Irwin Inc.

Hercowitz, Zvi and Michel Strawczynski. 2004. “Cyclical Ratcheting in Government
Spending: Evidence from the OECD.” Review of Economics and Statistics 86:
353-361.

Hibbs Jr., Douglas A. 1977. “Political Parties and Macroeconomic Policy.” American
Political Science Review 71: 1467-1487.

_____. 1987. The American Political Economy. Cambridge, MA: Harvard University
Press.

Hockley, Graham C. 1992. Fiscal Policy: An Introduction. New York, NY: Routledge.

Holsey, Cheryl M. and Thomas E. Borcherding. 1997. “Why Does Government’s Share
of National Income Grow? An Assessment of the Recent Literature on the U.S.
Experience.” In Dennis C. Mueller (ed.) Perspectives on Public Choice. New
York, NY: Cambridge University Press, 562-589.

241
Inman, Robert P. and Michael A. Fitts. 1990. “Political Institutions and Fiscal Policy:
Evidence From the U.S. Historical Record.” Journal of Law, Economics, and
Organization 6: 79-132.

International Monetary Fund (IMF). 2001. “Brazil: Report on the Observance of
Standards and Codes (ROSC)—Fiscal Transparency Module.” International
Monetary Fund Country Report 01/217.

Jaimovich, Dany and Ugo Panizza. 2010. “Public Debt Around the World: A New Data
Set of Central Government Debt.” Applied Economic Letters 17: 19-24.

Johnson, Ronald N., and Gary D. Libecap. 1989. Agency Growth, Salaries, and the
Protected Bureaucrat.” Economic Inquiry 27: 431-451.

Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi 2010. “The Worldwide
Governance Indicators: Methodology and Analytical Issues.” World Bank Policy
Research Working Paper No. 5430.

Katzenstein, Peter J. 1985. Small States in World Markets: Industrial Policy in Europe.
Ithaca, NY: Cornell University Press.

Kiewiet, D. Roderick and Kristin Szakaly. 1996. “Constitutional Limitations on
Borrowing: An Analysis of State Bonded Indebtedness.” Journal of Law,
Economics, and Organization 12: 62-97.

Kim, Jae-On and Charles W. Mueller. 1978a. Introduction to Factor Analysis: What it is
and How to Do it. Newbury Park, CA: Sage Publications, Inc.

_____. 1978b. Factor Analysis: Statistical Methods and Practical Issues. Newbury Park,
CA: Sage Publications, Inc.

King, Gary. 1986. “How Not to Lie with Statistics: Avoiding Common Mistakes in
Quantitative Political Science.” American Journal of Political Science 30:
666-687.

King, Gary, Robert O. Keohane and Sidney Verba. 1994. Designing Social Inquiry:
Scientific Inference in Qualitative Research. Princeton, NJ: Princeton University
Press.

Kone, Susan L. and Richard F. Winters. 1993. “Taxes and Voting: Electoral Retribution
in American States.” The Journal of Politics 55: 22-40.

Kontopoulos, Yianos and Roberto Perotti. 2008. “Government Fragmentation and Fiscal
Policy Outcomes: Evidence from OECD Countries.” In James M. Poterba and
Jurgen Von Hagen (eds.) Fiscal Institutions and Fiscal Performance. Chicago,
IL: University of Chicago Press, 81-102.

242
Kopcke, Richard W., Geoffrey M.B. Tootell and Robert K. Triest. 2006. “Introduction:
The Macroeconomics of Fiscal Policy.” in Richard W. Kopcke, Geoffrey M.B.
Tootell and Robert K. Triest (eds.) The Macroeconomics of Fiscal Policy.
Cambridge, MA: MIT Press, 3-21.

Kopits, George. 2001. “Fiscal Rules: Useful Policy Framework or Unnecessary
Ornament?” International Monetary Fund Working Paper WP/01/145.

Kopits, George and Jon Craig. 1998. “Transparency in Government Operations.”
International Monetary Fund Occasional Paper No. 158.

Kopits, George and Steven Symansky. 1998. “Fiscal Policy Rules.” International
Monetary Fund Occasional Paper No. 162.

Kydland, Finn E. and Edward C. Prescott. 1977. “Rules Rather than Discretion: The
Inconsistency of Optimal Plans.” Journal of Political Economy 85: 473-491.

Larkey, Patrick D., Chandler Stolp and Mark Winer. 1981. “Theorizing About the
Growth of Government: A Research Assessment.” Journal of Public Policy 1:
157-220.

Laakso, Markku and Rein Taagepera. 1979. “`Effective’ Number of Parties: A Measure
with Application to West Europe.” Comparative Political Studies 12: 3-27.

Lasswell, Harold Dwight. 1936. Politics: Who Gets What, When, How. New York, NY:
Whittlesey House.

Levitsky, Steven and Lucan A. Way. 2002. “The Rise of Competitive Authoritarianism.”
Journal of Democracy 13: 51-65.

_____. 2010. Competitive Authoritarianism: Hybrid Regimes after the Cold War. New
York, NY: Cambridge University Press.

Lewis-Beck, Michael S. and Tom W. Rice. 1985. “Government Growth in the United
States.” The Journal of Politics 47: 2-30.

Lijphart, Arend. 1994. Electoral Systems and Party Systems: A Study of Twenty-Seven
Democracies, 1945-1990. New York, NY: Oxford University Press.

_____. 1999. Patterns of Democracy: Government Forms and Performance in Thirty-Six
Countries. New Haven, CT: Yale University Press.

Linz, Juan J. Totalitarian and Authoritarian Regimes. 2000. Boulder, CO: Lynne Rienner
Publishers, Inc.

243
Little, Roderick J. A., and Donald B. Rubin. 1989. “The Analysis of Social Science Data
with Missing Values.” Sociological Methods and Research 18: 292-326.

Lizzeri, Alessandro and Nicola Persico. 2001. “The Provision of Public Goods Under
Alternative Electoral Incentives.” American Economic Review 91: 225-239.

Lowery, David and William D. Berry. 1983. “The Growth of Government in the United
States: An Empirical Assessment of Competing Explanations.” American Journal
of Political Science 27: 665-694.

Lowry, Robert C., James E. Alt, and Karen E. Ferree. 1998. “Fiscal Policy Outcomes and
Electoral Accountability in American States.” American Political Science Review
92: 759-774.

MacKay, Robert J., and Carolyn L. Weaver. 1979. “On the Mutuality of Interest Between
Bureaus and High Demand Review Committees: A Perverse Result.” Public
Choice 34: 481-491.

Mackelprang, A.J. 1970. “Missing Data in Factor Analysis and Multiple Regression.”
Midwest Journal of Political Science. 14: 493-505.

MacRae, C. Duncan. 1977. “A Political Model of the Business Cycle.” Journal of
Political Economy 85: 239-263.

Mahoney, James. 2010. “After KKV: The New Methodology of Qualitative Research.”
World Politics 62: 120-147.

Mankiw, N. Gregory. 1998. Principles of Macroeconomics. Fort Worth, TX: The Dryden
Press.

Marshall, Monty G. Ted Robert Gurr, Christian Davenport, and Keith Jaggers. 2002.
“Polity IV, 1800-1999.” Comparative Political Studies 35: 40-45.

Marshall, Monty G. and Keith Jaggers. 2010. Polity IV Project: Political Regime
Characteristics and Transitions, 1800-2009. Fairfax, VA: George Mason
University, Center for Systemic Peace.

Masters, Geoff N. 1982. “A Rasch Model for Partial Credit Scoring.” Psychometrika 47:
149-174.

McCubbins, Matthew D., Roger G. Noll, and Barry R. Weingast. 1987. “Administrative
Procedures as Instruments of Political Control.” Journal of Law, Economics, and
Organization 3: 243-277.

244
McKnight, Patrick E., Katherine M. McKnight, Souraya Sidani, and Aurelio Jose
Figueredo. 2007. Missing Data: A Gentle Introduction. New York, NY: The
Guilford Press.

Meltzer, Allan H. and Scott F. Richard. 1981. “A Rational Theory of the Size of
Government.” Journal of Political Economy 89: 914-927.

_____. 1983. “Tests of a Rational Theory of the Size of Government.” Public Choice 41:
403-418.

Migue, Jean-Luc and Gerard Belanger. 1974. “Toward a General Theory of Managerial
Discretion.” Public Choice 17: 27-47.

Milesi-Ferretti, Gian Maria. 2000. “Good, Bad or Ugly? On the Effects of Fiscal Rules
with Creative Accounting.” International Monetary Fund Working Paper
WP/00/172.

_____. 2003. “Good, Bad or Ugly? On the Effects of Fiscal Rules with Creative
Accounting.” Journal of Public Economics 88: 377-384.

Milesi-Ferretti, Gian Maria, Roberto Perotti, and Massimo Rostagno. 2002. “Electoral
Systems And Public Spending.” Quarterly Journal of Economics 117: 609-657.

Miller, Gary J. and Terry M. Moe. 1983. “Bureaucrats, Legislators, and the Size of
Government.” American Political Science Review 77: 297-322.

Moe, Terry M. 1989. “The Politics of Bureaucratic Structure.” In John E. Chubb and Paul
E. Peterson (eds.) Can the Government Govern? Washington D.C.: Brookings
Institution Press, 267-329.

_____. 1995. “The Politics of Structural Choice: Towards a Theory of Public
Bureaucracy.” In Oliver E. Williamson (ed.) Organization Theory: From Chester
Barnard to the Present and Beyond. New York, NY: Oxford University Press,
116-153.

_____. 1997. “The Positive Theory of Public Bureaucracy.” In Dennis C. Mueller (ed.)
Perspectives on Public Choice. New York, NY: Cambridge University Press,
455-480.

Mulas-Granados, Carlos. 2004. “Economics or Politics? A Theoretical Review of the
Determinants of Fiscal Policy,” Hacienda Pública Española. Monografía 2004,
17-34.

Mukherjee, Bumba. 2003. “Political Parties and the Size of Government in Multiparty
Legislatures: Examining Cross-Country and Panel Data Evidence.” Comparative
Political Studies 36: 699-728.

245
Munck, Gerardo L. and Jay Verkuilen. 2002. “Conceptualizing and Measuring
Democracy: Evaluating Alternative Indices.” Comparative Political Studies 35:
5-34.

Musgrave, Richard A. 1959. The Theory of Public Finance: A Study in Public Economy.
New York, NY: McGraw Hill.

Niemi, Richard G., Harold W. Stanley, and Ronald J. Vogel. 1995. “State Economies and
State Taxes: Do Voters Hold Governors Accountable?” American Journal of
Political Science 39: 936-957.

Niou, Emerson M.S. and Peter C. Ordeshook. 1988. “Universalism in Congress.”
American Journal of Political Science 29: 246-258.

Niskanen, William A. 1971. Bureaucracy and Representative Government. Chicago, IL:
Aldine Press.

Nordhaus, William D. 1975. “The Political Business Cycle.” Review of Economic Studies
42: 169-190.

Olsen, Randall. 2008. “Self-Selection Bias.” In Paul J. Lavrakas (ed.) Encyclopedia of
Survey Research Methods. Thousand Oaks, CA: Sage Publications, Inc, 809-811.

Organization for Economic Co-operation and Development (OECD). n.d. OECD Budget
Practices and Procedures Survey. Available at:
http://www.oecd.org/governance/budgeting/39466141.pdf

_____. 2006. “OECD Budget Practices and Procedures Database Phase II: Final
Glossary.” Available at:http://www.oecd.org/governance/budgeting/39466131.pdf

Ostrom, Elinor. 1990. Governing the Commons: The Evolution of Institutions for
Collective Action. New York, NY: Cambridge University Press.

Paldam, Martin. 1979. “Is There an Electoral Cycle?: A Comparative Study of National
Accounts.” Scandinavian Journal of Economics 81: 323-342.

_____. 1989a. “Politics Matter After All: Testing Alesina’s Theory of RE Partisan Cycles
on Data for 17 Countries.” Aarhus University Institute of Economics Working
Paper 1989-6.

_____. 1989b. “Politics Matter After All: Testing Hibbs’ Theory of Partisan Cycles on
Data For 17 Countries.” Aarhus University Institute of Economics Working Paper
1989-8.

_____. 1997. “Political Business Cycles,” in Dennis C Mueller (ed.) Perspectives on
Public Choice. Cambridge, UK: Cambridge University Press, 342-370.

246
Parente, Stephen L. and Edward C. Prescott. 2000. Barriers to Riches. Cambridge, MA:
MIT Press.

Peltzman, Sam. 1980. “The Growth of Government.” Journal of Political Economy 23:
209-287.

_____. 1992. “Voters as Fiscal Conservatives.” Journal of Quarterly Economics 107:
327-361.

Perotti, Roberto and Yianos Kontopoulos. 2002. “Fragmented Fiscal Policy.” Journal of
Public Economics 86: 191-222.

Persson, Torsten. 2002. “Do Political Institutions Shape Economic Policy?”
Econometrica 70: 883-905.

_____. 2004. “Consequences of Constitutions.” Journal of the European Economic
Association 2: 139-161.

Persson, Torsten and Lars E. O. Svensson. 1989. “Why a Stubborn Conservative Would
Run A Deficit: Policy with Time-Inconsistent Preferences.” Quarterly Journal of
Economics 104: 325-345.

Persson, Torsten, and Guido Tabellini. 1999. “The Size and Scope of Government:
Comparative Politics with Rational Politicians.” European Economic Review 43:
699-735.

_____. 2000. Political Economics: Explaining Economic Policy. Cambridge, MA: MIT
Press.

_____. 2002. “Do Constitutions Cause Large Governments? Quasi-Experimental
Evidence.” European Economic Review 46: 908-918.

_____. 2003a. “Do Electoral Cycles Differ Across Political Systems?” Bocconi
University Innocenzo Gasparini Institute for Economic Research Working Paper
232.

_____. 2003b. The Economic Effects of Constitutions. Cambridge, MA: MIT Press.

_____. 2004. “Constitutional Rules and Fiscal Policy Outcomes.” American Economic
Review 94: 25-45.

Persson, Torsten, Gerard Roland and Guido Tabellini. 2000. “Comparative Politics and
Public Finance.” Journal of Political Economy 108: 1121-1161.

_____. 2007. “Electoral Rules and Government Spending in Parliamentary
Democracies.” Quarterly Journal of Political Science 2: 155-188.

247
Persson, Torsten, Guido Tabellini, and Francesco Trebbi. 2003. “Electoral Rules and
Corruption.” Journal of the European Economic Association 1: 958-989.

Pierson, Paul. 2000. “Increasing Returns, Path Dependence, and the Study of Politics.”
American Political Science Review 94: 251-267.

Poterba, James. 1994. “State Responses to Fiscal Crises: The Effects of Budgetary
Institutions and Politics.” Journal of Political Economy 102: 799-821.

Powell Jr., G. Bingham and Guy D. Whitten. 1993. “A Cross-National Analysis of
Economic Voting: Taking Account of the Political Context.” American Journal of
Political Science 37: 391-414.

Przeworski, Adam. 2003. “Minimalist Conception of Democracy: A Defense.” In Robert
A. Dahl, Ian Shapiro, and Jose Antonio Cheibub (eds.) The Democracy
Sourcebook. Cambridge, MA: The MIT Press, 12-17.

Przeworski, Adam, Michael E. Alvarez, Jose Antonio Cheibub, and Fernando Limongi.
2000. Democracy and Development: Political Institutions and Well-Being in the
World, 1950-1990. New York, NY: Cambridge University Press.

Rabe-Hesketh, Sophia, Anders Skrondal, and Andrew Pickles. 2004. GLLAMM Manual.
University of California-Berkeley, Division of Biostatistics, Working Paper Series
No. 160.

Reckase, Mark D. 1997. “The Past and Future of Multidimensional Item Response
Theory.”Applied Psychological Measurement 21: 25-36.

Reich, Gary. 2002. “Categorizing Political Regimes: New Data for Old Problems.”
Democratization 9: 1-24.

Reinhart, Carmen M. and Kenneth S. Rogoff. 2009. This Time is Different: Eight
Centuries of Financial Folly. Princeton, NJ: Princeton University Press.

Ricciuti, Roberto. 2004. “Political Fragmentation and Fiscal Outcomes.” Public Choice
118: 365-388.

Rodden, Jonathan. 2003. “Reviving Leviathan: Fiscal Federalism and the Growth of
Government.” International Organization 57: 695-729.

_____. 2004. “Comparative Federalism and Decentralization.” Comparative Politics 36:
481-500.

Rodrik, Dani. 1997. “Sense and Nonsense in the Globalization Debate.” Foreign Policy
107: 19-37.

248
_____ 1998. “Why Do More Open Economies Have Bigger Governments?” Journal of
Political Economy 106: 997-1032.

Rogoff, Kenneth. 1990. “Equilibrium Political Budget Cycles.” American Economic
Review 80: 21-36.

Rogoff, Kenneth and Anne Sibert. 1988. “Elections and Macroeconomic Policy Cycles.”
Review of Economic Studies 55: 1-16.

Romer, Thomas and Howard Rosenthal. 1978. “Political Resource Allocation, Controlled
Agendas, and the Status Quo.” Public Choice 33: 27-43.

_____. 1979. “Bureaucrats versus Voters: On the Political Economy of Resource
Allocation by Direct democracy.” Quarterly Journal of Economics 93: 563-587.

Rose, Shanna. 2006. “Do Fiscal Rules Dampen the Political Business Cycle?” Public
Choice 128: 407-431.

Roubini, Nouriel and Jeffrey D. Sachs. 1989a. “Political and Economic Determinants of
Budget Deficits in the Industrial Democracies.” European Economic Review 33:
903-938.

_____. 1989b. “Government Spending and Budget Deficits in the Industrialized
Countries.” Economic Policy 8: 1989: 99-132.

Rubin, Donald B. 1976. “Inference and Missing Data.” Biometrika 63: 581-592.

_____. 1996. “Multiple Imputation after 18+ Years.” Journal of the American Statistical
Association 91: 473-489.

Ruggie, John Gerard. 1982. “International Regimes, Transactions, and Change:
Embedded Liberalism in the Postwar Economic Order.” International
Organization 36: 379-415.

Rummel, R. J. 1970. Applied Factor Analysis. Evanston, IL: Northwestern University
Press.

Samuelson, Paul A. and William D. Nordhaus. 1989. Microeconomics. New York, NY:
McGraw-Hill, Inc.

Schafer, Joseph L. and John W. Graham. 2002. “Missing Data: Our View of the State of
the Art.” Psychological Methods 7: 147-177.

Schafer Joseph L. and Maren K. Olsen. 1998. “Multiple Imputation for Multivariate
Missing Data Problems: A Data Analyst’s Perspective.” Multivariate Behavioral
Research 33: 545-571.

249
Schwartz, Gerd, and Benedict Clements. 1999. “Government Subsidies.” Journal of
Economic Surveys 13: 119-147.

Shadbegian, Ronald J. 1998. “Do Tax and Expenditure Limitations Affect Local
Government Budgets? Evidence from Panel Data.” Public Finance Review 26:
118-136.

Shepsle, Kenneth and Barry Weingast. 1981. “Political Preferences for the Pork Barrel: A
Generalization.” American Journal of Political Science 25: 96-111.

Shi, Min and Jakob Svensson. 2002. “Conditional Political Budget Cycles.” Center for
Economic Policy Research Discussion Paper 3352.

_____. 2006. “Political Budget Cycles: Do They Differ Across Countries and Why?”
Journal of Public Economics 90: 1367-1389.

Shugart, Matthew Soberg, and John M. Carey. 1992. Presidents and Assemblies:
Constitutional Design and Electoral Dynamics. New York, NY: Cambridge
University Press.

Sinharay, Sandip, Hal S. Stern, and Daniel Russell. 2001. “The Use of Multiple
Imputation for the Analysis of Missing Data.” Psychological Methods 6: 317-329.

Spector, Paul E. 1982. Research Designs. Beverly Hills, CA: Sage Publications, Inc.

Stein, Ernesto, Ernesto Talvi, and Alejandro Grisanti. 2008. “Institutional Arrangements
and Fiscal Performance: The Latin American Experience.” In James Poterba and
Jurgen Von Hagen (eds.) Fiscal Institutions and Fiscal Performance. Chicago,
IL: University of Chicago Press, 103-133.

Stokey, Nancy L. 2003. “’Rules vs. Discretion’ After Twenty-Five Years.” In Mark
Gertler and Kenneth Rogoff (eds.) NBER Macroeconomics Annual 2002 Vol. 17.
Cambridge, MA: MIT Press, 9-45.

Studenmund, A.H. 2006. Using Econometrics: A Practical Guide, 6
th
Ed. Upper Saddle
River, NJ: Prentice Hall.

Taagepera, Rein. 1997. “Effective Number of Parties for Incomplete Data.” Electoral
Studies 16: 145-151.

Tabellini, Guido. 2000. “A Positive Theory of Social Security.” Scandinavian Journal of
Economics 102: 523-545.

Tabellini, Guido and Alberto Alesina. 1990. “Voting on the Budget Deficit.” American
Economic Review 80: 37-49.

250
Takane, Yoshio, and Jan de Leeuw. 1987. “On the Relationship between Item Response
Theory and Factor Analysis of Discretized Variables.” Psychometrika 52:
393-408.

Tanzi, Vito. 1998. “Corruption around the World: Cases, Consequences, Scope, and
Cures.” IMF Staff Papers 45: 559-594.

_____. 2005. “The Economic Role of the State in the 21
st
Century.” Cato Journal 25:
617-638.

_____. 2006. “Fiscal Policy: When Theory Collides with Reality.” Centre for European
Policy Studies (CEPS) Working Document No. 246, June.

Tanzi, Vito and Ludger Schuknecht. 1997. “Reconsidering the Fiscal Role of
Government: The International Perspective.” American Economic Review 87:
164-168.

_____. 2000. Public Spending in the 20
th
Century: A Global Perspective. New York, NY:
Cambridge University Press.

Tarschys, Daniel. 1975. “The Growth of Public Expenditures: Nine Modes of
Explanation.” Scandinavian Political Studies 10: 9-31.

Teorell, Jan, Marcus Samanni, Soren Holmberg, and Bo Rothstein. 2011. The QoG
Standard Dataset version 6 April 2011. University of Gothenburg: The Quality of
Government Institute, http://www.qog.pol.gu.se

Thies, Michael F. 2001. “Keeping Tabs on Partners: The Logic of Delegation in Coalition
Governments.” American Journal of Political Science 45: 580-598.

Treier, Shawn and Simon Jackman. 2008. “Democracy as a Latent Variable.” American
Journal of Political Science 52: 201-217.

Treisman, Daniel. 2000. “The Causes of Corruption: A Cross-National Study.” Journal of
Public Economics 76: 399-457.

Truxillo, Catherine. 2005. “Maximum Likelihood Parameter Estimation with Incomplete
Data.” Proceedings of the Thirteenth Annual SAS® Users Group International
Conference Paper 111-30.

Van Buuren, S. H.C. Boshuizen, and D.L.Knook. 1999. “Multiple Imputation of Missing
Blood Pressure Covariates in Survival Analysis.” Statistics in Medicine 18:
681-694.

Vanhanen, Tatu. 2000. “A New Dataset for Measuring Democracy, 1810-1998.” Journal
of Peace Research 37: 251-265.

251
Van Schuur, Wijbrandt H. 2003. “Mokken Scale Analysis: Between the Guttman Scale
and Parametric Item Response Theory.” Political Analysis 11: 139-163.

Velasco, Andres. 2000. “Debts and Deficits with Fragmented Fiscal Decision-Making.”
Journal of Public Economics 76: 105-125.

_____. 2008. “A Model of Endogenous Fiscal Deficits and Delayed Fiscal Reforms.” In
James Poterba and Jurgen Von Hagen (eds.) Fiscal Institutions and Fiscal
Performance. Chicago, IL: University of Chicago Press, 37-58.

Volkerink, Bjorn and Jakob and De Haan. 2001. “Fragmented Government Effects on
Fiscal Policy: New Evidence.” Public Choice 109: 221-242.

Von Hagen, Jurgen. 1991. “A Note on the Empirical Effectiveness of Formal Fiscal
Restraints.” Journal of Public Economics 44: 199-210.

_____. 1992. “Budgeting Procedures and Fiscal Performance in the European
Communities.” Commission of the European Communities Economic Papers no.
96, October 1992.

_____. 2002. “Fiscal Rules, Fiscal Institutions, and Fiscal Performance.” The Economic
and Social Review 33: 263-284.

_____. 2006. “Political Economy of Fiscal Institutions.” In Barry Weingast and
Donald A. Wittman (eds.) The Oxford Handbook of Political Economy. Oxford,
UK: Oxford University Press, 464-479.

Von Hagen, Jurgen and Ian J. Harden. 1995. “Budget Processes and Commitment to
Fiscal Discipline.” European Economic Review 39: 771-779.

_____. 1996. “Budget Processes and Commitment to Fiscal Discipline.” International
Monetary Fund Working Paper WP/96/78.

von Hippel, Paul T. 2007. “Regression with Missing Ys” an Improved Strategy for
Analyzing Multiply Imputed Data.” Sociological Methodology 37: 83-117.

_____. 2009. “How to Impute Interactions, Squares, and Other Transformed Variables.”
Sociological Methodology 39: 265-291.

Vreeland, James Raymond. 2008. “The Effect of Political Regime on Civil War:
Unpacking Anocracy.” Journal of Conflict Resolution 52: 401-425.

Walkey, Frank and Garry Welch. 2010. Demystifying Factor Analysis: How it Works and
How to Use it. Bloomington, IN: Xlibris Corporation.

252

Wayman, Jeffrey C. 2003. “Multiple Imputation for Missing Data: What Is It and How
Can I Use It?” Paper presented at the 2003 Annual Meeting of the American
Educational Research Association, Chicago, IL.

Wehner, Joachim. “Assessing the Power of the Purse: An Index of Legislative Budget
Institutions.” Political Studies 54: 767-785.

_____. 2010. “Cabinet Structure and Fiscal Policy Outcomes.” European Journal of
Political Research 49: 631-653.

Weingast, Barry. 1979. “A Rational Choice Perspective on Congressional Norms,”
American Journal of Political Science 23: 245-262.

Weingast, Barry R., Kenneth A. Shepsle and Christopher Johnsen. 1981. “The Political
Economy of Benefits and Costs: A Neoclassical Approach to Distributive
Politics.” Journal of Political Economy 89: 642-664.

White, Ian R., Patrick Royston, and Angela M. Wood. 2011. “Multiple Imputation Using
Chained Equations: Issues and Guidance for Practice.” Statistics in Medicine 30:
377-399.

Wildavsky, Aaron. 1974. The Politics of the Budgetary Process, Second ed. Boston, MA:
Little Brown & Co.

Wilensky, Harold L. 1976. The “New Corporatism,” Centralization, and the Welfare
State. Beverly Hills, CA: Sage Publications.

Wintrobe, Ronald. 1997. “Modern Bureaucratic Theory.” In Dennis C. Mueller (ed.)
Perspectives on Public Choice. New York, NY: Cambridge University Press,
429-454.

Wooldridge, Jeffrey M. 2008. Introductory Econometrics: A Modern Approach. Mason,
OH: South-Western Cengage Learning.

_____. 2010. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA:
The MIT Press.

Wyckoff, Paul Gary. 1990. “The Simple Analytics of Slack-Maximizing Bureaucracy.”
Public Choice 67: 35-47.

253
APPENDIX A: DATA APPENDIX

All observations are averaged over the period 1997-2007 (or the sub-period for which
data are available).

Source Abbreviations:

Data sources are indicated by the last name of the author and the year of the work in
which they appear and can be found in the bibliography. Otherwise the source of the data
is denoted by one or more of the following databases.

ADB: Asian Development Bank’s Statistical Database System.

AfDB: African Development Bank’s Statistical Data Portal.

BPP: Organization for Economic Cooperation and Development’s Budget Practices and
Procedures Database.

CEPALSTAT: The Economic Commission for Latin America and the Caribbean
statistical database.

EURO: European Union’s Eurostat Database.

FDI: World Bank’s Fiscal Decentralization Indicators.

GFS: International Monetary Fund’s Government Finance Statistics.

IFS: International Monetary Fund’s International Financial Statistics.

GRRT: International Monetary Fund’s Guide on Resource Revenue Transparency.

OBP: International Budget Partnership’s Open Budget Survey.

OSE: Organization for Economic Cooperation and Development’s StatExtracts database.

NAS: Organization for Economic Cooperation and Development’s National Account
Statistics.

WEO: International Monetary Fund’s World Economic Outlook database.

WDI: World Bank’s World Development Indicators.

List of Variables:

BICAMERAL: binary variable indicating whether a legislature is bicameral. Source:
Teorell et al. (2011).

254
BMR: dummy variable for democracy coded 1 if all of the following criteria are satisfied,
and 0 otherwise: (1) the executive is elected either directly or indirectly in popular
elections and is responsible either directly to voters or to the legislature; (2) the
legislature (or the executive if directly elected) is chosen in free and fair elections, where
(2a) “free” signifies that voters are given multiple options on ballots, and (2b) “fair”
denotes the absence of fraud and/or government electoral abuse; and (3) suffrage for a
majority of adult males. Source: Boix et al. (forthcoming).

CGV: dummy variable for democracy coded 1 if all of the following criteria are satisfied,
and 0 otherwise: (1) the chief executive is either chosen directly by popular election or
indirectly through a body that was itself popularly elected; (2) the legislature is popularly
elected; (3) there is more than one party competing in elections; (4) an alternation in
power under electoral rules identical to the ones that brought the incumbent to office has
taken place. Source: Cheibub et al. (2010).

COLONY: nominal variable indicating whether a country was a former colony and, if so,
the name of the metropolitan state. Values range from 0-to-7 where 0 = none, 1 = Dutch,
2 = Spanish, 3 = U.S., 4 = British, 5 = French, 6 = Portuguese, and 7 = Australian.
Source: Teorell et al. (2011).

CORRUPTION: worldwide governance indicator of the control of corruption measuring
“perceptions of the extent to which public power is exercised for private gain, including
both petty and grand forms of corruption, as well as ‘capture’ of the state by elites and
private interests.” Scores range from -2.5 to 2.5 with higher values corresponding to
better outcomes. Source: Kaufmann et al. (2010: 4).

DEBT: central government debt as a share of GDP. Sources OSE, WDI, Reinhart and
Rogoff (2009), and Jaimovich and Panizza (2010).

DEFICIT: central government budget deficit (if negative) or surplus (if positive) as a
share of GDP. Source: AfDB, EURO, GFS, IFS, NSA, and WDI.

EFFECTIVENESS: worldwide governance indicator index of government effectiveness
measuring “perceptions of the quality of public services, the quality of the civil service
and the degree of its independence from political pressures, the quality of policy
formulation and implementation, and the credibility of the government’s commitment to
such policies.” Scores range from -2.5 to 2.5 with higher values corresponding to better
outcomes. Source: Kaufmann et al. (2010: 4).

ENLP: effective number of legislative parties. Source: Gallagher and Mitchell (2008) and
Teorell et al. (2001).

EXP: size of government measured as the central government expenditures as a share of
GDP. Source: ADB, AfDB, GFS, and NSA.

255
FEDERAL: indicator variable equal to 1 if a country has a federal system of government
and 0 otherwise. Source: Teorell et al. (2011).

FISCRULES_E[R,B,D]: budget institution index of fiscal rules measured on a scale of 0-
to-16 for revenue (R), deficit (B), and debt (D) rules, and on a scale of 0-to-20 for
expenditure (E) rules. Source: BPP.

FISCCEN: the degree of fiscal centralization measured as the ratio of central government
spending to general government spending as shares of GDP. Source: FDI and GFS.

GASTIL: average of indices for civil liberties and political rights, where each index is
measured on a 1-to-7 scale with 1 corresponding to the highest degree of freedom and 7
denoting the lowest degree of freedom. Source: Freedom House.

GINI: Gini index of income distribution. Source: WDI.

GOVFRAC: government fractionalization measured as the probability that two deputies
picked at random from among the government parties will be from different parties.
Source: Beck et al. (2001).

LGDPPC: natural logarithm of per capita GDP. Source: WDI.

LGSTPARTY: size of the largest party in the legislature (lower house). Counts the largest
party’s number of seats divided by the legislative assembly’s total number of seats.
Source: Teorell et al. (2011).

LPOPTOT: natural logarithm of total population (in millions). Source: WDI.

MAJ: indicator variable for majoritarian electoral systems, equal to unity if all the lower
legislative chamber country is elected under plurality rule and 0 otherwise. Source: Beck
et al. (2001), Persson and Tabellini (2003b), and Teorell et al. (2011).

MAJPAR = MAJ x (1 – PRES). Source: See MAJ and PRES.

MAJPRES = MAJ x PRES. Source: see MAJ and PRES.

OECD: dummy variable equal to 1 if a country is a member of the Organization for
Economic Cooperation and Development. Source: OECD.

PGOOD: spending toward public good provision measured as the sum of government
wage consumption, non-wage consumption, and government investment net of
depreciation, as a share of GDP. Source: GFS.

POLITY: score for democracy, computed by subtracting the autocratic score from the
democracy score and ranging from +10 (strongly democratic) to -10 (strongly autocratic).
Source: Marshall and Jaggers (2010).

256
POP1564: percentage of a country’s population between 15 and 64 years of age in the
total population. Source: WDI.

POP65: percentage of the population over the age of 65 in the total population. Source:
WDI.
PRES: indicator variable for presidential forms of government, equal to 1 in countries
with presidential regimes, and 0 for all else. Source: Beck et al. (2001), Persson and
Tabellini (2003b), and Teorell et al. (2011).

PROPAR = (1 – MAJ) x (1 – PRES). Source: See MAJ and PRES.

PROPRES = (1 – MAJ) x PRES. Source: See MAJ and PRES.

PRORULES: budget institution index of procedural rules measured on a scale of 0-to-19
where higher values indicate more hierarchical procedural rules. Source: BPP.

REGION: nominal variable indicating the geographic region to which a country belongs.
Values range from 1-to-8 where 1 = Africa, 2 = East Asia, 3 = Western Europe, 4 = Latin
America and the Caribbean, 5 = Middle East/North Africa, 6 = Oceania, 7 = Former
USSR, 8 = East-Central Europe. Source: Author.

REGULATORY: worldwide governance indicator of regulatory quality measuring
“perceptions of the ability of the government to formulate and implement sound policies
and regulations that permit and promote private sector development.” Scores range from -
2.5 to 2.5 with higher values corresponding to better outcomes. Source: Kaufmann et al.
(2010: 4).

RELSUC: relative size of the upper chamber measured as the ratio of the number of seats
in the upper chamber to the total number of seats in the legislature. Source: Beck et al.
(2001).

RESOURCES: indicator variable equal to one if the countries classifies as hydrocarbon-
and/or mineral-rich. Source: GRRT.

RESRENTS: total natural resources rents. The sum of oil rents, natural gas rents, coal
rents, mineral rents, and forest rents as a share of GDP. Source: WDI.

REV: size of government measured as the ratio of central government revenues as a share
of GDP. Source: ADB, AfDB, GFS, and NSA.

RULEOFLAW: worldwide governance indicator of rule of law measuring “perceptions of
the extent to which agents have confidence in and abide by the rules of society, and in
particular the quality of contract enforcement, property rights, the police, and the courts,
as well as the likelihood of crime and violence.” Scores range from -2.5 to 2.5 with
higher values corresponding to better outcomes. Source: Kaufmann et al. (2010: 4).

257
SEMI: indicator variable for semi-proportional electoral systems, equal to 1 if the
electoral formula for electing the lower legislative chamber in a country is neither strict
plurality nor strict proportionality, and 0 otherwise. Source: Beck et al. (2001), Persson
and Tabellini (2003b), and Teorell et al. (2011).

SSW: central government expenditures on social services and welfare as a share of GDP.
Source: ADB, CEPALSTAT, GFS, and Persson and Tabellini (2003b).

STABILITY: worldwide governance indicator index of political stability and absence of
violence/terrorism which measures “perceptions of the likelihood that government will be
destabilized or overthrown by unconstitutional or violent means, including politically
motivated violence and terrorism.” Scores range from -2.5 to 2.5 with higher values
corresponding to better outcomes. Source: Kaufmann et al. (2010: 4).

TAS: transfers and subsidies spending measured as the sum of social security benefits and
other transfers to households, and subsidies to firms, as a share of GDP. Source: GFS and
WEO.

TOTSEATS: total number of seats in the legislature (lower chamber). Source: Beck et al.
(2001).

TRADE: sum of exports and imports of goods and services measured as a share of GDP.
Source: WDI.

TRANSPARENCY: budget institution index of fiscal transparency measured on a scale of
0-to-37 and where higher values indicate a higher degree of transparency. Source: BPP
and OBS.

VOICE: worldwide governance indicator index of voice and accountability which
measures “perceptions of the extent to which a country’s citizens are able to participate in
selecting their government, as well as freedom of expression, freedom of association, and
free media.” Scores range from -2.5 to 2.5 with higher values corresponding to better
outcomes. Source: Kaufmann et al. (2010: 4).

Abstract (if available)

Abstract There is much interest in the factors determining fiscal policy in order to explain existing differences in policy outcomes between countries and within countries over time. Recently, political institutions have emerged as the set of determinants thought to be most influential in explaining these differences. However, extant studies are problematic since they have not derived the full range of implications from theory, they lack sufficient evidentiary support for their claims, and/or they do not pay particular attention to the actors and institutions most influential in determining fiscal policy. This study examines the effects of budget institutions on fiscal policy outcomes. These are the set of protocols and practices used to draft, approve, and implement budgets. Since the budget explicates the Government’s fiscal plans, budget institutions are the most proximal institutions determining fiscal policy. This study uses survey responses to develop indices of three budget institutions: fiscal rules, fiscal transparency, and procedural rules. Using Ordinary Least Squares regression this study examines the effects of these institutions on the size of government, the composition of expenditures, budget deficits, and government debt among a sample of 83 countries. The results suggest that countries with stringent fiscal rules, transparent budget procedures, and centralized procedural rules, are associated with smaller governments, less transfers and subsidies spending, smaller budget deficits, and less debt. While this is not the first study to examine budget institutions and fiscal policy, it contributes to this research agenda by being the largest such study to date, in terms of both country and regional coverage, and the most comprehensive in scope with regard to the set of policy outcomes examined. Most importantly, its conclusions provide countries with a set of initial policy prescriptions for improving fiscal policy performance.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Social policy and presidential ideology in Latin America: the political economy of social spending and anti-poverty programs

PDF

Power to the masses? The rise of direct democracy in Latin America

PDF

Reconstituting the fiscal bargain: the politics of taxation in Latin America's emerging markets

PDF

Circuit breakers: how policy entrepreneurs interrupted the electric flow with Peru’s first renewable energy legislation for the grid

PDF

Developing countries in the digital era: state and business interactions for industrial upgrading

PDF

Habits and policy: the socio-cognitive foundations of foreign policy stability

PDF

Decisions to ratify the Kyoto Protocol: a Latin American perspective on poliheuristic theory

PDF

International politics and domestic institutional change: the rise of executive war-making autonomy in the United States

PDF

Essays on fiscal outcomes of cities in California

PDF

Financial crises and trade policy in developing countries

PDF

Trade and legalization in East Asia: government-business collaboration in trade dispute settlement

PDF

California's adaptation to sea level rise: incorporating environmental justice communities along the California coastline

PDF

The role of local actors in international development: how incentive types affect foreign aid effectiveness

PDF

Politics behind appointments: policy, patronage, and party systems in the Andes

PDF

Economic policy-making in a changing environment: the politics of success in Peru vis-à-vis China

PDF

Securitizing language: political elites and language policy securitization in South Korea

PDF

Do sustainability plans affect urban sustainability outcomes in Santa Monica, San Francisco, and San Jose?

PDF

Politics is something we do together: identity and institutions in U.S. elections

PDF

The impact of globalization, economics, and educational policy on the development of 21st-century skills and education in science, technology, engineering, and mathematics in Costa Rican schools

PDF

Running for office? Three different models in U.S. political campaigns & elections

Asset Metadata

Creator Pérez, Michael David (author)

Core Title Budget institutions and the positive theory of fiscal policy

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Politics and International Relations

Publication Date 09/25/2013

Defense Date 09/06/2013

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag budget institutions,fiscal policy,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Wise, Carol (committee chair), James, Patrick (committee member), Pastor, Manuel, Jr. (committee member)

Creator Email perezmd@usc.edu,xmdperezx@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-330917

Unique identifier UC11296038

Identifier etd-PerezMicha-2053.pdf (filename),usctheses-c3-330917 (legacy record id)

Legacy Identifier etd-PerezMicha-2053.pdf

Dmrecord 330917

Document Type Dissertation

Format application/pdf (imt)

Rights Pérez, Michael David; Perez, Michael David

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA