University of Southern California Dissertations and Theses

Essays on economic modeling: spatial-temporal extensions and verification

(USC Thesis Other)

Essays on economic modeling: spatial-temporal extensions and verification

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
ESSAYS ON ECONOMIC MODELING:
SPATIAL-TEMPORAL EXTENSIONS AND VERIFICATION

by

Jiyoung Park

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
(PLANNING)

August 2007

Copyright 2007 Jiyoung Park
ii

DEDICATION

For my wife, HyeJin,
my sons, Tevin and an expected baby,
my parents and parents-in-law,
and my brother.
iii

ACKNOWLEDGMENTS

Work on this dissertation was supported by the PERISHIP Dissertation Fellowships in
Hazards, Risk, and Disasters, which is granted by the Natural Hazards Center at the
University of Colorado at Boulder and the Public Entity Risk Institute with support from the
National Science Foundation and Swiss Re. Also, the author wishes to acknowledge the
support of a Dissertation Fellowship from the University of Southern California’s School of
Policy, Planning, and Development. Research work leading up to dissertation was also
supported by the United States Department of Homeland Security through the Center for
Risk and Economic Analysis of Terrorism Events (CREATE) under grant number N00014-
05-0630 and by the University of Southern California’s METRANS Transportation Center
of U.S. Department of Transportation. I am grateful for the intellectual support of Profs.
Peter Gordon, Harry W. Richardson, and James E. Moore II. However, any opinions,
findings, and conclusions or recommendations in this document are those of the author and
do not necessarily reflect views of the cited people or groups.
iv

TABLE OF CONTENTS

DEDICATION......................................................................................................................... ii
ACKNOWLEDGMENTS ...................................................................................................... iii
LIST OF TABLES..................................................................................................................vi
LIST OF FIGURES .............................................................................................................. viii
ABSTRACT ........................................................................................................................... ix
INTRODUCTION ................................................................................................................. 1
Introduction references .................................................................................................. 15

CHAPTER I: ESSAY ONE ................................................................................................ 17
1. Introduction............................................................................................................... 17
2. Overview of the Supply-driven IO Model ................................................................ 19
2.1. Demand-driven vs. Supply-driven IO models .................................................. 19
2.2. Empirical applications of Ghoshian Supply-driven Model.............................. 25
3. Debate on the Implausibility of Supply-driven IO Models....................................... 28
4. Reinterpretation of Supply-driven IO Model............................................................ 30
5. Extension of the Supply-driven IO Model................................................................ 36
6. Conclusions............................................................................................................... 40
Chapter I references....................................................................................................... 43

CHAPTER II: ESSAY TWO.............................................................................................. 45
1. Introduction............................................................................................................... 45
2. Background............................................................................................................... 47
2.1. The Jackson et al (2006) Approach................................................................... 49
2.2. Lee’s Approach (1973) ..................................................................................... 51
3. Data Reconciliation................................................................................................... 53
4. Model ........................................................................................................................56
4.1. Two-Step Approach: Adjusted Flow Model (AFM)......................................... 56
4.2. Two-Step Approach: Doubly-constrained Fratar Model .................................. 62
5. Results....................................................................................................................... 69
6. Conclusions............................................................................................................... 75
Chapter II references...................................................................................................... 77

CHAPTER III: ESSAY THREE........................................................................................ 79
1. Introduction and Issue............................................................................................... 79
2. Trade Flows Estimation and Service Industries........................................................ 80
3. Data and Model......................................................................................................... 84
4. An Application: the Case of Education Services ...................................................... 90
5. Conclusions and Remarks......................................................................................... 97
v

Chapter III references .................................................................................................. 100

CHAPTER IV: ESSAY FOUR......................................................................................... 102
1. Introduction............................................................................................................. 102
2. Aggregation Errors and NIEMO............................................................................. 103
3. Data and Model....................................................................................................... 105
4. Results..................................................................................................................... 120
4.1. NIEMO Benefits from Spatial Information .................................................... 123
4.2. NIEMO costs .................................................................................................. 126
4.3. Model Accuracy.............................................................................................. 129
5. Conclusions............................................................................................................. 129
Chapter IV references.................................................................................................. 131

CHAPTER V: ESSAY FIVE ............................................................................................ 133
1. Introduction............................................................................................................. 133
2. Flexible Input Output Models (FlexIO) .................................................................. 135
3. Flexible National Interstate Economic Model (FlexNIEMO)................................. 151
4. Conclusions and Remarks....................................................................................... 162
Chapter V references ................................................................................................... 164
REMARKS......................................................................................................................... 166
Remarks references...................................................................................................... 172
REFERENCES .................................................................................................................. 174
APPENDICES.................................................................................................................... 181
Appendix A: Chapter II Appendices........................................................................... 181
Appendix A1: Data Reconciliation Steps, SCTG and IMPLAN ..................... 181
Appendix A2: Data Reconciliation and Definitions of USC Sectors............... 182
Appendix B: Chapter III Appendix............................................................................. 198
Appendix C: Chapter IV Appendix ............................................................................ 201
Appendix D: Chapter V Appendix …………………………………………………..204

vi

LIST OF TABLES

Table I-1: General Expanded Flow Matrix of a National Economy...................................... 20

Table II-1: IMPLAN Reconciliation with 1997/2002 CFS Producer Prices,
by USC Sector ..................................................................................................... 55

Table II-2: Summary of 1997 Estimated Trade Flows between U.S. States
by Adjusted Flow Model for USC Sector 15, ‘Plastics and Rubber’ ($M.) ........ 70

Table II-3: Summary of Estimated Trade Flows between U.S. States for
2001 by Doubly-Constrained Fratar Model for USC Sector 15,
‘Plastics and Rubber’ ($M.).................................................................................. 72

Table III-1: Definition of Variables, 2000............................................................................. 83

Table III-2: Selected Variables for the GWR: USC Sector 42, Education Service............... 91

Table III-3: Results of Geographically Weighted Regression: Case of mean_agg ............... 92

Table III-4: Results of Geographically Weighted Regression: Case of Agg_USC01-
Agge_USC09...................................................................................................... 93

Table IV-1: Model estimates of NIEMO and two USIOs using value added and
error comparisons with aggregations of IMPLAN and BEA total
output (units: $m.) ........................................................................................... 109

Table IV-2: Sectoral aggregation errors between 47 USIO and 509 IMPLAN IO
(units of Xs: $).................................................................................................. 122

Table IV-3: Spatial aggregation errors between NIEMO and USIO (units of Xs: $).......... 125

Table IV-4: Estimates of X
PI
and X
PN
, and various other errors (units of Xs: $) ............... 128

Table V-1: Comparison State-by-State Effects, NIEMO and FlexNIEMO. ....................... 160

Table R-1: Empirical Applications using the Various IO models ....................................... 169

Table A. II-2a: Selected IMPLAN Sector Aggregation to Two-Digit NAICS Codes......... 184

Table A. II-2b: Aggregated 1997 BEA Benchmark: Producer/Purchaser Values and
Ratios .......................................................................................................... 185

Table A. II-2c: Sales Values Matched to SCTG 43 from 2002 Wholesale Economic
Census........................................................................................................ 187
vii

Table A. II-2d: Calculation of SCTG 43 Value From 2001 IMPLAN Using
Economic Census........................................................................................ 188

Table A. II-2e: Comparison of Aggregated 2001 IMPLAN with 1997_CFS:
U.S. Total, Including SCTG16 .................................................................. 190

Table A. II-2g: Definitions of USC Two-Digit Sectors....................................................... 192

Table A. II-3: Trade flows matrix between States for USC sector 15
(=SCTG sector 24) from 1997 CFS ($M.).................................................... 193

Table A. III-1: Definition of USC sector............................................................................. 198

Table A. III-2: Total Domestic Import for each state by USC service sector, 2001............ 199

Table A. III-3: Total Intrastate Flows for each state by USC service sector, 2001 ............. 200

Table A. IV-1: Comparison of value added between IMPLAN and BEA data sets............ 201

Table A. IV-2: Comparison of total output between IMPLAN and BEA data sets............. 202

Table A. IV-3: Definitions of USC Two-Digit sectors........................................................ 203
viii

LIST OF FIGURES

Figure p-1: Four Types of U.S. Input-Output Systems............................................................ 3

Figure p-2: Expansion and Extension of Traditional Input-Output Model: Space,
Time, and ............................................................................................................... 6

Figure p-3: Economic Sector Classification System Conversions (Current $)...................... 13

Figure I-1: Four possible cases according to the change of value added factors................... 33

Figure I- 2: New Equilibrium via Market due to Economic Disturbance.............................. 37

Figure II-1: Estimated U.S. Trade Flows by the Adjusted Flow Model only for the
Incomplete Trade Flows of 1997 Commodity Flow Survey for USC
Sector 15, ‘Plastics and Rubber’........................................................................ 71

Figure II-2: Estimated within-State Trade Flows via the Doubly-constrained Fratar
Model for USC Sector 15 ................................................................................... 73

Figure II-3: Estimated Trade Flows without diagonal movements via the Doubly-
constrained Fratar Model for USC Sector 15 ..................................................... 74

Figure III-1: The Estimated and Actual Domestic Imports for The USC Sector 42,
Education Service, by Each State ..................................................................... 94

Figure III-2: Estimation of State-by-State Trade Flows: Case of Education Service,
USC Sector 42 .................................................................................................. 96

Figure IV-1: Various Types of IO Models Specified .......................................................... 106

Figure IV-2: Errors Resulting from Various Types of IO models....................................... 113

Figure IV-3: Estimated Errors Resulting from Various Types of IO Models ..................... 121

Figure V-1: Expanded Matrix of National Economic Transaction Flows........................... 135

Figure V-2: Monthly Multipliers Changes for FlexNIEMO................................................ 161

Figure R-1: Various Input-Output Models with Suggested Methodologies........................ 168

Figure A. II-1: Data Reconciliation Steps, SCTG and IMPLAN ........................................ 181

ix

Figure A. V-1: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct Losses of
Foreign Exports USC Sector 10 (Coal and Petroleum Products),
August 2005................................................................................................ 204

Figure A. V-2: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct losses of
Foreign Exports USC Sector 10 (Coal and Petroleum Products),
September 2005 .......................................................................................... 205

Figure A. V-3: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct losses of
Foreign Exports USC Sector 10 (Coal and Petroleum Products),
October 2005 .............................................................................................. 206

Figure A. V-4: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct losses of
Foreign Exports USC Sector 10 (Coal and Petroleum Products),
November 2005 .......................................................................................... 207

Figure A. V-5: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct Losses of
Foreign Exports in USC Sector 10(Coal and Petroleum Products),
December 2005........................................................................................... 208

Figure A. V-6: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct Losses of
Foreign Exports in USC sector 10 (Coal and Petroleum Products),
January 2006............................................................................................... 209

Figure A. V-7: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct Losses of
Foreign Exports in USC Sector 10(Coal and Petroleum Products),
February 2006............................................................................................. 210

Figure A. V-8: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct Losses of
Foreign Exports in USC Sector 10(Coal and Petroleum Products),
March 2006................................................................................................. 211

Figure A. V-9: The State-by-State Economic Impacts of the Customs District
of Louisiana: Application of FlexNIEMO, Direct Losses of
Foreign Exports in USC Sector 10(Coal and Petroleum Products),
Total Impacts Between Aug. 05 and Mar. 06.. ........................................... 212

x

Figure A. V-10: The State-by-State Economic Impacts of the Customs District of
Louisiana: Application of NIEMO, Direct Losses of Foreign
Exports in USC Sector 10(Coal and Petroleum Products),
Total Impacts Between Aug. 05 and Mar. 06 ........................................... 213

ix

ABSTRACT

Ever since the 2001 terrorist attacks on the U.S., several studies have evaluated the
socioeconomic impacts on the U.S. economy. Economic impacts are not restricted to just the
immediate impact area; rather, they would involve spatial incidence, spreading via social and
economic linkages. Therefore, to be useful for policy makers, impact analyses should
include information on the nature of spatially distributed impacts throughout the national
economy, especially for the evaluation of alternate defensive and mitigation measures. To
estimate such spatial impacts, the usefulness of interstate multiregional input-output (MRIO)
type models is clear. Also, it is also well known, as the various studies using IO-type models
have noted, that a key limitation of IO models is that all the coefficients in such models are
fixed. The constant coefficients in the IO matrix ignore substitution opportunities and/or
different relations between industries that might be prompted by market signals. Therefore,
IO model applications are only useful for impact analyses relevant to very short periods,
where we can reasonably assume that most market behaviors do not change. If some of the
assumptions of IO models can be relaxed so that they are relevant beyond the short-run, their
usefulness and application would increase tremendously.
Therefore, this dissertation addresses methodologies on spatial and/or temporal
economic IO models extended from the classic IO model, as well as from a temporally
extended national IO model. Also, this dissertation includes an essay that offers theoretical
support to the supply-driven IO model, especially in light of some well known interpretation
controversies. Hence, this dissertation includes three essays on extensions of supply-driven
IO and spatial expansions of the classic demand-driven IO: ‘The Supply-Driven Input-
x

Output Model: A Reinterpretation and Extension’, ‘A Two-Step Approach Estimating State-
by-State Commodity Trade Flows’, ‘Estimation of Interstate Trade Flows for Service
Industries’. A fourth essay, ‘An Evaluation of Input-Output Aggregation Error Using a New
MRIO Model’, provides information on the accuracy of the National Interstate Economic
Model (NIEMO) an operational multi-regional input-output model developed as part of this
research . The final essay on ‘Constructing a Flexible National Interstate Economic Model
(FlexNIEMO)’ deals with the temporal extensions of all types of IO models.
Several academic papers applying those models to hypothetical terrorist attacks or
natural disasters are being readied for publication.
1

INTRODUCTION

Ever since the 2001 terrorist attacks on the U.S., several studies have evaluated the
socioeconomic impacts on the U.S. economy from these and also various hypothetical
attacks. Although recent studies (see for example, Planning Scenarios; Howe, 2004)
illustrate preliminary estimates of the losses from various hypothetical terrorist attacks on
selected major targets, such reports typically contain no spatial information. However,
economic impacts from man-made or natural disasters would not be restricted to a just the
immediate impact area; rather, they involve spatial incidence, spreading via various social
and economic linkages.
Clearly, spatial impact information is important for at least two reasons in discussions of
homeland security.

i. Political representatives have an obvious interest in their own constituency and
jurisdiction.
ii. Subnational impacts can cancel each other in the aggregate, causing national
measures to obscure key dimensions of events.

Therefore, to be useful for policy makers, impact analyses should include information on the
nature of spatially distributed impacts throughout the national economy, especially for
alternate defensive and mitigation measures.
To identify such spatial impacts, they should be estimated by tracing impulses through
inter-related industries as well as via inter-regional commodity flows. In this sense, the
2

usefulness of interstate multiregional input-output (MRIO) type models is clear. Many
mitigation and precautionary approaches to disasters are conducted at the local level, and
therefore, those various hypothetical impacts cannot easily be evaluated unless subnational
effects can be estimated.
It is also well known, as the various studies using IO-type models have noted, that a key
limitation of IO models is that all coefficients in the models are fixed. The constant
coefficients in the IO matrix ignore substitution opportunities and/or different relations
between industries that might be prompted by market signals. Therefore, IO model
applications are only useful for impact analyses relevant to very short periods, where we
assume that most market behaviors do not change. If some of the assumptions of IO models
can be relaxed so that they are relevant beyond the short-run, their usefulness and application
would increase tremendously. Therefore, this dissertation investigates spatial and temporal
expansions of the classic IO model. Also, because temporal expansion requires the use of an
additional type of IO model, that is, a supply-driven IO model, this dissertation includes the
development of theoretical support for some applications of the supply-driven IO model.
To understand the characteristics and differences of IO systems, I have summarized the
various IO models into four types, as in Figure p-1. The traditional Leontief IO system
reveals the interrelations between industries, taking account of ‘technical’ relations
throughout an economy via fixed-coefficient production functions. Hence, this classic IO
model is widely referred to as the demand-driven input-output model. However, Ghosh
(1958) noted that demand-driven IO models require many restricting conditions: “so long as
there is no scarce factor and so long as suppliers are able to offer more of any commodity at
the existing price” (ibid: p.59) on the production side. He suggested another version of the
3

IO model to identify the interrelationships among industries, relaxing the cited restricting
conditions, the so-called supply-driven IO model.

Figure p-1. Four Types of Input-Output Systems
Traditional Leontief
(Demand-driven) Input-
Output Model
Ghosh ’s Supply-driven
Input-Output Model
Demand-dirven
Multiregional Input -
ou tpu t ( MRIO ) Mo del
Supply-dirven
Multiregional Input -
ou tput ( MRIO) Model
Spatial Expansion

I-1 Figure p-1. Four Types of U.S. Input-Output Systems

The basic IO systems present at least three problems. First, to examine the full costs
throughout the U.S. economy using an integrated model obscures spatial connections
between the states that should be considered. The best known examples for the U.S. which
considered those limits are the 1963 U.S. data sets for 51 regions and 79 sectors published in
Polenske (1980), and the 1977 U.S. data sets for 51 regions and 120 sectors released by Jack
Faucett Associates (1983) and updated by various Boston College researchers and reported
4

in 1988 (Miller and Shao, 1990). Those spatially disaggregated IO models are constructed
based on demand-driven type models. While Shao and Miller (1990) elaborated the structure
of the supply-side commodity-industry MRIO for the U.S., Bon (1988) elaborated the
generalized theoretical approach to constructing the supply-driven MRIO. However, a
common problem in these extended models is that it is not easy to develop bridge
conversions between different sector classification systems.
Another major problem in developing a spatially expanded IO model for the U.S. is very
limited data representing interstate U.S. trade flows (Lahr, 1993). This difficulty actually has
two parts: state-to-state trade flows for physical commodity sectors and (non-physical)
service sectors. After discontinuing the U.S. Commodity Transportation Survey Data on
interregional trade flows for some years, the Bureau of Transportation Statistics (BTS) have
released commodity flow survey (CFS) data since 1993. Although the CFS data have been
widely used, they have several inherent problems (Erlbaum and Holguin-Veras, 2005).

The CFS data do not include complete trade flows even between states.
When considering information on service sector trade flows, the use of CFS
data is problematic because they only report physical commodity trade flows.

Due to the limited information on interstate trade in services, in many cases, gravity
models have been widely applied to the estimation of service trade flows, reflecting
presumed distance-decay effects. Indeed, territorial location is important for trading, and
gravity models typically reflect Euclidean distance. However, the possibly overused log-
transformed gravity model is based on ordinary least squares (OLS), and hence ignores
5

‘spatial dependency’ and ‘heteroskedasticity’ resulting from many inherently invisible
characteristics of each region (Anselin, 1988; LeSage 1999). Therefore, a more sophisticated
spatial econometric approach to capturing spatial effects is critical for estimating trade flows
(Porojan, 2001).
Finally, after Ghosh (1958) developed the supply-driven IO model, there was some
debate over its plausibility (Oosterhaven, 1988; 1989; 1996; Rose and Allison, 1989; Gruver,
1989). More recently, Dietzenbacher (1997) showed that the various implausibility issues are
mitigated if the Ghosh model is interpreted as a price model, possibly equivalent to
Leontief’s price model. Although this interpretation by Dietzenbacher provides a theoretical
defense against the criticism of the supply-driven model, it places limits on the empirical
applications of impact analyses.
Based on the conventional IO models, therefore, this dissertation will include five essays
expanding the traditional U.S. national IO model spatially and temporally, in addition to
extending the supply-driven models. The five spatial and temporal extensions are outlined in
Figure p-2. The map of this research is based on two models that were developed as prelude
to this work and were used throughout. These are the National Interstate Economic Model
(NIEMO) an MRIO for the 50 U.S. states and the District of Columbia (for 2001) and USIO,
a national aggregate of NIEMO. Both involve 47 economic sectors.

6

I-2Figure p-2. Expansion and Extension of Traditional Input-Output Model: Space, Time, and
Supply-Driven IO.
Traditional Leontief
(Demand-driven) Input-
Output Model :
Demand -driven U.S.
National IO Model
(USIO)
Ghosh ’s Supply-driven
Input-Output Model :
Supply -driven U.S.
National IO Model
(USIO)
Theoretical Support:
1. The Supply -Driven
Input -Output Model : A
New Reinterpretation
and Extension
Temporal Expansion:
5. Constructing a Flexible National Interstate
Economic Model (FlexNIEMO)
Demand-
driven
FlexUSIO
Supply-driven
FlexUSIO
Demand-
dirven
FlexNIEMO
Supply-dirven
FlexNIEMO
Evaluation of NIEMO:
4. An Evaluation of Input -Output Aggregation
Error Using a New MRIO Model
Spatial Expansion:
Estimation of Trade Flows
2. A Two-Step Approach Estimating
State-by-State Commodity Trade Flows
3. Estimation of Interstate Trade Flows
for Service Industries
Demand-dirven
Multiregional Input -
ou tpu t ( MRIO) Mo del :
Demand-driven National
Interstate Economic
Model (NIEMO)
Supply-dirven
Multiregional Input -
ou tpu t ( MRIO) Mo del :
Supply -driven National
Interstate Economic
Model (NIEMO)

7

Both NIEMO and USIO are developed for, both, demand and supply sides, as seen in
Figure p-2. This dissertation includes three essays on expansions and extensions of the
classic demand-driven IO: ‘The Supply-Driven Input-Output Model: A New Reinterpretation
and Extension’, ‘A Two-Step Approach Estimating State-by-State Commodity Trade Flows’,
‘Estimation of Interstate Trade Flows for Service Industries’. Then, a fourth essay in this
dissertation is ‘An Evaluation of Input-Output Aggregation Error Using a New MRIO
Model’, which provides information on the accuracy of NIEMO. Finally, because theoretical
suggestion of temporal extension is based on the demand- and supply-driven IO or NIEMO,
the last essay of ‘Constructing a Flexible National Interstate Economic Model
(FlexNIEMO)’ deals with the expansion of all types of IO models in Figure p-1. Brief
introductions follow.

The Supply-Driven Input-Output Model: A Reinterpretation and Extension
Most previous efforts have focused on constructing various demand-driven IO models
because of their widely accepted usefulness in regional science. After Ghosh’s suggestion of
the supply-driven IO model, a debate over its plausibility ensued. Much of this was resolved
with Dietzenbacher’s (1997) suggestion of the interpretation of a price model, similar to
Leontief’s price model but with the proviso that absolute price changes rather than relative
price changes are estimated. However, in static market equilibrium, producers will not
change current technical relationships which are based on historical sales during the
immediate period following an exogenous event. This reflects the fact that Ghosh’s supply-
driven model is in terms of monetarily expressed quantities and hence applicable when using
the supply-side IO in the circumstance of static market equilibrium with abnormal economic
8

cessations. To suggest a new interpretation for the supply-driven IO model, a four-quadrant
space of economic situations is introduced, using ‘price vs. price-quantity’ and ‘increase-
decrease’ axes. Furthermore, even in the case that normal market equilibrium is not
maintained, instead of the direct use of supply-side quantity model, Ghosh’s case can be
switched to a price-type supply-driven model, and play a role in estimating economic
impacts. To address this switching process, exogenous price elasticities of demand are
combined with the supply-driven model, adjusting quantity responses to price impacts. This
logic will underlie the theoretical background necessary to utilize the supply-side model, and
hence the first essay highlights the power and the usefulness of linear models by clarifying
the applicability of the supply model.

A Two-Step Approach Estimating State-by-State Commodity Trade Flows
A major problem in developing interregional-interindustry models, including
multiregional input-output (MRIO) models, is how to combine seemingly incompatible
databases. This research aims to estimate state-to-state commodity trade flow tables by
major industrial sectors for the U.S. useful for creating an MRIO-type National Interstate
Economic Model (NIEMO). The model is based on IMPLAN and related data for 2001.
Constructing NIEMO was challenging because of the limited availability of commodity
freight shipment data between the states. This helps to explain why a NIEMO-type model
has not been developed in recent years. As one of the two basic sets of tables along Chenery-
Moses lines necessary to construct NIEMO, interregional trade tables providing interstate
trade coefficients by industry had been available from the U.S. Commodity Transportation
Survey Data since 1977, but reporting was discontinued for some years. For the years since
9

1993, this data deficit can be met to some extent with the recent (CFS) data from the Bureau
of Transportation Statistics (BTS). Available since 1993, the CFS data are widely used, but
they have several inherent problems (Erlbaum and Holguin-Veras, 2005: 3). Recent attempts
estimate interregional trade flows using data from the 1997 Commodity Flow Survey (CFS),
based on IMPLAN data, made this clear to us. This study suggests a new approach to
estimating the trade tables (between all 50 States plus D.C. and the rest of the world)
assembled via a two-step method, to adjust for incomplete reported trade flows and to update
the adjusted trade flows by estimating values via the Doubly-constrained Fratar Model,
based on the 1997 Commodity Flow Survey (CFS) data. Reconciliation of the IMPLAN and
CFS databases present various problems that are also addressed in this essay.

Estimation of Interstate Trade Flows for Service Industries
Multiregional input-output models have been discussed for many years, but their
implementation has been rare. The limitations are mostly because of the difficulty of adding
spatial detail representing trade flows between the 50 states. Since 1993, however,
Commodity Flow Survey (CFS) data have been widely used. But these data have several
inherent problems. The most serious one is that the CFS does not report trade flows below
the state level and also these are not complete estimates of the trade flows even between the
states. To construct trade flows as the basic data set for a U.S. interstate MRIO model, there
have recently been various attempts to estimate interregional trade flows based on the 1997
CFS. However, the common problem with the all of these trials is that there has been too
little attention directed to the problems of estimating trade flows among the service sectors.
In the modern information economy, this is a serious omission. Therefore, the research
10

reported in this essay addresses new approaches to relaxing the assumption of no interstate
trade in services and, instead, proposes estimates of interstate trade flows for the service
sectors. Using Geographically Weighted Regressions (GWR) econometric analysis, this
study proposes and implements a sequence of computational and spatial econometric steps
for estimating inter-state trade flows for the major service sectors required for implementing
a U.S. interstate MRIO model. Furthermore, the approach can be expanded to examine the
economic relationships between sub-state level areas, as well as to forecast future trade
flows.

An Evaluation of Input-Output Aggregation Error Using a New MRIO Model
Studies of estimation errors in multi-regional input-output (MRIO) models have focused
on spatial aggregation problems on the assumption that two MRIO models developed by
Polenske (1980) and Jack Faucett Associates (1983) estimate direct and indirect effects
accurately. Two empirical studies based on these MRIO models reported that spatially
aggregated MRIO models, consisting of a target region and others, produce small biases.
However, before studying spatial aggregation errors, it was necessary to verify that MRIO-
type models can be substituted for an aggregate U.S. Input-Output (USIO) model. A new
MRIO-type model, called the 2001 National Interstate Economic Model (NIEMO), was
recently developed, with 47 sectors and 52 regions (50 states, D.C. and the rest of world).
For the purposes of model comparison, another 2001 spatially aggregated (USIO) model was
constructed, consisting of the same 47 sectors. This one was aggregated from the IMPLAN
national IO model which has 509 sectors. Because NIEMO is developed from the same data
sources as USIO, comparisons among the three models are appropriate. One of the
11

conclusions of this study is that a multiregional input-output model for the U.S., containing
approximately six-million multipliers, can be constructed at low cost given the assumption
that IMPLAN’s outputs are plausible. And, with respect to the estimation of overall model
accuracy, I found only relatively small errors when comparing the aggregates of all sectors
and also only minor errors on an individual sector-by-sector comparison basis. I have also
demonstrated that the sectoral aggregation required to go from IMPLAN to NIEMO imparts
only minor errors. Therefore, all things considered, it is more useful to construct an MRIO-
type model instead of a general one-region IO model which aggregates over economically
diverse subregions.

Constructing a Flexible National Interstate Economic Model (FlexNIEMO)
Spatial extensions of the classic IO model include Interregional or Multiregional Input
Output (IRIO or MRIO) models (Isard, 1951; Chenery, 1953; Moses, 1955); and empirical
models of the US constructed in the late 1970s (Polenske, 1980) and early 1980s (Jack
Faucett Associates, 1983). Park et al. (2007) constructed a new demand-driven MRIO model,
the National Interstate Economic Model (NIEMO). Park (2006) and Park et al. (2006)
elaborated a supply-driven version of NIEMO, including empirical tests. As many have
noted, a key limitation of conventional IO models is that the coefficients in the models are
fixed, and the models ignore substitution opportunities that would be prompted by market
signals. Based on the approach of Gordon et al (2006) built on both the demand- and supply-
driven models, where they relaxed the assumption of fixed coefficients in IO models by
applying the RAS method to adjust coefficient matrices to account for empirical changes in
value added and final demand. Applying this approach, I extended the classic IO model,
12

making the standard model a useful tool for studying economic resiliency. The coefficients
in the resulting IO model can be adjusted across time periods to account for substitution-
driven adjustments resulting from exogenous events such a natural disaster or a terrorist
attack. This study, therefore, suggests an approach for constructing such a ‘resilience’ IO
model, one that reflects substitution effects, based on demand- and supply-driven IO models
or NIEMO. The construction of flexible IO or FlexNIEMO describes how to utilize post-
event information on concurrent demand and value added changes to identify the
technological changes that actually occurred after a major disruption.
These five essays address some major limitations of conventional IO analysis. As
suggested previously, the spatial extensions or the expansions to a supply-driven IO model
have been difficult, because of incompatible sector system among different data sets,
incomplete trade flows, or theoretical weakness supporting supply-driven IO model. Further,
although temporal expansion of the classic IO model is crucial, it has been very hard to find
a theoretical basis for constructing the model.
Although it is not reported in this dissertation, the first requirement to be resolved for the
all extensions is to construct appropriate bridges to combine different industry classification
systems. My recent development of a conversion code system bridging various standard
industrial data classifications summarizes an economic sector classification and
reconciliation system. The brief conversion table is shown in Figure p-3. The bridge matrix
in the Figure p-3 provides a key answer on the question ‘how can we combine not easily
compatible databases?’ This has been a major problem in developing integrated
interregional-interindustry models.
13

I-3 Figure p-3. Economic Sector Classification System Conversions (Current $)
Sector
System USC SCTG BEA NAICS
IMPLAN
(2001)
SIC HS SITC WCUS
USC
SCTG C, E
BEA C, E C, E
NAICS C, E C, E A
IMPLAN
(2001)
C, E C, E A A
SIC C, W P P C, W P
HS C, E C, E A C, E C, E P
SITC C, W C, W P P P P C, W
WCUS C, W C, W P P P P C, W C, E
Notes:
C: Complete mapping
A: Available from other sources
P: Possible to create mapping
E: Mappings constructed without any weights (Bayesian allocations)
W: Mappings constructed with plausible weights informed by additional data sources

Sector Classification Systems:
USC: USC Sectors newly created
SCTG : Standard Classification of Transported Goods
(http://www.bts.gov/cfs/sctg/welcome.htm)
BEA: Bureau of Economic Analysis (http://www.bea.doc.gov)
NAICS : North American Industry Classification System
(http://www.census.gov/epcd/www/naics.html)
2001 IMPLAN: IMPLAN 509-sector codes
SIC : Standard Industrial Classification (http://www.osha.gov/oshstats/sicser.html)
HS : Harmonized System (http://www.statcan.ca/trade/htdocs/hsinfo.html)
SITC: Standard International Trade Classification available from WISERTrade
(http://www.wisertrade.org)
WCUS: Waterborne Commerce of the United States
(http://www.iwr.usace.army.mil/ndc/data/datacomm.htm)
14

While the system matrix provides bridges for all codes itemized in the matrix, I
developed the matrix in order to convert any of them to a set of 47 common sectors (29
commodity sectors and 18 service sectors), here called the ‘USC Sectors’. Therefore, this
economic conversion matrix plays a basic role in constructing all models described in this
dissertation.
The next five chapters include the essays addressing the issues discussed above and
suggested, in the same order as marked in the Figure p-2 and in the above abstracts.
15

I. INTRODUCTION REFERENCES

Anselin, T., 1988, Spatial Econometrics: Methods and Models. Dorddrecht: Kluwer
Academic Publishers.

Bon, R., 1988, Supply-side Multiregional Input-Output Models, Journal of Regional Science,
28 (1): 41-50.

Chenery, H.B., 1953, Regional Analysis, in The Structure and Growth of the Italian
Economy, edited by H.B. Chenery, P.G. Clark and V.C. Pinna, U.S. Mutual Security
Agency, Rome: 98-139.

Dietzenbacher, E., 1997, In Vindication of the Ghosh Model: A Reinterpretation as a Price
Model, Journal of Regional Science, 37: 629-651.

Erlbaum, N. and J. Holguin-Veras, 2005, Some Suggestions for Improving CFS Data
Products, Paper presented at Commodity Flow Survey (CFS) Conference, Boston Seaport
Hotel & World Trade Center, Boston, Massachusetts, Jul. 8~9.

Ghosh, A., 1958, Input-Output Approach to an Allocative System, Economica, XXV: 58-64.

Gordon, P., J.Y. Park, and H. W. Richardson, 2006, Modeling Economic Impacts in Light of
Substitutions in Household Sector Final Demand, submitted to Economic Modeling.

Gruver, G.W., 1989, On the Plausibility of the Supply-driven Input-Output Model: A
Theoretical Basis for Input-coefficient Change, Journal of Regional Science, 29, 441-
450.

Howe, D., 2004, Planning Scenarios, available at the
(http://132.160.230.113:8080/revize/repository/CSSPrototype/simplelist/Planning_Scena
rios_Exec_Summary_.pdf).

Isard, W., 1951, Interregional and Regional Input-Output Analysis: A Model of a Space
Economy, Review of Economics and Statistics, 33: 318-328.

Jack Faucett Associates, INC, 1983, The Multiregional Input-Output Accounts, 1977:
Introduction and Summary, Vol. I (Final Report), prepared for the US Department of
Health and Human Services, Washington.

Lahr, M.L., 1993, A Review of the Literature Supporting the Hybrid Approach to
Constructing Regional Input-Output Models, Economic Systems Research 5: 277-293.

Miller, R.E. and G. Shao, 1990, Spatial and Sectoral Aggregation in the Commodity-
Industry Multiregional Input-Output Model, Environment and Planning A 22: 1637-
1656.
16

Moses, L.N., 1955, The Stability of Interregional Trading Patterns and Input-Output
Analysis, American Economic Review, 45: 803-832.

Oosterhaven, J., 1988, On the Plausibility of the Supply-driven Input-Output Model, Journal
of Regional Science, 28: 203-217.

Oosterhaven, J., 1989, The Supply-driven Input-Output Model: A New Interpretation but
Still Implausible, Journal of Regional Science, 29: 459-465.

Oosterhaven, J., 1996, Leontief versus Ghoshian Price and Quantity Models, Southern
Economic Journal, 62(3): 750-759.

Park, J.Y., 2006, The Economic Impacts of a Dirty-Bomb Attack on the Los Angeles and
Long Beach Port: Applying Supply-driven NIEMO, Paper presented at 17th Annual
Meeting of the Association of Collegiate Schools of Planning, Fort Worth, TX, USA,
November 9-12.

Park, J.Y., P. Gordon, S.J. Kim, Y.K. Kim, J.E. Moore II, and H.W. Richardson, 2006,
Estimating the State-by-State Economic Impacts of Hurricane Katrina, Paper presented
at CREATE symposium: Economic and Risk Assessment of Hurricane Katrina,
University of Southern California, California, USA, August 18-19.

Polenske, K.R., 1980, The U.S. Multiregional Input-Output Accounts and Model, DC Health,
Lexington, MA.

Porojan, A., 2001, Trade Flows and Spatial Effects: The Gravity Model Revisited. Open
Economies Review 12 (3): 265-280.

Rose, A. and T. Allison, 1989, On the Plausibility of the Supply-driven Input-Output Model:
Empirical Evidence on Joint Stability, Journal of Regional Science, 29: 451-458.

Shao, G. and R.E. Miller, 1990, Demand-side and Supply-side Commodity-Industry
Multiregional Input-Output Models and Spatial Linkages in the US regional Economy,
Economic Systems Research, 2 (4): 385-405.

17

II. CHAPTER I: ESSAY ONE

The Supply-Driven Input-Output Model: A Reinterpretation and Extension

1. Introduction
Regional scientists are often engaged in the study of economic impacts (Hewings, 1985).
In recent years, the increased threat of terrorist attacks and the possibility of a higher
frequency of natural disasters due to climate change have renewed policy makers’ interest in
impact studies. Yet, have our analytic tools really improved? The picture is mixed. While we
have better data, better software and hardware than ever, is it straightforward to apply
currently existing models to those impact studies? In this essay, I argue that a fundamental
reinterpretation of some old models does shed new light and does make it possible to
develop more insightful and useful impact analyses.
In the input-output (IO) world, two standard models have been developed since
Leontief’s first contributions (1936, 1941). One is the ‘Leontief’ or demand-driven IO model,
generalizing interdependences between industries in an economy. To address “the highly
complex network of interrelationships which transmits the impulses of any local primary
change into the remotest corners of the economic system”, the general static equilibrium for
an economy can be represented (Leontief, 1976: 34). In the classic IO system, therefore, the
interrelations between industries take account of ‘technical’ relations throughout an
economy via fixed-coefficients production functions.
Two key assumptions implicit in the Leontief model, a competitive market system and
non-scarce resources, were noted by Ghosh (1958) who suggested another version of the
18

model to identify the interrelationships among industries. The technical coefficients from the
Leontief model are assumed fixed and yield new industrial total inputs necessary for an
economy as in response to changed final demands. This requires conditions as “so long as
there is no scarce factor and so long as suppliers are able to offer more of any commodity at
the existing price” (ibid: p.59) on the production side, even in the short-term. However, a
monopolistic or a centrally planned economy, where all resources are scarce except for one,
considers the best feasible combination of the non-scarce sector based on the rest of the
scarce resources with respect to a welfare function, not the optimized technical combination
of production for the non-scarce sector (Ghosh, 1964). The economic situation can, therefore,
denote a new approach to examining the ‘allocation’ of a non-scarce resource, so called
‘Ghosh’ model or supply-driven IO model. Here, two new key conditions are found for the
‘Ghosh’ model, that depart from Leontief’s assumptions:

i. Fixed (by authority or stable equilibrium market during the short run not
allowing any behavioral adjustments between industries) allocation coefficients
not affected by final demand changes.
ii. Scarce capacity for all industrial sectors except the sectors targeted.

Although the two assumptions are basic when interpreting economic conditions in order
to apply the supply-driven model, the theoretical criticisms (especially by Oosterhaven
(1988; 1989)) on the implausibility of the supply-driven model still do not contain full
considerations of them in their interpretation of specific economic conditions, and remain to
be solved.
19

The rest of this essay, therefore, will deal with

i. The difference between the Leontief and Ghoshian IO model in basic terms and
which economic conditions are taken into consideration in the empirical
applications of supply-driven models.
ii. What criticisms and defenses of supply-driven models have been offered.
iii. Whether there are new approaches possible to reinterpret the supply-driven IO
model or not.
iv. What the appropriate conditions applied to impact studies might be.

Based on these discussions, a new price-type supply-driven IO model combining
traditional supply-driven IO with price elasticities of demand is suggested, and then some
conclusions are followed.

2. Overview of the Supply-driven IO Model
2.1. Demand-driven vs. Supply-driven IO models
To discuss the demand-driven and supply-driven IO models, it is helpful to begin with
definitions of the interindustry flows matrix expanded to include final demand and value
added sectors. Table I-1 shows the national expanded economic transaction flows for an
economy, along with matrix (in parentheses) and notation.
In the general IO model, it is assumed that
T
s d
X X = , where superscript T means the
transpose of the matrix, and the standard Leontief IO model is easily expressed in matrix
form of which notations shown in Table I-1.
20

d
X =
N
Zu +
k
Yu (1.)
Because the Leontief technical coefficients are interindustry coefficients to produce total
inputs corresponding to demand requirements, the input coefficients matrix A can be
obtained from a flow matrix Z but divided by the total inputs, that is,
A =
1
)
ˆ
(
− d
X Z =
1
)
ˆ
(
− s
X Z , where X
ˆ
means the diagonal matrix of X and hence
d
T
s s
X X X
ˆ ˆ ˆ
= = . Note that the input coefficients matrix A examines the backward effects
of interindustry relationships, because the coefficient
ij
a (=
s
j ij
x z / ) is based on total input.

Table I-1 General Expanded Flow Matrix of a National Economy
NN N
ij
N
z z
z
z
z z z
1
21
1 12 11
O
M
O
L
NK N
ik
K
y y
y
y
y y y
1
21
1 12 11
O
M
O
L
) ) ( (
) ) ( (
) ) ( (
2 2 2
1 1 1
∑∑
∑∑
∑ ∑
+ =
+ =
+ =
jk
Nk Nj
d
N
jk
k j
d
jk
k j
d
y z x
y z x
y z x
M
LN L
lj
N
v v
v
v
v v v
1
21
1 12 11
O
M
O
L

-
) (
) (
) (
2 2
1 1
∑
∑
∑
=
=
=
j
Lj L
j
j
j
j
v v
v v
v v
M
s
N
s s
x x x L
2 1
1

K
y y y L
2 1
2

-
Note: 1.
∑∑
+ =
i
lj
l
ij
s
j
v z x ) (
2.
∑
=
i
ik k
y y
3. Description of notation
( Z ) (Y ) (
d
X )
(V )
(V )
(
s
X ) (Y )
21

Z is the NxN matrix of intermediate interindustry flows and its
element
ij
z denotes the deliveries in dollar values from industry sector
i to j .
Y is the NxK matrix showing various kinds of final demands and
has its element
ik
y denoting the deliveries in dollar values from
industry sector i to final users k . Generally, k contains private
consumers, governments, investments, and exports.
V denotes the LxN matrix showing various kinds of value added
factors and has its element
lj
v meaning the dollar values going to
product sector j with factor inputs l . Generally, l contains various
kinds of labor, capital, taxes by governments, and imports.
d
X denotes the monetary value column vector of total outputs for
each sector and its elements are expressed as
d
i
x , which is the column
sum of intermediate flows and final demands of sector i .
s
X denotes the monetary value row vector of total inputs for each
sector and its elements are expressed as
s
j
x , which is the row sum of
intermediate flows and value added factors of sector j , as shown in
Note 1.
V is a column vector, that is, column sum of value added factor l
and same as
N
Vu , where
T
N
u is N element unit row vector, i.e. (1, …, 1)
and superscript T means the transpose of u .
22

Y is a row vector, that is, row sum of final demands k and same as
Y u
T
N
.

Then, the equation (1.) can be rewritten as equation (2.)

T
s
X =
N
T
s
u X A
ˆ
+
k
Yu (2.1)

=
T
s
AX +
k
Yu
(2.2)

From equations (2.), it is simple to obtain the ‘Leontief’ inverse matrix which is fixed
due to the assumption of constant input coefficients matrix A as shown in equation (3).

T
s
X =
1
1
) ( u Y A I
k
−
− (3.)

If final demands change exogenously only for the k
th
final user (
k
Y ), new total inputs
necessary to satisfy the required changes can be derived via equation (4.).

T
s
X ∆ =
k
Y A I ∆ −
−1
) ( (4.)

23

Then, to obtain the ‘Ghoshian’ supply-driven model requires construction of an
allocation (output) coefficients matrix B , which allocates (sales) the current total inputs to
each sector. Hence the allocation coefficients matrix B should be measured as a fraction of
total outputs (
ij
b =
d
i ij
x z / ) in order to examine allocation processes of input industry sectors.
This examines ‘bottleneck’ effects according to change in the value added factors. Then,
B = Z X
s 1
)
ˆ
(
−
= Z X
d 1
)
ˆ
(
−
, where interindustry relationships are examined in a forward
direction. Note the relation between the allocation coefficients matrix and the Leontief
technical coefficients,
s s d d
X A X X A X B
ˆ
)
ˆ
(
ˆ
)
ˆ
(
1 1 − −
= = from
d d
X A B X Z
ˆ ˆ
= = .
Therefore, the ‘Ghoshian’ inverse allocation matrix is easily obtained from (6.) via
equation (5.), and changes of total outputs according to the changes of value added vector l
are estimated via (7.), similarly to the process executed in the ‘Leontief’ inverse solution.

s
X = Z u
T
N
+ V u
T
l
(5.1)
T
d
X = Z u
T
N
+ V u
T
l
(5.2)
= B X u
T
d T
N
ˆ
+ V u
T
l
(5.3)
= B X
T
d
+ V u
T
l
(5.4)
and hence,
T
d
X =
1
1
) (
−
− B I V u
l
T
(6.)
and
T
d
X ∆ =
1
) (
−
− ∆ B I V
l
(7.)
24

where
l
V is a row vector only containing l
th
value added factor.

According to equations (4.) and (7.), significant assumptions for both IO models are
implicit. The positive changes of total input (
T
s
X ∆ ) in equation (4.) assume that newly
required value added factors should be enough to support the changes,
T
s
X ∆ , in the
production function. In other words, from equation (4.) obtain the newly required value
added vector ] ) [(
ˆ
1
k
X
T
Y A I R V
s
∆ − = ∆
−
where
1
)
ˆ
(
−
=
s
X
s
R V X under the assumption
that
1
)
ˆ
(
−
=
s
X
X V R
s
. This requires that the condition, ≤ ∆V
V
U , should be satisfied at
least for the economy, where
V
U is the upper-bound of the available value-added factors for
all sectors in an economy (Ghosh, 1958: 58~59). Therefore, only a perfect market system
which happens to contain enough resources for all sectors can plausibly afford to support the
final demand changes.
Similarly but also differently, the supply-driven model rests on the assumption that the
forward interindustry allocation processes work providing that the new changes of final
demands via allocation distributions are only higher than the lower-bound required by final
users due to its welfare conditions. That is, ≥ ∆Y
Y
L , where
d
X
l
T
R B I V Y
ˆ
] ) ( [
1 −
− ∆ = ∆
from the equation (7.). The Y R X
d
X
d 1
)
ˆ
(
−
= , based on the definition Y X R
d
X
d
1
)
ˆ
(
−
= .
The important implication of the condition that ≥ ∆Y
Y
L is that it is not necessary any
longer to let value-added factors increase in order to increase total outputs if the required
final demands satisfy the minimum requirements. The key assumption of enough supplies to
25

produce total outputs, as shown in the Leontief IO model, does not matter any more with
respect to the welfare of users, irrespective of the inefficiency. The condition ≥ ∆Y
Y
L ,
therefore, will be useful for discussing and interpreting the ‘Ghoshian’ supply model as
shown below.
The interpretation of the changes of ‘total outputs’ according to changes of value added,
however, was severely criticized by Oosterhaven (1988) and various debates followed until
Dietzenbacher’s (1997) novel interpretation. Before looking into these debates, it useful to
review some empirical discussions in order to understand some criteria that can apply to the
‘Ghoshian’ models. Those are addressed in Section 2-2.

2.2. Empirical applications of Ghoshian Supply-driven Model
Empirical applications since Ghosh’s suggestion have been few, although there have
been many possibilities, e.g., oil shocks, cartels, earthquakes, and so on, to apply the supply-
driven model. The demand-driven IO model, meanwhile, has been widely used for various
impact analyses. Besides, ever since Oosterhaven’s consecutive criticisms (1988, 1989,
1996) on the implausibility of supply-side model theoretically, it is hard to find any
dominant studies of impact analysis. Although Dietzenbacher (1997) showed that the
supply-driven model can be interpreted as an ‘absolute price’ model and that the
interpretation is easier to understand its price effects than the ‘Leontief’ (relative) price
model, empirical applications still seem to be limited.
1
In this sense, selecting and comparing

1
Recently, Cai et al. (2005) reported a case study of economic impacts about regulation of longline
fishing using Goshian supply-side model, but it cannot still avoid Oosterhaven’s criticism in the sense
they depend on the hypothetical sector extraction approach neglecting the impacts on the targeted
sector(s), remarking the result of the Goshian model should be understood as the potential impacts of
the regulation, instead of being ignored.
26

two dominant modes for impact analysis will guide which criteria can be applied to the
supply-driven model.
First, Giarratani (1976) applied the supply-driven national IO model to trace national
economic impacts by restricting each energy sector (coal mining, crude petroleum and
natural gas, and energy activity sector) based on the 1967 U.S. interindustry table. That
study stressed two key criteria, monopolistic markets and scarce resources, the same as
suggested by Ghosh (1958) when choosing study definitions. It is reasonable to accept that
energy sectors are in some sense ‘monopolistic’ because it is not easy to find easy
alternatives in the short-run (Giarratani, 1976: 449~450). As Ghosh (1958) and Chen and
Rose (1991) noted, the balanced equilibria of industries in the long-run period even in
competitive markets remain stable by rationing without substitutions during the short-run,
because producers will depend on their previous sales even in the case that sudden
disruptions do occur. However, scarcities among other sectors except the three-targeted
sectors were not clarified in his application. Further, when simulating other impact analyses
in the study, two row vectors of coefficient sectors were extracted and multiplied by the total
inputs, in order to add the amounts to the remaining value added sectors as exogenous values.
To rerun the model with the increased value added, the supply-driven IO model was
reconstructed without the two sectors. This approach would avoid the problem of ‘output
changes without input changes’ criticized by Oosterhaven (1988) later. Still, this effort
disregarded the interconnections related to the two deleted sectors (ibid: 221).
Another important analysis was conducted by Davis and Salkin (1984), who applied the
supply-driven model to Kern county in California so as to estimate the economic impacts
27

from a hypothetical limitation of water supply to the agricultural sectors of the county. They
clarified two points;

i. Water supply in this area for agricultural industry is subject to the agency’s
distribution policy, denoting monopoly.
ii. No alternate sources of the water would be supported from the small area and
hence economic behaviors of producers would be preserved in the local area, not
adapted efficiently to minimize the costs by choosing alternatives.

The second point attached to using the supply-driven model has significant implications
especially when an economy depends very much on imports because it is hard for producers
to find alternatives in the economy itself, and hence the cited scarcity condition might be
significant at least in the short run.
In the circumstances that our current economic system is not a centralized planned
economy, it is important to specify the relevant conditions applying to supply-driven models.
From Ghosh’s idea and previous studies, therefore, four major conditions applying to
supply-driven model can be summarized:

i. Monopoly.
ii. Scarcity (of other inputs except targeted sectors).
iii. Short period.
iv. Small region (depending targeted resources much upon other regions).

28

Those four criteria should be applied to a case study in order to examine the applicable
possibilities of the supply-driven model. The next section will highlight the debates on the
plausibility of the supply-driven model and the state of the art.

3. Debate on the Implausibility of Supply-driven IO Models
While some comments on the supply-driven model since Ghosh had been addressed,
serious criticisms of the implausibility of the supply-driven model were made by
Oosterhaven (1988). He convincingly questioned that based on given final demands, if
“local consumption or investment reacts perfectly to any changes in supply” for example,
“purchases are made, e.g., of cars without gas and factories without machines”, it does not
require any production function because the final demands and input factors might be
combined without any technological relationship. Using Taylor’s expansion, he concluded
that “both as a general description of the working of any economy and as a way to estimate
the effects of loosening or tightening the supply of one scarce resource, the supply-driven
model may not be used” and suggested an alternative model instead of using the supply-
driven model directly.
In the following year, two studies by Gruver (1989) and Rose and Allison (1989) added
comments to the Oosterhaven’s critique. Gruver (1989) suggested an alternative
interpretation that the implicit production function, in spite of its perfect substitutability, and
agreeing with Oosterhaven that the supply-driven IO model could be interpreted as a cost-
minimizing choice to produce constant-returns-to-scale outputs, under the assumption of
constant relative prices. Similarly, Rose and Allison (1989) argued that the supply model
could still be useful to approximate the impacts of supply-driven disasters, if the allocation
29

coefficients are tolerably fixed. However, both studies did not touch on the core debate on
the implausibility of the model and Oosterhaven had succeeded with his argument (1989,
1996) until Dietzenbacher’s interesting interpretation appeared.
The contribution of Dietzenbacher’s (1997) interpretation is that the supply-driven IO
model is equivalent to the Leontief price IO model and hence that the supply model is to be
interpreted as an (absolute) price model instead of a quantity model. Then, the implausibility
problem raised by Oosterhaven would vanish. This interpretation contains one condition,
that the changes in value added should be followed by the price changes for the value added
inputs, not quantity changes. Given that quantities are fixed in the value-added vectors, here
is a simple proof that follows from equation (7.) and the fact that
s s
X A X B
ˆ
)
ˆ
(
1 −
= .

T
d
X ∆ =
1 1
]
ˆ
)
ˆ
( [
− −
− ∆
s s
l
X A X I V (8.1)

=
1 1 1
]
ˆ
)
ˆ
(
ˆ
)
ˆ
[(
− − −
− ∆
s s s s
l
X A X X X V (8.2)

=
s s
l
X A I X V
ˆ
) ( )
ˆ
(
1 1 − −
− ∆ (8.3)

From equation (8.3), let price changes in the l
th
labor factor be
p
i
v δ , let the row vector
without price changes in other value added sectors be ) )
ˆ
( (
1 −
∆ = ∆
s
l
p
l
X V V , and price
changes of total outputs be P ∆ . Then,

30

1
)
ˆ
(
−
∆
s
T
d
X X =
1
) (
−
− ∆ A I V
p
l
(9.)

and hence,

P ∆ =
1
) (
−
− ∆ A I V
p
l
(10.)

Equation (10.) is exactly the same as the Leontief price model, which suggests the
supply-driven model is the ‘Ghoshian’ price model. Although this interpretation provides a
theoretical defense against the criticism of the supply-driven model, the interpretation places
limits on the empirical applications of impact analyses. This might be due to difficulties of
interpretation that direct and indirect impacts caused by a disruption on the supply-side are
presumably dollar quantity losses rather than price decreases. This is the reason why recent
studies focus on forward linkages (Dietzenbacher, 2002; Cai et al., 2006) or structural
changes within an economy (Wang, 1997; Bon and Yashiro, 1996; Bon, 2001) using the
supply-driven model. Therefore, some further explanations are necessary to apply the
supply-driven model to impact studies, beyond the interpretation of the ‘Goshian’ price
model.

4. Reinterpretation of Supply-driven IO Model
The key assumption in Dietzenbacher’s (1997) is the suggestion that there are only price
changes among the value added changes. Corresponding changes of total outputs,
consequently, result from price changes, not quantity changes. This interpretation
successfully met the attacks on the implausibility of supply-driven IO models. However,
31

accepting this interpretation limits our understanding of the results of impact analyses,
because quantity losses, e.g. labor losses or capacity losses of a facility from unexpected
disasters are general and cannot be applied to the ‘Ghoshian’ price model.
Unfortunately, it has not been suggested how to find applicable conditions of supply-
driven models in the current market environment. Even Oosterhaven (1988) took only
cautious steps when dealing with the two basic conditions, monopoly and scarcity, in his
first criticism. Contrary to Ghosh’s explanation of the welfare function, however, one
defender (Gruver, 1989) concluded that producers behave to minimize their costs, but
unfortunately his conclusion was undermined by Oosterhaven’s next response (1989:
460~461), which missed Ghosh’s point with respect to the welfare function. Since
Dietzenbacher (1997), any attempts to deal with these conditions have not been made.
However, the market mechanism, although often modeled as “perfect”, includes some
market power at given prices and quantities. Due to limited accesses to market information
between sellers and buyers or even between industries, that is, due to asymmetry of
information, various shades of market power are common. This is the reason why a long-run
solution maintains that the equilibrium is a result of best negotiations among numerous
efforts by the actors in normal economic environments and induces resource scarcity without
efficient distributions especially during the short-run due to capacity constraints (Pindyck
and Rubinfeld, 1998: 21). This leads to, as Ghosh (1958) mentioned, the result that
producers will not decrease their previous outputs or factors during the short run, even if
there are huge shocks to the economy. The effort for finding substitute products requires
many other unexpected costs, e.g. costs of searching for substitutes, additional transportation
costs, and so on, and hence unless those are expected for the long-run experience, these
32

reactions will not happen. Even in the market system, therefore, two conditions of monopoly
and scarcity of resources can be verified during the short-run. Of course, because smaller
regions are more dependent on other regions, they will loose some market power and be
subject to exogenous market power which is not easily changed. So, it can be said ‘the more
common the market power, shorter the period and smaller the region’.
Therefore, under normal equilibrium economic status and characteristics, four possible
quadrants according to changes amon the value added factors could be identified, as shown
in Figure I-1. Because economic factors are expressed as money, I assumed two changes:

Only price changes without changes in quantity.
Monetary quantity (=quantity x price) changes.

These changes could be increased or decreased. Among the quadrants, the upper quadrants
are only affected by the price changes under the assumption that quantities are fixed, while
the lower quadrants refer to monetary quantity changes. The right sides of the quadrants
indicate the increases of price or monetary quantity, while the left sides show the decreases.
Although Dietzenbacher’s (1997) suggestions with respect to the ‘Goshian’ price model still
play an important role for all kinds of spaces, because under normal economic equilibrium
conditions, quantities are more or less fixed, but prices change, it is not enough to
understand just some of the cases and hence requires some additional explanation in order to
be extended to the study of impact analyses. The detailed discussions on each quadrant
follow.
33

The upper-left side, the first quadrant, of decreased value added factors in price might be
experienced in a deflated economic state. The U.S. experienced two sustained deflations
associated with depressions during the 1890s and 1930s and a temporary deflation
experience was experienced for one year 1954~1955 (Samuelson and Nordhaus, 1995:
575~576). This space is rarely observed in reality and, therefore, it will be hard to find an
appropriate case to use the supply-driven IO model, although Dietzenbacher’s suggestion
can be applicable.

I-1 Figure I-1 Four possible cases according to the change of value added factors

The upper right quadrant shows normal economic equilibrium conditions, where only
price increases are generally observed for the value added factors, given that quantity is
fixed. Most economic systems experience this sort of inflation and it might be investigated
whether or not there are different structures in an economic system during time intervals
Only price change without change in quantity
Monetary quantity (value) change
Decreased Increased
Dietzenbacher’s
(D) space

Dietzenbacher’s
(D) space

Oosterhaven’s
(O) space

Ghosh’ s
(G) space
34

based on such a price-deflator or decomposition methods. Also, sudden increases of labor
price in one industry sector due to e.g. a labor strike without increasing the number of
laborers will lead increases of output prices in other industry sectors even if there are no
changes in quantities in all other value added factor. Therefore, this quadrant wholly matches
Dietzenbacher’s explanation and is labeled as Dietzenbacher’s (D) space.
However, the assumption of only price changes without quantity changes is idealized
compared to an actual economic world, and monetary quantity changes including price
changes in value added factors are common. Although Dietzenbacher noted the
mathematical relation between the Leontief price model and the supply-driven model, it is
more generally true that the only price increase in value added is not reflected wholly in total
output, because market equilibrium still works during the short term period and hence would
not sustain the price increase. Rather, the D-space is more useful to examine these cases e.g.
the change of economic structures or linkages of multipliers for long-run analysis than short-
run impact studies.
Therefore, it is more useful to deal with monetary changes among value added factors
because they are more realistic. The lower quadrants are the cases including both price and
quantity changes simultaneously. Under normal economic equilibrium conditions, the third
(lower right) quadrant indicates that monetary value added factors have increased. I labeled
this space as O-space, because this it is the space criticized by Oosterhaven due to its
implausibility, as discussed in section III. However, if monetary increases in value added
factors due to, e.g., an increase of number of laborers for one sector temporarily induces
prices in all other sectors to increase in the forward direction without increases of value
added factors in all other sectors, because all economic factors only recognize the ‘dollar’
35

valuess then Dietzenbacher’s interpretation might be still useful to defend against
Oosterhaven’s concerns.
However, the monetary increases in factors would be relevant with the demand-driven
model because in many cases increases of the value added factors might result from the
market signals required by final demands. That is, although the supply-driven model might
be still useful in O-space to verify the price increases in total outputs of all sectors due to the
increase of only one value added factor during the short-run, the fundamental changes from
normal market equilibrium should be noticed on both the supply and demand sides. For
example, an increased demand for cars will induce an increase of the number of labors and
thus value added and output increases of other linked sectors. These changes require the
movement of market equilibrium for each period. Therefore a new approach reflecting both
sides at the same time might be helpful to investigate the impacts. In that sense, an
alternative model by Oosterhaven (1988) combining supply- and demand-driven models is
understandable, but his implausibility suggestion on the O-space was met via
Dietzenbacher’s suggestion.
For the above three quadrants, the supply-driven model is surely applicable using the
Ghoshian price-model. However, for the final remaining case, the Ghoshian price-model is
not as easily applied as the other quadrants. According to the Ghoshian price-model, a
sudden decrease in the monetarily expressed value added factors will decrease the absolute
price of total output losses. This result is wholly opposite to actual experience, because a
sudden decease in valued added sectors induces decreases in total output for the sector via
the allocation interrelations, and hence the absolute price for the sector generally would
increase.
36

Therefore, the fourth lower left-side quadrant indicates the exceptional economic
situation of sudden monetary value added losses such caused by terrorist attacks or
unexpected natural disasters, requiring the four conditions for an impact analysis. While the
basic interpretation of the supply-driven model might be focused on the price interrelations,
under static market equilibrium, producers will not change their current technical rationing
during the short term, after a man-made or natural disaster, as discussed for the four
conditions. In other words, only if they verify changes of final demands by the factor losses
( Y ∆ ) are higher than the lower-bounds (
Y
L ) required by final users at least, or ≥ ∆Y
Y
L ,
not upper-bound (or maximum) requests, will suppliers continue their sales until the market
loses its power from other pressures, even in the case that they lose their benefits during the
short-term. This examination of Ghosh’s supply-driven model shows a relation with
monetary quantity losses and is theoretically applicable to the man-made or natural impact
analyses under the four conditions. Therefore, this quadrant might be labeled as Ghosh’s (G)
space. The outstanding defense of Dietzenbacher, seeing the supply-driven model as a
Goshian (absolute) price model is very useful for the other spaces except for the G-space.
But taking into consideration that many impact analyses are conducted in this space, the G-
space should be differentiated.

5. Extension of the Supply-driven IO Model
The common limitation in IO models is their linear characteristic which can be expected
to over-estimate total impacts in a relatively long-term duration, because of the fixed
coefficients assumption. Or the market system might react relatively quickly, loosening one
or two conditions among the four conditions. It is common to note that market power is
37

relatively weak because of sizable regions or a relatively long-run period enabling
production substitutions and violating static market equilibrium will require more or less
adjusted equilibrium via markets. Empirically, however, we observe consumer behavior as
price changes according to quantity changes in the market in terms of price elasticity. Using
this price elasticity, even in the case of loosening the four conditions in the G-space, the
supply-driven IO model still might be useful. However, the supply-driven model using a
price elasticity is not the quantity-type supply-driven model, but price-type supply-driven
model, which is possibly converted from the quantity-type demand-driven model.

I-2 Figure I- 2. New Equilibrium via Market due to Economic Disturbance

1
S
0
S
0
D
Quantity
Price
0
Q
1
Q
0
P
1
P
38

As shown in Figure I-2, total input losses ) (
0 1
Q Q Q − = ∆ due to a disaster will
increase price by shifting the original supply curve
0
S to the left
1
S supply curve, following
the
0
D demand curve. Then, a new market equilibrium is decided at the new price
1
P ,
where consumers’ demands are reflected. From the changes of monetary quantity and price
on demand curve and an exogenous price elasticity of demand (
p
ε ) vector, new total input
losses are estimated as follows, reflecting consumers’ demands.
2

To use the exogenous price elasticity of demand,
p
ε , first, the quantity-type demand-
driven model shown in the equation (4.) should be converted to a price-type supply-driven
model as shown in equations (11.) and (12.).

T
s
X ∆ =
k
T
s
T
s
Y X B X I ∆ −
− − 1 1
] )
ˆ
(
ˆ
[ (11.1)
=
k
T
s
T
s
Y X B I X ∆ −
− − 1 1
)
ˆ
( ) (
ˆ
(11.2)
and hence,
T
s
T
s
X X ∆
−1
)
ˆ
( =
k
T
s
Y X B I ∆ −
− − 1 1
)
ˆ
( ) ( (12.1)
T
s
P ∆ =
k
Y
P B I ∆ −
−1
) ( (12.2)
Because the price elasticity of demand for sector i ,
i p,
ε , is defined as
i i
s
i
s
i
p p
x x
/
/
δ
δ
, based
on the
s
X Q = , the price change
i
p δ is obtained based on the exogenous price elasticity of
demand
i p,
ε as,

2
Price elasticities of demand for energy, for example, are available from
http://www.eia.doe.gov/smg/asa_meeting_2004/fall/files/exe/Elasticity%20Estimates.htm
39

i
p δ =
i i p
s
i
s
i
p
x x
/
/
,
ε
δ
(13.1)
=
s
i i p
i
s
i
x
p x
,
ε
δ
(13.2)
=
i
s
i
x π δ (13.3)

where
i
π =
s
i i p
i
x
p
,
ε
is exogenous for sector i , because it is relatively easier to find
the fixed
i
p and
s
i
x right before the event than to find the exogenous price
elasticities of demand.

Here, the column vector of price changes due to the disaster,
d
t
P
1 =
∆ (= Π ∆
s
X
ˆ
, where
Π is a column vector of
i
π ), is changed to
k
Y
t
P
1 =
∆ as,

k
Y
t
P
1 =
∆ =
d
t
X
P R
d
1
ˆ
=
∆ where Y X R
d
X
d
1
)
ˆ
(
−
= (14.)

Therefore, based on the
k
Y
t
P
1 =
∆ and equation (12.2), the vector of derived total
(relative) price changes
T
s
t
P
~
1 =
∆ is obtained as,

T
s
t
P
~
1 =
∆ =
k
Y
t
P B I
1
1
) (
=
−
∆ − (15.)
40

Because the total input vector in the next period,
T
s
t
X
1 =
, is the sum of total input vector
of pre-disaster and the monetary quantity changes vector after-disaster, that is,
T
s
t
X
0 =
+
T
s
X ∆ , the total input changes in post-disaster
T
s
t
X
1 =
∆ can be obtained by
multiplying of total inputs and price changes in the next period as
T
s
t
T
s
t
P X
~
ˆ
1 1 = =
∆ .
Therefore, even in the case that markets are out of equilibrium due to quantity losses
caused by a disaster, the price-type supply-driven model can still be applied to G-space to
estimate the total input losses for consumers due to the increase of prices if there are
exogenous price elasticities of demands.

6. Conclusions
In the Leontief comparative static analysis, we pass from one equilibrium to another. It
is clear, however, that in reality the economy traverses a period of disequilibrium in between.
In fact, it is possible to comment on the temporary disequilibria in terms of the various
models that have been discussed along with exogenously provided information on selected
price elasticities of demand. Under normal economic equilibrium conditions, quantities are
more or less fixed, but prices change, and hence Dietzenbacher’s (1997) suggestion is useful.
However, abnormal economic cessations such as caused by natural disasters will
temporarily produce quantity losses and lead to further economic losses via interindustrial
and interregional relations. As Dietzenbacher has noted, the basic interpretation of the
supply-driven model is via price interrelations. However, in static market equilibrium,
producers will not change the current technical relationships that are based on historical sales
during the short run immediately after an exogenous event. This reflects the fact that
41

Ghosh’s supply-driven model is in terms of monetarily expressed quantities and hence is
applicable.
Although Ghosh’s allocation IO model is suggested based on limiting conditions, the
implausibility debates have not focused on these aspects. According to his first conditions
and the two dominant empirical applications of economic impacts by the hypothetical
inoperability of facilities, it is reasonable to assume four conditions when running the supply
IO model: Monopoly characteristics, scarcity of inputs, short period, and small region
depending much upon other regions. Based on these conditions, I have supplemented
Dietzenbacher’s interpretation that showed weaknesses in understanding economic impact
analyses by monetary quantity losses.
To suggest a new interpretation for the supply-driven IO model, I introduced four
quadrant spaces of economic situations based on ‘price vs. price-quantity’ and ‘increase-
decrease’ axes. The analysis of the model space shows that Dietzenbacher’s suggestion is
useful for three quadrants, although his focus is only on the first and second ‘price increase’
quadrant, or D-space. Also, Oosterhaven’s focus on the supply-driven model is in the third
‘price- (monetary) quantity increase’ quadrant, or O-space. Finally, Ghosh’s suggestion is
most useful for (equivalent to) the ‘price- (monetary) quantity decrease’ quadrant, or G-
space, in market systems, where the supply-driven model can be used if normal static
equilibrium continues. Furthermore, even in the case that normal equilibrium could not be
maintained, the G-space can still be useful because the price-type supply-driven model plays
an important role in estimating the economic impacts in the abnormal equilibrium status. To
address the switching processes, I introduced an exogenous price elasticity of demand and
42

combined it with the classic supply-driven model, adjusting monetary quantity impacts to the
price impacts.
Input-output models are attractive because they can be made operational and accessible
at low cost. I have tried to unscramble the various positions taken and have contrasted them
in Figure I-1. When the general price level moves up or down, the supply model highlights
the actual absolute price transmission linkages. This is the Dietzenbacher view.
Oosterhaven’s implausibility criticism is most applicable when a positive value added
change is presumed to increase all forward transactions. This leaves us with the lower-left
quadrant. Ghosh’s original position is most plausible in the downward direction: a
downward shift in value added inputs can put limits on the forward transactions.
The power and the usefulness of linear models are enhanced once the applicability of the
supply model is clarified.

43

CHAPTER I REFERENCES

Bon, R., 2001, Comparative Stability Analysis of Demand-side and Supply-side Input-
Output Models: Toward Index of Economic “Maturity”, in Lahr, M.L. and E.
Dietzenbacher eds., Input-Output Analysis: Frontiers and Extensions, NY, Palgrave.

Bon, R. and T. Yashiro, 1995, Comparative Stability Analysis of Demand-side and Supply-
side Input-Output Models: The Case of Japan, 1960-90, Applied Economics Letters, 3:
349-354.

Cai, J., P. Leung, and J. Mak, 2006, Tourism’s Forward and Backward Linkages, Journal of
Travel Research, 45 (1): 36-52.

Cai, J., P. Leung, M. Pan, and S. Pooley, 2005, Linkage of Fisheries Sectors to Hawaii's
Economy and Economic Impacts of Longline Fishing Regulations, University of Hawaii
- NOAA, Joint Institute for Marine and Atmospheric Research, Honolulu, Hawaii.

Chen, C.Y. and A. Rose, 1991, The Absolute and Relative Joint Stability of Input-Output
Production and Allocation Coefficients, in Peterson, W., ed., Advances in Input-Output
Analysis. NY, Oxford University Press.

Davis, H.C. and E.L. Salkin, 1984, Alternative Approaches to the Estimation of Economic
Impacts Resulting from Supply Constraints, The Annals of Regional Science, 18: 25-34.

Dietzenbacher, E., 1997, In Vindication of the Ghosh Model: A Reinterpretation as a Price
Model, Journal of Regional Science, 37: 629-651.

Dietzenbacher, E., 2002, Interregional Multipliers: Looking Backward, Looking Forward,
Regional Studies, 36 (2):125-136.

Ghosh, A., 1958, Input-Output Approach to an Allocative System, Economica, XXV: 58-64.

Ghosh, A., 1964, Experiments with Input-Output Models: An Application to the Economy of
the United Kingdom, 1948-55, London, Cambridge University Press.

Giarratani, F., 1976, Application of an Interindustry Supply Model to Energy Issues,
Environmental Planning A, 8: 447-454.

Gruver, G.W., 1989, On the Plausibility of the Supply-driven Input-Output Model: A
Theoretical Basis for Input-coefficient Change, Journal of Regional Science, 29, 441-
450.

Hewings, G.J.D., 1985, Regional Input-Output Analysis, Beverly Hills, CA: Sage
Publications, Inc.

44

Leontief, W., 1936, Quantitative Input and Output Relations in the Economic System of the
United States, Review of Economic Statistics, XVIII (3): 105-125.

Leontief, W., 1941, The Structure of American Economy, 1919-1929: An Empirical of
Equilibrium Analysis, Cambridge, MA, Harvard University Press.

Leontief, W., 1976, The Structure of American Economy, 1919-1939: An Empirical of
Equilibrium Analysis, White Plains, NY, International Arts and Science Press, Inc.

Oosterhaven, J., 1988, On the Plausibility of the Supply-driven Input-Output Model, Journal
of Regional Science, 28: 203-217.

Oosterhaven, J., 1989, The Supply-driven Input-Output Model: A New Interpretation but
Still Implausible, Journal of Regional Science, 29: 459-465.

Oosterhaven, J., 1996, Leontief versus Ghoshian Price and Quantity Models, Southern
Economic Journal, Vol. 62, No.3: 750-759.

Pindyck, R.S. and D.L. Rubinfeld, 1998, Microeconomics, NJ. Prentice-Hall, Inc.

Rose, A. and T. Allison, 1989, On the Plausibility of the Supply-driven Input-Output Model:
Empirical Evidence on Joint Stability, Journal of Regional Science, 29: 451-458.

Samuelson, P.A. and W.D. Nordhaus, 1995, Economics, McGraw-Hill, Inc.

Wang, E.C., 1997, Structural Change and Industrial Policy in Taiwan, 1966-91: An
Extended Input-Output Analysis, Asian Economic Journal, 11 (2): 187-206.

45

III. CHAPTER II: ESSAY TWO

A Two-Step Approach to Estimating Detailed State-to-State Commodity Trade Flows

1. Introduction
Many economic disruptions from natural or man-made disasters have led economists to
evaluate the socioeconomic impacts on the U.S. economy, especially since the 2001 terrorist
attacks in New York. This is because even a small disaster that does not involve many
fatalities could cause enormous economic losses. The most widely used impact models are
input-output (IO) models. They offer considerable sectoral detail along with computability.
National IO models aggregate over large numbers of diverse regions. However,
economic impacts have an inevitable spatial incidence. Consquently, the disasters under
discussion should be studied by tracing effects on commodity flows between regions and
industries. This suggests adding spatial detail to traditional IO. First, political representatives
have an obvious interest in their own constituency and jurisdiction. Second, subnational
impacts can cancel each other in the aggregate, causing national measures to obscure key
dimensions of the event. So far, the U.S. multi-regional input-output models (MRIO)
present an attempt at regional disaggregation, but these MRIO models are difficult to
construct because consistent sub-national trade data are hard to develop.
Spatial connections between states must be considered to examine the full-costs
throughout the U.S. economy using an integrated model of losses. A major problem in
developing integrated interregional-interterindustry models is how to combine otherwise
46

incompatible databases. Although Chenery (1953) and Moses (1955) formulated an
interregional framework based on the early discussion of Isard (1951), data problems still
stymie most applications. This explains why an operational Chenery-Moses model has not
been available, aside from the work of Polenske (1980) and Jack Faucett Associates (1983).
Also, U.S. Commodity Transportation Survey reports on inter-regional trade flows since
1977 have been discontinued. This data deficit can be met to some extent with the
Commodity Flow Survey (CFS) from the Bureau of Transportation Statistics (BTS), which
is published every five years. However, the currently available CFS data are incomplete
with respect to interstate flows.
The primary purpose of this study is to suggest a useful way to create trade flows
between U.S. states as the basis for a new National Interstate Economic Model (NIEMO) for
the U.S. Direct economic impacts are relatively easily estimated in the aftermath of an
attack. If plausible scenarios for the time-profile of reduced shipping facilities are available,
spatially detailed indirect and induced economic effects can be estimated with a NIEMO-
type model. Standard applications of IO determine indirect and induced impacts that
typically do not include interactions among industries and states. Estimating such short-term
impacts requires multi-regional models consisting of two sets of tables, regional coefficient
tables and trade coefficient tables (Miller and Blair, 1985). These NIEMO-type Chenery-
Moses models can be used to estimate inter-state industry effects as well as inter-industry
impacts on each state. To proceed this way, it is necessary to calculate multi-regional inter-
industry coefficients among U.S. states, based on, i) regional tables that give intra-regional
industry coefficients by state, and ii) interregional trade tables that give trade coefficients by
47

industry. This essay suggests and applies a sequence of computational steps for estimating
inter-state trade flows required implementing such a model.
To construct the trade tables - between all 50 states plus D.C. and the rest of the world -,
I applied an Adjusted Flow Model (AFM), a Doubly-constrained Fratar Model (DFM) based
on 1997 CFS and 2001 IMPLAN data. Due to the different industrial code systems that
characterize the two data sources, however, reconciliation of the IMPLAN and CFS
databases presented several problems. Below, I describe the “USC (reconciled) Sectors”
which were developed to enable a matching of two code systems used in the North American
Industry Classification System (NAICS), the Bureau of Economic Analysis (BEA), or the
new 509 IMPLAN industry codes. The third section of this essay explains the methodologies
used to improve the available data. I applied an AFM to estimate missing entries in the 1997
CFS data and a Doubly-onstrained Fratar Model (DFM) to update the 1997 CFS flows to
2001 estimates using 2001 IMPLAN data. This permits us to construct tables for trade
between all 50 states plus the District of Columbus, the rest of the world. The estimated
results using the methodologies are shown in the fourth section.

2. Background
As Lahr (1993) noted, a major problem in developing an MRIO-type model stems from
the fact that it is difficult to obtain data representing U.S. trade flows between the states, and
difficult to reconcile what data are available. The U.S. Commodity Transportation Survey
Data on interregional trade flows had been available since 1977, but the reporting was
discontinued for some years. For the years since 1993, this data deficit can be met to some
extent with the recent Commodity Flow Survey (CFS) data from the Bureau of
48

Transportation Statistics (BTS). While the CFS data have been widely used, the data have
several inherent problems (Erlbaum and Holguin-Veras, 2005: 3). The most serious problem
is that the CFS data do not include trade flows below the state level. Also they are not
complete trade flows even at the state level. For these reasons, there has been no
comprehensive inventory of flows, despite the foundation offered by Polensky’s (1980) and
Faucett Associates’ (1983) trial models.
The presence of many unreported values in the CFS data requires relying on other data
sources to improve completeness. Harrigan et al. (1981) compared old methodologies to
estimate interregional trade flows; and demonstrated “more information, better results,”
based on 1973 data for Scotland. All of their techniques are simple ratio-based
methodologies. Lie and Vilain’s (2004) used location quotients with CFS data in a similar
trial to estimate trade inflows below the state-level. This approach requires very restrictive
assumptions, resulting in sizable errors in the estimates. Canning and Wang (2005) used an
approach suggested by Wilson and Batten 1982 to suggest a new method for estimating
interregional trade flows for MIRO and models and provide an empirical test of performance.
Most recently, there have been two attempts using data from the 1997 Commodity Flow
Survey (CFS) to estimate interregional trade flows for a MRIO. The two recent studies used
the CFS and IMPLAN data as their basic data set. Jackson et al. (2006) used IMPLAN data
to adjust incomplete CFS information primarily by adopting a Box-Cox transformation and
double-log distance-decay functions. The second attempt relies on a doubly-constrained
gravity model based on county-level data from IMPLAN and ton-mile data from CFS
(Lindall et al, 2005). Details are given below.

49

2.1. The Jackson et al (2006) Approach
Jackson et al (2006) suggest the following formula to obtain parameter estimates γ and
θ , which minimize the percentage error between the trade flows predicted using CFS data
(
mn
i
Y
ˆ
) and buffer (b )-minimized regressed trade flows (
mn
i
T ) based on IMPLAN data that
are adjusted to SCTG (Standard Classification of Transported Goods) codes.

2
,
ˆ
∑∑
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
−
mn
mn
i
mn
i
mn
i
T
T Y
Min
θ γ
(1.)
where
mn
i
Y
ˆ
=
m
i
n
mn
i
n
i i
mn
i
n
i i
y
d f p g
d f p g
∑
) ( ) (
) ( ) (
, (2.)
i
x x g
i
γ
= ) ( , (3.)
x
i
i
e x f
θ −
= ) ( , (4.)
∑
≠
=
m n
mn
i
m
i
y y , and (5.)
mn
i
T = τ − Γ
∑
m
mn
i
b
Min , (6.)

The
∑
Γ
m
mn
i
is the regression-based total import demand for region n , sector i ; and τ is the
corresponding IMPLAN import demand. Commodity flows
mn
i
Γ from region m to n with
regression-based estimates
0
ˆ
β and
1
ˆ
β are given as
50

mn
i
Γ =
r
i
b C d L b C d L
X e
m mn m mn
*
)} (
ˆ
) (
ˆ
{ − − − + −
, (7.)
where
) ln(
ˆ ˆ
) (
ˆ
1 0
x x L β β + = (8.)

and
r
i
X are domestic exports of sector i obtained from IMPLAN data. Note that the
estimates
0
ˆ
β and
1
ˆ
β are obtained from a Box-Cox transformation regression

i
i
i
i
i i
d h
λ
β β
λ
λ λ
~
1 0
+ = , (9.)

or a double-log regression

)
~
ln( ) ln(
1 0 i i
d h β β + =

(10.)

in the case unfitted by Box-Cox transformation for sector i , where
i
d
~
is the distance range
and
i
λ is the Box-Cox transformation parameter.
The estimated parameters γ and θ are used in the equation (2.), and hence the
updated values for
mn
i
Y
ˆ
are obtained via the minimization of errors in equation (1.).
However, because the CFS includes no trade data for service sectors, Jackson et al. used the
average values of
0
ˆ
β and
1
ˆ
β across all other services and then estimated γ , θ , and
mn
i
Y
ˆ
for
the service sectors.
51

According to the CFS definitions, foreign imports that are transported from their port
of entry to the destination state are included in the CFS inter-state commodity flows.
However, in the IMPLAN data, foreign imports refer to the imports that are consumed in the
local area. The foreign imports that are not consumed in the local area, and are instead
transported to other state(s), are excluded from the state or county-level IMPLAN data (Park
et al, 2006a; Giuliano et al, 2006). Therefore, it is necessary to reconcile the two data sets.

2.2. Lee’s Approach (1973)
Doubly-constrained gravity models reflect the interactive effects of trades in addition
to allocations of exports to regions. The model is premised on the assumption that the
attractiveness of a regional economy is proportional to the trade flows between regions,
inversely proportional to distances between regions. Given gross domestic supply and
demand for which the sum over all regions is same,
mn
i
T , the trade flow for sector i
between region m and n , is as follows (Lee, 1973: 82~87).

ω −
=
mn
n
i
m
i
n
i
m
i
mn
i
d D O B A T , (11.)
where
m
i
A =
1
) (
− −
∑
n
mn
n
i
n
i
d D B
ω
, (12.)
n
i
B =
1
) (
− −
∑
m
mn
m
i
m
i
d O A
ω
, (13.)
m
i
O =
∑
n
mn
i
T = total supply originating in region m of sector i , (14.)
n
i
D =
∑
n
mn
i
T = total demand consumed in region n of sector i , and (15.)
52

ω −
mn
d = a distance decay function with parameter ω . (16.)

Lindall et al (2005) used data from three sources to estimate this model, Oak Ridge
National Laboratory (ORNL) for county-to-county distances by mode of transportation, the
CFS for ton-miles by sector, and IMPLAN for total supply and demand by county. The
ORNL data support an “impedance index as a combination of distance, time, and cost fact”
(Lindall et al, 2005: 4) by reference to employment centroids. The CFS data are used for a
criterion index to determine whether the average ton-miles estimated from the Lee’s
approach matches the CFS ton-miles values.
However, there are at least three problems associated with using the CFS data. First,
it is still not clear whether Lindall et al (2005) constructed a data bridge providing
correspondence between the CFS and IMPLAN data. Second, although the IMPLAN data
are expressed in dollar values, not tons, Lindall, et al. did not explain how to convert
IMPLAN dollar values to ton values for comparison with CFS averages. And finally, the
IMPLAN data are not shipments. They are transaction values, so the data must be adjusted
according to the CFS definitions.
Further, these authors set the value of the parameter ω at 2 for non-service sectors.
They set the corresponding values for service sectors at high (unreported) magnitudes to
induce low volumes in these sectors’ trades. Because there is no ton-mile data for service
sectors in the CFS, how Lindall et al (2006) estimated trade flows for service sectors remains
unclear. Despite their many caveats, their results may be the first substantive attempt to
estimate national trade flows at the county level. They reported that 36 CPU hours were
53

required to estimate a 3,140 x 3,140 county-by-county matrix for 509 IMPLAN sectors,
excluding service sectors.
None of these attempts provides clear discussions of how to estimate trade flows based
on a careful reconciliation between IMPLAN’s sectors and the SCTG sectors of the CFS.
Also, it might be an oversimplification to assume low levels of trade between service sectors,
thus distorting these results.

3. Data Reconciliation
Basic data for this study are obtained from the 1997 CFS and the 2001 version of
IMPLAN. The 1997 CFS reports trade flows between U.S. states, although the flow data are
not complete because of ‘high sampling variability’ or ‘disclosure constraints’ with respect to
data associated with individual companies. Yet, these data can provide a useful base-line for
updating the CFS data to 2001 based on the 2001 IMPLAN data. However, the industrial
classification systems for these different data sources are incompatible. Estimating 2001
trade flows from the 1997 CFS requires various intermediate conversion steps between the
SCTG code system used in the 1997 CFS and the code system for the IMPLAN sectors. This
is because there is not always a one-to-one match between elements of the BEA and NAICS
codes. This approach executes the data reconciliation process suggested by Park et al
(2006a) to create a SCTG-IMPLAN conversion bridge enabling aggregation of the 509
IMPLAN sectors into 43 SCTG sectors. The various matches are shown in Appendix A1.
Another reconciliation task between IMPLAN and CFS data concerns basic concepts. In
other words, the concepts in these two data sources should refer to the same thing. For
example, based on CFS definitions, foreign imports transported from a port of entry to the
54

destination state are included in CFS inter-state commodity flow. However, in the IMPLAN
data, foreign imports refer to imports that are consumed in the local area. Those foreign
imports that are not consumed in the local area and transported to other state(s) are excluded
from the state or county-level IMPLAN data.
In order to make the concept of “inter-state commodity flow” consistent within these two
data sources, foreign imports in IMPLAN data (
f
i
IM
I ) are adjusted and defined as adjusted
foreign imports (
f
i
a
I ) by dividing by the 29 new ratios. The ratios are calculated as
proportions of foreign imports of sum of every states over U.S. foreign import by 29 USC
commodity sector. The
f
i
a
I consists of the foreign imports consumed in other states (
f
i
c
I )
and the foreign imports consumed in the local area that is used as a port of entry (
f
i
IM
I ). In
this way, CFS and IMPLAN data could be reconciled conceptually.
With this reconciliation, some minor manual adjustments are still required on the basis of
judgment, using sector names of 5-digit SCTG and 6-digit NAICS. This adjusts default
equal-proportions assumptions arising from aggregation in the case of ‘one- or multi-sectors
to multi-sectors’. Also, a producer/purchaser dollar value adjustment was conducted because
the IMPLAN data include producer values, while CFS data are based on purchaser values
which include transportation cost, wholesale markup, and retail markup besides the producer
values.

55

Table II-1 . IMPLAN Reconciliation with 1997/2002 CFS Producer Prices, by USC Sector
Sectors
2001
IMPLAN
2002
CFS_Revised
1997
CFS_Revised
2002 Ratio
1997
Ratio
USC V1* P1** V4 P4 V5 P5 V1/V4 P1/P4 V1/V5 P1/P5
USC01 192,478 3.18% 171,981 2.92% 153,997 3.03% 1.12 1.09 1.25 1.05
USC02 130,536 2.16% 131,504 2.24% 115,470 2.27% 0.99 0.97 1.13 0.95
USC03 45,911 0.76% 41,433 0.70% 50,130 0.99% 1.11 1.08 0.92 0.77
USC04 86,329 1.43% 86,226 1.47% 79,122 1.56% 1.00 0.97 1.09 0.92
USC05 302,706 5.01% 263,970 4.49% 252,361 4.96% 1.15 1.11 1.20 1.01
USC06 80,602 1.33% 76,558 1.30% 58,148 1.14% 1.05 1.02 1.39 1.17
USC07 54,172 0.90% 49,519 0.84% 36,191 0.71% 1.09 1.06 1.50 1.26
USC08 20,141 0.33% 19,396 0.33% 17,936 0.35% 1.04 1.01 1.12 0.94
USC09 11,054 0.18% 14,729 0.25% 11,794 0.23% 0.75 0.73 0.94 0.79
USC10 480,173 7.94% 270,347 4.60% 253,304 4.98% 1.78 1.73 1.90 1.59
USC11 104,099 1.72% 120,479 2.05% 126,464 2.49% 0.86 0.84 0.82 0.69
USC12 174,086 2.88% 300,630 5.11% 158,114 3.11% 0.58 0.56 1.10 0.93
USC13 22,231 0.37% 29,431 0.50% 23,606 0.46% 0.76 0.73 0.94 0.79
USC14 159,819 2.64% 172,452 2.93% 154,153 3.03% 0.93 0.90 1.04 0.87
USC15 231,896 3.83% 248,130 4.22% 201,484 3.96% 0.93 0.91 1.15 0.97
USC16 122,282 2.02% 115,614 1.97% 113,525 2.23% 1.06 1.03 1.08 0.91
USC17 154,827 2.56% 160,021 2.72% 158,010 3.11% 0.97 0.94 0.98 0.82
USC18 133,501 2.21% 106,600 1.81% 202,729 3.99% 1.25 1.22 0.66 0.55
USC19 292,878 4.84% 316,653 5.39% 236,813 4.66% 0.92 0.90 1.24 1.04
USC20 113,064 1.87% 114,330 1.94% 87,240 1.72% 0.99 0.96 1.30 1.09
USC21 169,411 2.80% 213,769 3.64% 240,745 4.73% 0.79 0.77 0.70 0.59
USC22 200,391 3.31% 199,880 3.40% 193,294 3.80% 1.00 0.97 1.04 0.87
USC23 433,014 7.16% 424,514 7.22% 347,545 6.83% 1.02 0.99 1.25 1.05
USC24 844,544 13.96% 799,929 13.60% 733,800 14.43% 1.06 1.03 1.15 0.97
USC25 654,570 10.82% 620,959 10.56% 481,910 9.48% 1.05 1.02 1.36 1.14
USC26 143,113 2.37% 157,354 2.68% 124,723 2.45% 0.91 0.88 1.15 0.96
USC27 160,050 2.65% 166,576 2.83% 118,491 2.33% 0.96 0.93 1.35 1.14
USC28 92,277 1.53% 82,582 1.40% 59,471 1.17% 1.12 1.09 1.55 1.30
USC29 436,417 7.22% 404,687 6.88% 295,358 5.81% 1.08 1.05 1.48 1.24
ALL 6,047,838 100% 5,880,253 100% 5,085,927 100% 1.03 1.00 1.19 1.00
* Units: million$
** {(Each SCTG sector value)x100}/ (ALL value).

56

4. Model
Based on the data bridges to reconcile different data code systems, a two-step approach
via an Adjusted Flow Model (AFM) and a Doubly-constrained Fratar Model (DFM) was
developed. Estimated 2001 commodity trade flows among all 50 states plus Washington,
D.C. and the rest of the world were developed from the original 1997 CFS for 29 USC
Commodity Sectors. The first step in order to use the DFM is to complete the 1997 CFS
unreported values for a variety of commodities. This is the AFM approach and includes
some marginal values such as total shipments originating in each state, total shipments
destined for each state, and the matrix of cells representing commodity trade flows between
pairs of states. The completion is conducted based on the 2001 IMPLAN data, which report
total origin and destination values by state. Then, the 2001 commodity trade flows could be
estimated with the DFM. First, the procedures for missing value estimation are followed as
described in Section IV-1.

4.1. Two-Step Approach: Adjusted Flow Model (AFM)
To calculate the values in each unreported cell of the trade flows between states, first,
total origin and total destination values should be fixed. Let reported total origin and
destination from CFS be
T
i
O and
T
j
D respectively. To calculate unreported total origin
(output) value of state i (=
UT
i
O ), the ratio of 2001 IMPLAN total origin of state i (=
T
i
IM
O )
to the sum and 1997 CFS reported total value of each USC sector m
(=
∑ ∑
+ = + =
j
UT
j
T
j
m
i
UT
i
T
i
m
T
m
D D O O V ) ( ) ( , 29 ,..., 1 = m ) was used as shown in
equation (17.), based on a specific USC sector m .
57

UT
i
O =
T
i
T
i
IM
T
i
IM
V
O
O
×
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
∑
, (17.)

where,
i
IM
S is total supply commodity,

d
i
IM
N is net domestic products (= j i N
d
j
IM
= , ),

d
i
IM
E is domestic exports, and

f
i
IM
E is foreign exports.
All these are supported by IMPLAN data. Also,

T
i
IM
O =
i
IM
S +
f
i
a
I . (18.)

Because
i
IM
S =
d
i
IM
N +
d
i
IM
E +
f
i
IM
E , (19.)

then the
T
i
IM
O shown in equation (18) would be rewritten as,

T
i
IM
O =
d
i
IM
N +
d
i
IM
E +
f
i
IM
E +
f
i
a
I . (20.)

This is because foreign imports should be counted in the trade flows in U.S. domestic trade,
or
f
i
IM
I (IMPLAN foreign imports in state i , that remains in each state) plus
f
i
c
I (foreign
58

imports for state i consumed in other states), once commodities are imported. Hence,
adjusted foreign imports are shown as,

f
i
a
I =
f
i
IM
I +
f
i
c
I , (21.)

although IMPLAN data only count
f
i
IM
I . Also, foreign imports are more compactly related
to regional economic conditions than foreign exports (
f
i
IM
E ). Therefore,
f
i
IM
E assumes no
trade to other states because i) they cannot separate which amount directly goes to the rest of
world from each state and which amount goes outbound and then to the rest of world, and ii)
economically those are only related to the transportation services sector once they are
produced.
Then, IMPLAN total destination
T
j
IM
D of state j can be calculated as,

T
j
IM
D =
d
j
IM
N +
d
j
IM
I +
f
j
a
I , (22.)

where
d
j
IM
N is the sum of net domestic products of state j ,

d
j
IM
I is domestic imports obtained from IMPLAN data of state j , and

f
j
a
I is adjusted foreign imports of state j .

Then, for a specific USC sector m , unreported total destination (input) values of state j ,
UT
j
D , were calibrated as,
59

UT
j
D =
T
j
T
j
IM
T
j
IM
V
D
D
×
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
∑
(23.)

From the estimated and original total origin/destination values, unreported trade flow
values between states i and j (
U
ij
V ) can be filled in the matrix. In this computation, the
cross-effects of origin and destination values are considered to estimate any unreported cell
values. For instance, the unreported destination (=
U
ij
D ) value can be calculated from total
unreported residuals

U
ij
p
R =
T U
j
D
) (
-
∑ p p
j ij
V , (24.)

by multiplying the portions of total origin corresponding to unpublished cells sector
U
ij
V as
shown in equation (25.1). Similarly, unreported origin cells (=
U
ij
O ) are computed as in
equation (25.2). However, because two matrices (
U
ij
D and
U
ij
O ) are adjusted only based on
total origin or total destination from the two equations of (25.1) and (25.2), I took the mean
value of the two in equations (25.3) This allows the one side based estimates to yield the
adjusted values of each cell.

60

U
ij
D =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
×
∑
k
T U
k
T U
k U
ij
O
O
R
p
) (
) (
(25.1),

U
ij
O =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
×
∑
k
T U
k
T U
k U
j i
D
D
R
p
) (
) (
(25.2),
U
ij
V

=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛ +
2
U
ij
U
ij
O D
(25.3).

where subscripts
p
i and
p
j indicate only published cells, and hence
p
ij
V or
j i
p
V mean only
reported or 0 values of each cell in the given trade matrix. Also, subscript k of
T U
k
O
) (
or
T U
k
D
) (
indicates the corresponding cells in
T U
k
O
) (
or
T U
k
D
) (
to the unreported cells
U
ij
V in the given matrix, irrespective of the estimated
UT
k
O (or
UT
k
D ) or the known
T
k
O (or
T
k
D ).

To obtain the optimal
U
ij t
V , equations (25.) should be iterated as shown in (26.).

U
ij t
D =
⎟
⎟
⎟
⎠
⎞
⎜
⎜
⎜
⎝
⎛
×
∑ −
−
−
k
T U
k t
T U
k t U
ij
t
O
O
R
p
) (
1
) (
1
1
(26.1),

61

U
ij t
O =
⎟
⎟
⎟
⎠
⎞
⎜
⎜
⎜
⎝
⎛
×
∑ −
−
−
k
T U
k t
T U
k t U
j i
t
D
D
R
p
) (
1
) (
1
1
(26.2),

U
ij t
V

=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛ +
2
U
ij t
U
ij t
O D
(26.3).

Then, the optimal value
U
ij t
V in the
th
t iteration was chosen as the maximum value
(=
U
ij
MV ) in equation (27.) to satisfy the following criteria:

U
ij
MV =
∑∑
ji
U
ij t
V MAX (27.)

subject to 1)
T
m
ij
U
ij t
ji
U
ij t
V V V ≤
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
=
∑∑ ∑∑
, or
2) 999 . 0
1 1
<
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
=
∑∑ ∑∑ − −
ji
U
ij t
ji
U
ij t
V V

Note that the optimal value
U
ij m
MV for USC Sector m from this model is the closest
value to
T
m
V , but only considers destination attractions and origin supply power without
distance decay effect.

62

4.2. Two-Step Approach: Doubly-constrained Fratar Model
Fratar models are useful for estimating updated commodity trade flows, where the
starting matrices include numerous estimated values for missing entries in the CFS data.
However, the traditional Fratar model calibrates only off-diagonal interregional cells. In this
application, new diagonal values accounting for intrastate trade flows had also to be
estimated. A Doubly-constrained Fratar model (DFM) that updates the diagonal values in the
CFS matrix combined with the traditional Fratar model to estimate the off-diagonal values
was developed for this estimation. The DFM iteratively estimates all the updated CFS values
simultaneously and consistently. The estimated values for each USC sector from the AFM
are the base values for the next iterative step of the DFM.
Define
T
i
O
ˆ
and
T
j
D
ˆ
as the observed or estimated values of
T U
i
O
) (
, the total origin
(output) value for state i , and
T U
j
D
) (
, the total destination (input) values for state j ,
respectively. These estimates are provided by the procedure used to estimate missing values
in the 1997 CFS data with the AFM. Also, define net destination
i
IM
ND (=
d
i
IM
N +
f
i
IM
I =
i
IM
NO ) as the diagonal entries in a matrix consisting of
IMPLAN’s net domestic products (
d
i
IM
N ) plus the remaining IMPLAN foreign imports for
each state i (
f
i
IM
I ).
Then, IMPLAN net total originating values (outputs) in statei ,
T
i
IM
NO , are defined in
equation (28.)

T
i
IM
NO =
T
i
IM
O –
f
i
IM
E (28.1)
63

= (
d
i
IM
N +
d
i
IM
E +
f
i
IM
E +
f
i
a
I ) –
f
i
IM
E (28.2)

= (
d
i
IM
N +
d
i
IM
E +
f
i
IM
E +
f
i
IM
I +
f
i
c
I ) –
f
i
IM
E (28.3)

=
d
i
IM
N +
d
i
IM
E +
f
i
IM
I +
f
i
c
I (28.4)

=
i
IM
NO +
d
i
IM
E +
f
i
a
I (28.5)

Similarly, IMPLAN’s net total destined values (inputs) for state j ,
T
j
IM
ND , is defined as,

T
j
IM
ND =
T
i
IM
D –
f
i
c
I (29.1)

= (
d
j
IM
N +
d
j
IM
I +
f
j
a
I ) –
f
i
c
I (29.2)

= (
d
j
IM
N +
d
j
IM
I +
f
i
IM
I +
f
i
c
I )–
f
i
c
I (29.3)

=
d
j
IM
N +
d
j
IM
I +
f
i
IM
I (29.4)

64

Then, by excluding the corresponding diagonal outputs
d
i
IM
N and
d
j
IM
N respectively,
net values (
T
i
NO and
T
j
ND ) of
T
i
IM
NO and
T
j
IM
ND can be obtained as,

T
i
NO =
T
i
IM
NO –
d
i
IM
N (30.1)

=
d
i
IM
E +
f
i
c
I , (30.2)

and

T
j
ND =
T
j
IM
ND –
d
j
IM
N (31.1)

=
d
j
IM
I (31.2)

Therefore, by excluding corresponding diagonal outputs
d
i
IM
N and
d
j
IM
N respectively,
net values (
T
i NO
∧
and
T
i ND
∧
) of
T
i
O
ˆ
and
T
j
D
ˆ
can be obtained as,

T
i NO
∧
=
T
i
O
ˆ
–
d
i
IM
N , (32.)

and

65

T
j ND
∧
=
T
j
D
ˆ
–
d
j
IM
N . (33.)

The growth factors for origin states i and destination states j ,
i
G and
j
G , are calculated
from equations (34.) and (35.),

T
i
T
i
i
NO
NO
G
∧
= , (34.)

and

T
j
T
j
j
ND
ND
G
∧
= . (35.)

These growth factors are substituted into equations (34.) and (35.) to obtain balance
factors
i
L and
j
L , which are used to update off-diagonal CFS entries iteratively. Let
ij
MV
be the observed and estimated cell values from the AFM and
ij
FV
1
be the starting values to
estimate the 2001 CFS off-diagonal flows from state i to state j .

i
L =
∑
×
∧
j
j ij
T
i
G MV
NO
) (
, (36.)
and
66

j
L =
∑
×
∧
i
i ij
T
j
G MV
ND
) (
(37.)

This is a standard application of the traditional Fratar model that relies on the calibrated
factors provided by equations (34.) to (37.).

ij
FV
1
=
⎭
⎬
⎫
⎩
⎨
⎧ +
× × ×
2
) (
j i
j i ij
L L
G G MV for all i ≠ j . (38.)

Equations (39.) to (40.) define
i
DG and
j
DG , diagonal entry growth factors for origin
state i to destination state j .

T
i
T
i
i
O
O
DG
∧
= , (39.)
and
T
j
T
j
j
D
D
DG
∧
= . (40.)

Similarly, equations (41.) and (42.) define
i
DL and
j
DL , the diagonal entry balance
factors used to update the diagonal (intrastate) entries of the CFS matrix iteratively.

67

i
DL =
∑
×
∧
j
j ij
T
i
DG MV
O
) (
, (41.)

and
j
DL =
∑
×
∧
i
i ij
T
j
DG MV
D
) (
. (42.)

Estimated Diagonal Values (
ij
DV
1
) are calculated via equation (43.), which defines a
second Fratar model estimating trade flows within each statei . These results also account for
new foreign imports remaining within each state.

ij
DV
1
=
⎭
⎬
⎫
⎩
⎨
⎧ +
× × ×
2
) (
j i
j i ij
DL DL
DG DG MV for all i = j . (43.)

These initial estimates of the updated diagonal values,
ii
DV
1
, the diagonal entry growth
factors,
i
DG and
j
DG , and the diagonal entry balance factors,
i
DL and
j
DL , are all
updated iteratively until they converge to consistent values across equations (39.) to (43.).

ij t
DV =
⎪
⎭
⎪
⎬
⎫
⎪
⎩
⎪
⎨
⎧ +
× × ×
− −
− − −
2
) (
1 1
1 1 1
j t i t
j t i t ij t
DL DL
DG DG DV for alli = j . (44.)

68

The
ii t
DV replaces
i
IM
NO if and only if
ii t
DV >
i
IM
NO . Therefore, the finally iterated
diagonal values at t times
ii t
DV replace the diagonal values
ii t
DV
1 −
in the CFS matrix if
and only if
ii t
DV >
ii t
DV
1 −
. Note the CFS totals for each state are reduced by the difference
between the corresponding values
ii t
DV and the original diagonal values
i
IM
NO , and hence,
the sum of off-diagonal flows (or residuals) for the corresponding state will be decreased.
The initial estimates of the updated off-diagonal CFS flows,
ii
FV
1
, the growth factors
for origin states i and destination states j ,
i
G and
j
G , and the balance factors,
i
L and
j
L
are all updated iteratively until they converge to consistent values across equations (34.) to
(35.).

ij t
FV =
⎪
⎭
⎪
⎬
⎫
⎪
⎩
⎪
⎨
⎧ +
× × ×
− −
− − −
2
) (
1 1
1 1 1
j t i t
j t i t ij t
L L
G G FV for alli ≠ j . (45.)

The stopping rule to identify the optimal values of
ij t
FV

from equations (44.) and (45.)
is shown in equation (46.). The stopping condition is met by maximizing.

∑∑
ij
ij t
FV MAX

(46.)

subject to

0.999 < (
∑
i
T
i
IM
NO /
∑ ∑
ij
ij t
FV ) < 1.001, and (47.1)
69

0.999 <(
∑
j
T
j
IM
ND /
∑ ∑
ij
ij t
FV ) < 1.001; or, alternatively, (47.2)
0.999 <
∑∑ −
ij
ij t
FV
1
/
∑ ∑
ij
ij t
FV ) < 1.001. (47.3)

Due to only limited information available about interstate trade in services, trade in
services between states was assumed to be negligible. However, it is essential to estimate
service sector trades for the 18 non-service sectors to complete the trade flow estimates
between states and it is a caveat that this two-step approach cannot provide a useful way to
estimate trade for the 18 service sectors. Another approach developed in Park (2006) will
address this caveat by using the geographical weighted regression (GWR) method for the
service sectors.

5. Results
The 2001 trade flows between the U.S. states for 29 USC Sector commodities are
estimated according to the AFM and DFM. In Table II-2 and Table II-3. I suggest a
summary of the estimated trade flow matrices for USC Sector 15 (Plastics and Rubber) as an
example, respectively, to show AFM and DFM estimates.
Values in Table II-2 are rounded off to the nearest integer and only unreported values
are estimated, represented as Symbol 2 or 3, without changing all of the given values shown
in Table A.II-3 of Appendix A2. The total estimated value for USC Sector 15 is 278,797
million dollars, showing $35 million difference from 1997 CFS’s, corresponding to very
high levels of accuracy.

70

Table II-2 Summary of 1997 Estimated Trade Flows between U.S. States by Adjusted Flow
Model for USC Sector 15, ‘Plastics and Rubber’ ($M.)
State

O
T
i
∑
i
V
ij
(∑
i
V
ij
)/ O
T
i

D
T
j
∑
j
V
ij
(∑
i
V
ij
)/ D
T
j

Alabama 3869 3914 1.012 3809 3826 1.005
Alaska 79 84 1.069 148 182 1.227
Arizona 1789 1908 1.067 3106 3105 1.000
Arkansas 3114 3138 1.008 3775 3601 0.954
California 19867 19953 1.004 24374 24451 1.003
Colorado 1911 1924 1.007 2632 2636 1.001
Connecticut 2768 2765 0.999 2596 2582 0.995
Delaware 791 788 0.997 1715 1574 0.918
District of Columbia 77 40 0.512 113 121 1.075
Florida 6385 6423 1.006 9291 9312 1.002
Georgia 9658 9536 0.987 11559 11621 1.005
Hawaii 188 201 1.071 320 394 1.230
Idaho 557 626 1.123 830 922 1.111
Illinois 22300 22039 0.988 16948 16968 1.001
Indiana 7732 7843 1.014 9600 9434 0.983
Iowa 4148 4175 1.006 3233 3252 1.006
Kansas 3140 3173 1.011 2687 2641 0.983
Kentucky 4436 4486 1.011 5384 5335 0.991
Louisiana 7343 7377 1.005 2803 2846 1.015
Maine 479 477 0.995 1004 1022 1.018
Maryland 2165 2248 1.039 3710 3720 1.003
Massachusetts 6765 6713 0.992 5731 5631 0.983
Michigan 9923 10005 1.008 14241 14259 1.001
Minnesota 5073 5094 1.004 5610 5700 1.016
Mississippi 2993 2862 0.956 2743 2699 0.984
Missouri 4329 4442 1.026 6054 6077 1.004
Montana 143 159 1.115 530 561 1.058
Nebraska 1231 1237 1.005 1639 1637 0.999
Nevada 554 646 1.166 1026 1124 1.095
New Hampshire 1284 1201 0.935 1312 1238 0.944
New Jersey 13705 13574 0.990 10216 10243 1.003
New Mexico 311 350 1.126 670 744 1.111
New York 10325 10376 1.005 13047 12805 0.981
North Carolina 11207 11250 1.004 9723 9788 1.007
North Dakota 216 247 1.142 626 630 1.006
Ohio 20427 20434 1.000 17050 17111 1.004
Oklahoma 2775 2779 1.001 2281 2295 1.006
Oregon 1978 2150 1.087 2752 2886 1.049
Pennsylvania 14060 13827 0.983 12099 12104 1.000
Rhode Island 953 766 0.804 1001 921 0.920
South Carolina 6012 6027 1.003 6025 6001 0.996
South Dakota 412 400 0.971 470 502 1.069
Tennessee 9947 9771 0.982 7608 7539 0.991
Texas 29313 29425 1.004 24996 25091 1.004
Utah 1014 1078 1.063 1825 1760 0.964
Vermont 377 388 1.028 703 729 1.036
Virginia 6007 5915 0.985 5812 5530 0.951
Washington 3126 3323 1.063 4507 4608 1.023
West Virginia 2582 2449 0.949 1832 1938 1.058
Wisconsin 8769 8677 0.989 6713 6752 1.006
Wyoming 155 113 0.726 352 349 0.992
TOTAL 278763 278797 1.000 278831 278797 1.000
Note: V=Value, O
T
i
= Total Origin (Output) value of State i, D
T
j
= Total Destination (Input)
value of State j.

71

Figure II-1 Estimated U.S. Trade Flows by the Adjusted Flow Model only for the Incomplete
Trade Flows of 1997 Commodity Flow Survey for USC Sector 15, ‘Plastics and Rubber’

Note: Order of States follows the order shown in Table II-1.

Figure II-1 shows the estimated trade flows using the Adjusted Flow Model only for the
incomplete 1997 CFS trade flows, represented as Symbol 2 or 3 in Table A.II-3 of Appendix
A2. This summarizes the state-to-state trade flows at the 15 degree heights and the largest
amount of the hide trade for the sector is $911 million from Illinois to Indiana. This figure
reveals that the AFM approach provides the information of unreported trades without
touching any published amounts in CFS

72

Table II-3 Summary of Estimated Trade Flows between U.S. States for 2001 by Doubly-
Constrained Fratar Model for USC Sector 15, ‘Plastics and Rubber’ ($M.)
State
a
I
f
i ∑
j
V
ij

IM
ND
T
j
Sale_VD
IM
E
f
i
∑
i
V
ij

IM
NO
T
i
Sale_VO
Alabama 390 2967 2963 4662 598 3503 3500 5695
Alaska 55 300 299 494 1 83 81 118
Arizona 354 2776 2776 4347 142 1662 1662 2506
Arkansas 249 2216 2215 3424 341 2510 2510 3961
California 2432 21377 21373 33068 1778 16527 16525 25424
Colorado 382 2910 2910 4572 167 1732 1732 2638
Connecticut 298 2740 2740 4220 280 2252 2252 3518
Delaware 88 605 605 962 157 900 900 1467
District of Columbia 53 296 295 485 2 61 61 88
Florida 1148 8254 8238 13057 457 4521 4508 6914
Georgia 687 6292 6287 9694 708 5853 5847 9113
Hawaii 82 490 486 794 6 152 149 219
Idaho 109 728 728 1163 23 331 330 491
Illinois 1143 10437 10436 16084 1573 12397 12399 19403
Indiana 739 6344 6347 9837 1073 8782 8782 13687
Iowa 354 2787 2787 4362 411 2954 2954 4673
Kansas 285 2231 2230 3494 255 1999 2000 3131
Kentucky 447 3506 3505 5490 836 4498 4498 7408
Louisiana 349 2346 2343 3743 867 2788 2783 5076
Maine 105 834 834 1305 45 536 536 808
Maryland 403 3130 3130 4906 162 2034 2033 3049
Massachusetts 506 4927 4926 7546 651 4996 4996 7843
Michigan 1290 9712 9711 15280 1252 10120 10122 15796
Minnesota 478 4339 4339 6691 457 4121 4121 6359
Mississippi 204 1835 1834 2832 261 2068 2067 3234
Missouri 502 4408 4407 6820 551 4101 4100 6460
Montana 80 475 475 770 7 145 145 211
Nebraska 224 1485 1484 2373 112 1113 1112 1701
Nevada 148 1170 1171 1831 77 792 792 1206
New Hampshire 117 1046 1046 1616 150 1513 1513 2310
New Jersey 642 6317 6316 9665 792 6435 6435 10037
New Mexico 127 784 772 1265 14 421 415 605
New York 1334 10901 10899 16993 1087 7826 7825 12378
North Carolina 751 6662 6658 10295 981 7363 7358 11589
North Dakota 75 466 466 752 16 216 216 323
Ohio 1397 11142 11141 17416 2328 16044 16042 25516
Oklahoma 323 2192 2191 3492 350 2496 2497 3953
Oregon 264 2145 2145 3345 161 1452 1453 2240
Pennsylvania 950 9680 9679 14764 1652 11798 11798 18681
Rhode Island 76 709 709 1090 126 925 925 1461
South Carolina 506 3390 3375 5412 949 5463 5450 8906
South Dakota 78 545 545 864 33 334 334 509
Tennessee 625 5160 5158 8034 984 5949 5948 9629
Texas 1645 14284 14281 22125 2369 14954 14950 24059
Utah 174 1418 1418 2212 81 890 890 1349
Vermont 60 475 475 743 25 260 260 395
Virginia 563 4738 4737 7362 733 4909 4909 7836
Washington 460 3743 3744 5838 198 2333 2333 3514
West Virginia 136 1041 1040 1634 472 1758 1759 3098
Wisconsin 605 5588 5587 8601 901 7646 7647 11872
Wyoming 48 292 292 471 12 118 118 180
TOTAL 24538 204634 204546 318294 27665 204634 204571 322637
Note: Sale_VO={(
∑
i
ij
V +
f
i
IM
E )/0.72}, Sale_VD= (
∑
j
ij
V +
f
i
a
I )/0.72}

73

Based on the completed trade flows matrix via the AFM, the DFM was used to estimate
2001 trade flows between states. The summary of the estimation is shown in Table II-3 and
the state-to-state trade flows are suggested in Figure II-2 for the trade flows within-state and
Figure II-3 for the state-to-state trade flows without the diagonal movements for USC Sector
15, ‘Plastic and Rubber’. The highest within-state trade flows for 2001 occurs in California,
at $13 billion, while Texas accounted for the largest outbound flow to California, with
$1,119 million in current dollars, among all states.

Figure II-2 Estimated within-State Trade Flows via the Doubly-
constrained Fratar Model for USC Sector 15

74

Figure II-3 Estimated Trade Flows without diagonal movements via the
Doubly- constrained Fratar Model for USC Sector 15

Note: Order of States follows the order shown in Table II-1.

For the sake of accuracy of the estimated values, I suggest the sum of trade flow
between states (
∑
i
ij
V or
∑
j
ij
V ) and IMPLAN total value (
T
i
IM
O or
T
j
IM
D ). The ratios of
the two are very close to 1.00 for every state. Foreign exports and foreign imports are
suggested as the trade flows to/from the Rest of World, although foreign imports are already
included in domestic trade flows between the states.
Because the values in the trade matrix estimated by the DFM are producer values,
dividing by 0.72, the producer/purchaser ratio for USC Sector 15, the sales value enables
comparisons with the raw CFS trade flows or the trade flows matrix via the AFM. For
instance, the estimated producer value for California-to-California shipments in 2001 is
$12,944 million, about 43 percent (=100*(12,944/0.72 -12,557)/ 12,557), increasing from
1997 total origin value (=12,557 million dollars) in current dollars.
75

Similarly, all other values in Table II-3 can be compared with those in Table II-2.
However, because the current estimated values of trade flows in the states with ports does
not include foreign exports, the values might be overestimated if
f
i
IM
E is added to the
diagonal value in trade flows and compared with the CFS which counts foreign exports as
domestic flows. Also, Sale_VO or Sale_VD, meaning {(
∑
i
ij
V +
f
i
IM
E )/0.72} or
{(
∑
j
ij
V +
f
i
a
I )/0.72} respectively, is suggested in Table II-3 to compare to total values of
Table II-2. For instance, the sum of origin flows from California is increased by 30 percent
(=100*(25,424-19,953)/19,953) in 2001 as the nominal value, while the total sum of the
estimated trade flows is different by 15.7 percent from the total sum in Table II-2
(=100*(322,637-278,797)/ 278,797).

6. Conclusions
Although a large variety of IO models have been developed, the construction of
multiregional IO models has remained a challenging task. In this study, I suggest how trade
flows between the U.S. states can be estimated and updated using secondary data. These
results can serve as a basis on which to build a NIEMO-type multiregional IO model for the
U.S.
I applied a two-step method, based on incomplete 1997 CFS trade flow data between
states and IMPLAN regional commodity balance data. Before doing any estimating, I
created several kinds of conversion tables to reconcile different data code systems. With the
adjusted flow model, incomplete trade flows for 1997 CFS are filled out. Based on this trade
flows matrix, including foreign imports/exports in the U.S. trade flows, I estimated the 2001
76

trade flows matrix, only including foreign imports using a doubly-constrained Fratar Model.
Those 2001 trade flows are constructed for 29 USC commodity sectors in the final step. As
an example, USC Sector 15 is highlighted in the results section, where I can verify that the
model and estimations are acceptable at a reasonable level of accuracy.
However, the 2001 model based on 1997 data has some limitations. Limited data on the
sources of service trade flows has restricted reporting the economic interrelationships of the
services sectors between regions. The rapid increases in telecommunications, especially
web-based industries, however, require us to investigate the amount of service trades
between regions. Although there are some suggestions on how to estimate service trade
flows, these still require strong assumptions. Therefore, in order to overcome these
limitations an alternative methodology is required.
77

IV. CHAPTER II REFERENCES

Batten D.F., 1982, The Interregional Linkages Between National and Regional Input-Output
Models. International Regional Science Review 7:53-67.

Canning, P. and Z. Wang, 2005, A Flexible Mathematical Programming Model to Estimate
Interregional Input-Output Accounts. Journal of Regional Science 45(3): 539-563.

Chenery, H.B., 1953, Regional Analysis, in The Structure and Growth of the Italian
Economy, edited by H.B. Chenery, P.G. Clark and V.C. Pinna, U.S. Mutual Security
Agency, Rome: 98-139.

Erlbaum, N. and J. Holguin-Veras, 2005, Some Suggestions for Improving CFS Data
Products, Paper presented at Commodity Flow Survey (CFS) Conference, Boston Seaport
Hotel & World Trade Center, Boston, Massachusetts, Jul. 8~9.

Giuliano, G., P. Gordon, Q. Pan, J.Y. Park, and L. Wang, 2007, Estimating Freight Flows for
Metropolitan Area Highway Networks Using Secondary Data Sources, Networks and
Spatial Economics (forthcoming).

Harrigan, F., J.W. McGilvray, and I.H. McNicoll, 1981, The Estimation of Interregional
Trade Flows, Journal of Regional Science, 21(1): 65-78.

Isard, W., 1951, Interregional and Regional Input-Output Analysis: A Model of a Space
Economy, Review of Economics and Statistics, 33: 318-328.

Jack Faucett Associates, INC, 1983, The Multiregional Input-Output Accounts, 1977:
Introduction and Summary, Vol. I (Final Report), prepared for the U.S. Department of
Health and Human Services, Washington.

Jackson, R.W., W.R. Schwarm, Y. Okuyama, and S. Islam, 2006, A Method for
Constructing Commodity by Industry Flow Matrices, Annals of Regional Science,
(forthcoming).

Lahr, M.L., 1993, A Review of the Literature Supporting the Hybrid Approach to
Constructing Regional Input-Output Models, Economic Systems Research, 5: 277-293

Lee, C., 1973, Models in Planning: An Introduction to the Use of Quantitative Models in
Planning, NY, Pergamon Press Inc.

Lindall, S., D. Olsen, and G. Alward, 2005, Deriving Multi-Regional Models Using the
IMPLAN National Trade Flows Model, Paper presented at the 2005 MCRSA/SRSA
Annual Meeting, April 7-9, Arlington, VA.

78

Liu, L.N. and P. Vilain, 2004, Estimating Commodity Inflows to a Substate Region Using
Input-Output Data: Commodity Flow Survey Accuracy Tests, Journal of Transportation
and Statistics, 7: 23-37.

Miller, R.E. and P. D. Blair, 1985, Input-Output Analysis: Foundations and Extensions, New
Jersey: Prentice-Hall.

Moses, L.N., 1955, The Stability of Interregional Trading Patterns and Input-Output
Analysis, American Economic Review, 45: 803-832.

Park, J.Y., 2006, “Estimation of State-by-State Trade Flows for Service Industries”, Paper
presented at North American Meetings of the Regional Science Association International
53rd Annual Conference, Fairmont Royal York Hotel, Toronto, Canada, November 16-
18.

Park, J.Y., P. Gordon, J.E. Moore II, and H.W. Richardson, 2006, Simulating the State-by-
State Effects of Terrorist Attacks on Three Major U.S. Ports: Applying NIEMO
(National Interstate Economic Model), available at the
http://www.metrans.org/nuf/documents/MooreII.pdf.

Park, J.Y., P. Gordon, J.E. Moore II, H.W. Richardson, and L. Wang, 2007, Simulating the
State-by-State Effects of Terrorist Attacks on Three Major U.S. Ports: Applying NIEMO
(National Interstate Economic Model): 208-234, in H.W. Richardson, P. Gordon and J.E.
Moore II, eds., The Economic Costs and Consequences of Terrorism. Cheltenham:
Edward Elgar.

Polenske, K.R., 1980, The U.S. Multiregional Input-Output Accounts and Model, DC Health,
Lexington, MA.

Wilson, A.G., 1970, Inter-regional Commodity Flows: Entropy Maximizing Approaches.
Geographical Analysis 2:255-282.

79

V. CHAPTER III: ESSAY THREE

Estimation of Interstate Trade Flows for Service Industries

1. Introduction and Issue
National economic models of the U.S. aggregate over large numbers of diverse
regions. However, many regional scientists are interested in evaluating socioeconomic
impacts that involve the states, especially in terms of their policy significance. A U.S. multi-
regional input-output (MRIO) model is an example of useful spatial disaggregation, but
models like this are still difficult to construct because of the difficulty of developing detailed
state-by-state trade data (Lahr, 1993).
The U.S. Commodity Transportation Survey data on interregional trade flows have been
available since 1977, but reporting was discontinued for some years. For the years since
1993, this data deficit can be met to some extent with the recent Commodity Flow Survey
(CFS) data from the Bureau of Transportation Statistics (BTS). Since 1993, CFS data have
been widely used, but the data have several inherent problems (Erlbaum and Holguin-Veras,
2005). The most serious one among them is that the CFS data do not include trade flows
below the state level but also that they are not complete even between the states. Since
Polenske (1980) and Faucett Associates (1983), there has been no comprehensive inventory
of flows for probably these reasons.
Furthermore, even though the commodity flow data between the states of the U.S. are
published every five years, there is no inventory of trade flows for services. Recent
approaches to estimating state-by-state trade flows of U.S. based on 1997 CFS, therefore,
80

have paid too little attention to the problems of estimating the trade flows among service
sectors and maintained strong assumptions of no or small trades in these sectors. However,
in the modern information economy, this is a serious omission.
Therefore, this research addresses new approaches to relaxing these assumptions and,
instead, proposes estimates of interstate trade flows for the service sectors. Using
Geographically Weighted Regressions (GWR) econometric analysis, this study proposes and
implements a sequence of computational and spatial econometric steps for estimating inter-
state trade flows among all of the major service sectors, especially as required for
implementing a U.S. interstate MRIO model. Furthermore, the approach can be expanded to
examine the economic relationships between sub-state-level areas, as well as to forecast
future trade flows.
The next section of this essay develops the background for estimating state-by-state
trade flows. In the following section, based on specially prepared data, the Geographically
Weighted Regressions (GWR) econometric methodology and an application is explained. In
the final section, conclusions and some remarks are elaborated.

2. Trade Flows Estimation and Service Industries
The existence of many unreported values in trade flow data has required relying on other
data sources for completeness. Harrigan et al (1981) compared several old methodologies for
estimating interregional trade flows and showed ‘more information, better results’, based on
1973 Scotland data. This is because all techniques used as examples are simple ratio-based
methodologies. Using the CFS, based on an approach of location quotients, Lie and Vilain
81

(2004) estimated trade inflows of subregional levels below the states. However, this requires
very restrictive assumptions, resulting in sizable errors in the estimates.
More recently, in order to construct trade flows as the basic data set for an MRIO, there
have been some attempts to estimate interregional trade flows. Using data from the 1997
Commodity Flow Survey (CFS), Jackson et al. (2006) combined IMPLAN data to adjust
incomplete CFS information using an error-minimizing equation via Box-Cox
transformation regressions and double-log regressions. Another attempt included a doubly-
constrained gravity model based on the Oak Ridge National Labs (ORNL) data for county-
to-county distances by mode of transportation, CFS for ton-miles by sector, and IMPLAN
data for total supply and demand by county (Lindall et al, 2005). The CFS data were used for
a criterion index, whether the average of the estimated ton-miles is matched to the CFS ton-
miles or not. Generally, doubly- constrained gravity models reflect interactive effects of
trades, but not only to allocate the exports to regions. The model basically accepts the fact
that attractiveness of an economy is proportional to the trade flows, but distances between
two regions are inversely proportional. Different from these studies, Canning and Wang
(2005) developed a new approach estimating interregional trade flows basically based on the
techniques developed by Wilson (1970) and Batten (1982), and tested the performance using
Global Trade Analysis Project (GTAP) data. Park et al. (2007) used the same basic data
sources as Jackson et al. (2006) and Lindall et al (2005), but adopted a different estimation
approach relying on an AFM (adjusted flow model) and a DFM (doubly-constrained Fratar
model). This two-step approach allows the incomplete CFS to be completed with the AFM
and updated with the DFM.
82

However, the common problem in these all trials is that there has not been a way to fully
estimate the trade flows for service sectors. Only average coefficients of commodity sectors
(Jackson et al, 2006) or high (but not specified in the study) exponents for the distance
functions were used for the estimation of service industries, excluding trade flows over long
distances (Lindall et al, 2005). Or the strong assumption of no service trade flows, mainly
due to the implausibility of estimates was applied (Park et al. 2007).
However, the problem should be addressed by estimating state-by-state or sub-state level
flows for the service industries. Unfortunately, the problem residing in the all studies results
partly from an applicable methodology to estimate trade volumes or partly from inexistence
of appropriate data to apply the regional economic models available currently. To conduct a
survey to verify, at least, the state-level trade flows of service industries incurs huge costs,
although the problem is widely acknowledged in light of the characteristics of the modern
information society. But there is some good news: we have access to total imports and
exports obtained from the widely used IMPLAN data and an appropriate methodology which
has heretofore not been applied.
83

Table III-1 Definition of Variables, 2000
V ariab le s D e sc rip tio n N o te
D e pend ent U S C se rv ice se c to rs R efe r to T ab le A 1 in A ppend ic e s
(U S C 30 to U S C 47)
Ag g _ u s c 0 1
Ag g _ u s c 0 2
Ag g _ u s c 0 3
Ag g _ u s c 0 4 R e fe r to T ab le A 1 in A p p e nd ic e s
Ag g _ u s c 0 5
Ag g _ u s c 0 6
Ag g _ u s c 0 7
Ag g _ u s c 0 8
Ag g _ u s c 0 9
m ean _ a g g A v erag e o f A gg_usc se cto rs (su m o f agg_usc i)/(su m o f nu m b er o f i)
Pop P o p u latio n o f e ac h state U nit: 1000
De n P o p u latio n/state size U nit: 1000/sq u are m ile s
Pop _ c h a P erc en t o f po p u latio n c h an g e betw een 19 90 an d 20 00 U nit: % , 100*{= (2000- 1990)/1990}
I_ a c A g ed-c hild index A ged=o v er o r at 65, C hild = u nd er 18
I_ de p D e pend ency index 100*(under18 + o v e r6 5)/(b e tw een1865)
P_ n _ w h P e rc e nt o f no n-w hite re sid e nts
P_ t_ im m P erc en t o f to tal im m ig ran ts
P_ oth _ s t P erc en t o f p o p u latio n b o rn at o th er areas
M_ t e m p A v erag e tem p eratu re F a hre n he it
R_ c r i m e C rim e rates per 1000 po pulatio n (V io le nt+ P ro p e rty )/p o p
I_ e c o n de p
P erc en t o f ec o no m ic d ep end en c y
100*(nu m b er o f unem plo y ed)/(num b e r
of e m p loy e d )
P_ b e lo w p o v Perc en t of belo w p overty statu s d uring last one y ear
HR _ i n d e x H o m e o w ne r and R e nter Index
U S Index= 200: R e calc u late o w ne r
o ccu pied ho use pric es and re nt
p ay m e nts to b e ind e x e d
I n depen d en t G _ st_ tax G en eral states tax U nit: c e nts p e r d o llar
I _liv c o st L iv in g c o st ind ex
U S ind e x = 1 0 0 : (ho stp ital c o st)+ (e ne rg y
e x p e nd itu re s)+ (g aso line p ric e s)
D is p _inc D ispo sab le inc o m e per c ap ita U nit: $1000
Go v _ e x p G o v ern m ent ex pend iture U nit: $ B illio ns
GS P i G ro ss state p ro d u c ts fo r U S C se c to r i U nit: $ M illio ns, i= 3 0 to 4 7
N _ pub _1 2 N u m b e r o f p u b lic e nro llm e nt u nd e r 1 2 U nit: 1000, U S C 4 2
N _ pu b_ h ig h N u m b e r o f p u b lic e nro llm e nt hig he r e d u c atio n U nit: 1000, U S C 4 2
Ta x _ g a s T ax fo r G as U nit: c e nts/g allo n, U S C 3 0 -3 5
I_ E n er_exp E nergy expend itu re index U S = 100, U S C 30-35
R ev _ tele R e v e nu e fro m te le c o m m u nic atio n U nit: $ M illio ns,U S C 3 6
S p n_d tra v E x p e nd itu re o f d o m e stic _ trav e l U nit:$ M ., 2001, U S C 4 5
C o mmo n
Va r i a b l e s
se t* *
S p e c ific
va ria b le s
s e t***
T he G W R is re g resse d fo r e ach U S C
se c to r, tally 1 8 tim e s.
Co r e
va r i a b l e s
se t*

Note: All variables are based on year 2000, except one variable, Spn_dtrav.
* All core variables are used for the GWR regression, and the unit for all variables is $million.
** Some independent variables in common variables set are selected for the GWR regressions
*** Specific variables set is only for specified sector shown in Note

84

The following basic processing steps involve building a database from Park et al (2007),
developing the new approach of this study and relaxing the assumption of no interstate trade
in services and, instead, estimating interstate trade flows for all the major service sectors of
the USC-Sector system (29 commodity sectors and 18 service sectors as shown in Table
A.III-1 in the Appendix B). The latter is easily converted to many other sector systems and
introduced in the next section.

3. Data and Model
Data sources for this study are various. Dependent variables for each USC service
sectors are selected from 2001 IMPLAN domestic imports and the values are shown at Table
A.III-2. The IMPLAN data support estimates of five kinds of economic transactions: total
commodity outputs, domestic/foreign imports/exports. From these data, intra-state flows
within each state are calculated, which can be converted to a diagonal state-by-state matrix
( T
ˆ
) in order to be added to the fitted non-diagonal trade flows (
*
T ).
See Table A.III-3 for the fixed intra-state flow of service sectors for T
ˆ
, where the intra-
state flows are adjusted by including foreign imports. This is because the foreign imports
which are not consumed in the local area but transported to other state(s) are excluded from
the state- or county-level IMPLAN data (Park et al., 2007; Giuliano et al., 2007). Refer to
Table III-1 for the definition of each variable. To estimate the dependent variable for
estimation of the various USC service sector trade flows, basic independent variables are
drawn from the State and Metropolitan Area Data Book
(http://www.census.gov/compendia/smadb/) and County Business Patterns
(http://censtats.census.gov/cbpnaic/cbpnaic.shtml) for 2000.
85

Table III-1 shows which independent variables are selected for each dependent (service)
variable, where all independent variables are classified into three categories. Core variables
used for the GWR regression include USC commodity sectors, aggregated to nine sectors
corresponding to the aggregation sector in the CFS. This is to examine the effects of physical
commodity sectors on service sectors. To compare individual effects of commodity sectors
and overall effect of all commodities, average values of commodity sectors are added. This
separation reveals whether or not it is acceptable to use average parameters obtained from
commodity sectors for the estimation of service trade flows. From independent variables in a
common variables set, some variables are selected for the GWR regressions, according to the
expected relation with each dependent industry. The specific variables set is only for
specified sectors noted in the last column of Table III-1. All variables are based on year 2000,
except one variable of expenditure of domestic travel (Spn_dtrav), because of the limitations
obtaining the data for 2000.
There is only limited information available on interstate trade in services. In general, the
gravity model is widely used to estimate trade flows, because it reflects spatial effects.
Indeed, territorial location is important for trading, and gravity models reflect most
importantly inherent distance. However, the possibly overused log-transformed gravity
model is based on ordinary least squares (OLS), and hence ignores ‘spatial dependency’ and
‘heteroscadasticity’ resulting from many inherent invisible characteristics of each region
(Anselin, 1988; Anselin and Griffith, 1988; LeSage 1999). Therefore, an econometric
approach to reflecting spatial effects is critical when estimating trade flows (Porojan, 2001;
LeSage and Pace, 2006).
86

However, direct use of the revised gravity model based on spatial autoregressive models
(e.g. SAR, SEM, or SAC) cannot deliver the direct estimates of trade flows, but fix only
some parameters. Still, in those approaches, distance is critical. However, for service sectors,
other socioeconomic factors are more important due to special characteristics. To directly
estimate trade flows of service sectors between states, therefore, it is important to consider
other socioeconomic and environmental effects as well as distance effects simultaneously.
The Geographically Weighted Regressions (GWR) econometric model (Brunsdon et al,
1996) can be applied to this approach. However, the locally linear regression model by
McMillen (1996) is not appropriate in the estimation of trade flows for service sectors
because it only reflects distances between states, although the approach is widely used as a
GWR approach.
The GWR model can be rewritten according to LeSage (1999, p.205~206) as applying
Weighted Least Squares (WLS). If
i
ε has heteroscadasticity according to spatial ( i )
characteristics, a new variance matrix of error term
i
ε can be specified as follows.

⎥
⎥
⎥
⎥
⎥
⎦
⎤
⎢
⎢
⎢
⎢
⎢
⎣
⎡
=
2
2
2
2
1
2
0 0
0 0
0 0
n
σ
σ
σ
σ
ε
L
M O M M
L
L
(1.)

Therefore, net domestic import of service sector vector
i
y can be weighted by
2
ε
σ
letting error term
i
ε follow normal distributions.

87

i
K
k
i k i
x y ε β + =
∑
=1
(2.)

i
i
K
k
i k
i i
i
x
y
σ
ε
β
σ σ
+ =
∑
=1
1

i
i
σ
ε
~ ) , 0 ( I N (3.)

If
i
W is a similar weight to adjust a regional heteroscadasticity, then equation (3.)
can be shown as,

i i i i i
W X W y W ε β + = (4.)

where, an i is spatial observation (e.g. state, county, etc.) and
i
β is K x 1
parameter column vector related to region i .
i
W represents n x n diagonal matrix
including distance-based weights for i and hence reflects the distance between i and
all other regions,
i
d .

Here, distance-based weights can be suggested via three types. First, Brunsdon et al.
(1996) introduce the “bandwidth” decay parameter θ shown in equations (5.) and (7.),
where different θ s will produce different exponential decay results varying over regions.

) / exp( θ
i
B
i
d W − = (5.)

88

The second set of weights were developed by McMillen (1996) using a tri-cube
function, where
i
q indicates the distance of the
th
q nearest neighbor to region i . This
weights the relationship between regions once more using
i
d and
i
q .

⎩
⎨
⎧ < −
=
otherwise
q d if q d
W
i i i i
M
i
0
, ) ) / ( 1 (
3 3
(6.)

Finally, the Gaussian standard normal density function can be applied to
i
W , where
σ indicates the standard deviation of the distance vector
i
d .

) / ( σθ φ
i
G
i
d W = (7.)

The “bandwidth” decay parameter θ relies on cross-validation value that uses a score
function shown in equation (8.) and indicator
i
q can be computed as shown in equation (6.).
However, because the indicator
i
q in tri-cube function of equation (6.) only depends
distances, it is not useful in this study. Hence, to estimate the optimal decay parameter θ ,
new iteration approach was used as,

∑
=
≠
−
n
i
i i
y y
1
2 *
)) ( ( θ (8.)

where,
*
i
y
≠
is the optimally fitted value of
i
y omitting region i .
89

Equation (8.), therefore, shows that the θ is selected when sum of residual is
minimized using the similar weighted least squares in equation (4.) via iterations. In other
words, the most optimally estimated
*
i
y
≠
reflects all effects of the independent variables and
invisible spatial relations given at a fixed distance. Therefore, the weights
λ
i
W ( λ = B or
G ) are not fixed as other spatial autoregressive models are, but flexibly changed, depending
on the independent variables. This study used the GWR approach based on “bandwidth”
decay parameter θ to adjust the distance-effects with various independent variables.
Because the optimal “bandwidth”, θ
ˆ
, is selected from equation (8.) omitting
regressed region i itself, I separate the trade matrix T into two types. One is the diagonal
matrix of intra-state trade movement by each state, denoted as T
ˆ
. Another is non-diagonal
trade flow ( T
~
), empty in the main diagonal. The existing data set includes only total
domestic imports without the non-diagonal state-by-state trade flows T
~
. Therefore, given T
ˆ
,
the T
~
is estimated using GWR. Based on the prediction vector
h
i
y for each service sector
after being estimated optimally based on the θ
ˆ
from the equation (4.), the estimated trade
flows
*
~
T of T
~
is calibrated,

s i
R
i
h
i
s W W y T )
ˆ
ˆ (
~
1 * * − Σ
= (9.)

90

where,
*
i
W is the fitted weighted matrix,
1
ˆ
− Σ
i
R
W is inverse matrix of
i
R
W
ˆ
Σ
where
i
R
W
ˆ
Σ
is diagonal matrix of row sum of
*
i
W ,
h
i
y ˆ is diagonal matrix of
h
i
y , and s
indicates each service sector ( s =USC Sector 30 to USC Sector 47).

Finally, the estimated trade flows are obtained from the equation (10.).

*
s
T =
*
~
s
T +
s
T
ˆ
(10.)

An application is shown in the next section for s =USC Sector 42, the education
services industry.

4. An Application: the Case of Education Services
As described in the previous section, in the case exploiting data on the sectors without
trade flow data but with only domestic outflows or inflows data, limited applications to
estimate trade flows have been implemented. For the estimation of service sectors, I applied
the GWR methodology with the
G
i
W bandwidth to USC Sector 42, the education service
sector. Another application using
B
i
W bandwidth does not show better estimates than the
application of the
G
i
W bandwidth for all cases. From Table III-1, the selected variables are
described in Table III-2.
The GWR approach yields different coefficient results by each state and hence each
state can have its own fixed coefficients. To understand the effects of commodity sectors,
two regressed results are shown in Table III-3 and III-4. While Table III-3 only has average
91

domestic imports of all 29 commodities as an independent variable, Table III-4 shows 9
types of aggregate USC commodity sectors, corresponding to the classification of the CFS.
From adjusted R-squares in both tables, I can verify that the GWR results explain more than
those of OLS.

Table III-2 Selected Variables for the GWR: USC Sector 42, Education Service
Variables Mean Standard Deviation
Dependent USC42 791 1144
Agg_usc01 3272 3321
Agg_usc02 5284 5298
Agg_usc03 438 428
Agg_usc04 4718 5014
Agg_usc05 4469 4184
Agg_usc06 7477 7924
Agg_usc07 9626 9734
Agg_usc08 12011 13118
Agg_usc09 5925 6100
mean_agg 5913 5979
Independent Pop 5518 6164
M_temp 53 9
R_crime 41 10
HR_index 229 14
Disp_inc 21 26
Gov_exp 25 3
GSP_USC42 926 1080
N_pub_12 300 367
N_pub_high 1554 2121
Common
Variables
set
Specific
variables
set
Core
variables
set

Note: The independent variable of mean_agg is used instead of the 9 types
of aggregated USC sectors to verify the effects by commodity type. Those
results are shown in Table III-3 and III-4, respectively.

92

Table III-3 Results of Geographically Weighted Regression: Case of mean_agg
Intercept m ean_agg Pop M _tem p R _crim e H R _index D isp_inc G ov_exp N _pub_12
N _pub_
hi g h
GSP_US
C42
R_sq Adj_R_sq Bandw idth D W N
GW R
al 1150.17 -0.2496
***
0.5523
***
12.9856 -9.4452
*
-11.8662
**
9.1019 56.7000
***
-1.3386
***
2.1068
***
-0.3272
***
0.9636 0.9546 1.2469 51
ak -2306.24
*
-0.1319
***
0.3309
*
6.0023 -1.8352 1.2458 17.0324 76.4152
***
-1.2723
**
3.7760
***
-0.4155
***
az -144.20 -0.2681
***
0.5319
***
6.5346 -3.5243 -6.9871
*
18.3912
**
70.0723
***
-1.3914
***
2.6544
***
-0.3923
***
ar 963.21 -0.2596
***
0.5525
***
11.3061 -8.4431
*
-10.9313
**
10.9697 58.2389
***
-1.3454
***
2.2295
***
-0.3390
***
ca -653.07 -0.2522
***
0.5126
***
4.7965 -2.3022 -4.9825 18.9991
**
73.0609
***
-1.4074
***
2.8038
***
-0.4004
***
co 176.92 -0.2692
***
0.5309
***
8.3196 -4.9340 -8.1594
*
17.3680
**
66.4490
***
-1.3465
***
2.5345
***
-0.3770
***
ct 1304.42 -0.2302
***
0.5367
***
15.5556
*
-10.4539
**
-12.8597
**
7.4765 54.8795
***
-1.2888
***
1.9284
**
-0.3122
***
de 1283.96 -0.2333
***
0.5392
***
15.1821
*
-10.3109
**
-12.7279
**
7.7203 55.1926
***
-1.2990
***
1.9684
**
-0.3148
***
dc 1270.43 -0.2344
***
0.5394
***
15.0261
*
-10.2379
**
-12.6512
**
7.8718 55.3199
***
-1.3007
***
1.9832
**
-0.3159
***
fl 1242.01 -0.2458
***
0.5563
***
13.7338
*
-9.9336
**
-12.3279
**
8.0252 56.0574
***
-1.3434
***
2.0234
**
-0.3210
***
ga 1212.97 -0.2443
***
0.5500
***
13.7954
*
-9.8278
**
-12.2331
**
8.4280 56.1356
***
-1.3309
***
2.0517
**
-0.3223
***
hi -2102.81
*
-0.1787
***
0.4682
***
0.7952 -0.5771 1.1933 12.6037 77.3049
***
-1.6984
***
4.0093
***
-0.4564
***
id -326.65 -0.2498
***
0.4931
***
8.0792 -3.6822 -6.6463 19.6522
**
71.2110
***
-1.3081
***
2.6478
***
-0.3873
***
il 1041.09 -0.2472
***
0.5412
***
12.9028 -9.1897
*
-11.4369
**
9.9647 57.1173
***
-1.3094
***
2.1109
***
-0.3285
***
in 1127.61 -0.2423
***
0.5407
***
13.6921
*
-9.6139
*
-11.8816
**
9.1452 56.4144
***
-1.3058
***
2.0562
**
-0.3234
***
ia 834.81 -0.2543
***
0.5370
***
11.6235 -8.2685
*
-10.5090
**
11.8625 58.6900
***
-1.3044
***
2.2151
***
-0.3386
***
ks 602.37 -0.2680
***
0.5439
***
9.6029 -6.8146 -9.5079
**
14.1362 61.4360
***
-1.3418
***
2.3902
***
-0.3567
***
ky 1161.89 -0.2431
***
0.5438
***
13.7419
*
-9.6842
*
-12.0267
**
8.9084 56.3069
***
-1.3153
***
2.0621
**
-0.3233
***
la 1007.54 -0.2631
***
0.5606
***
10.9795 -8.4563
*
-11.0324
**
10.6292 58.0821
***
-1.3635
***
2.2317
***
-0.3394
***
me 1339.35 -0.2244
***
0.5343
***
16.1823
*
-10.7339
**
-13.0726
**
6.9275 54.2960
***
-1.2721
***
1.8208
**
-0.3065
***
md 1271.87 -0.2342
***
0.5392
***
15.0520
*
-10.2471
**
-12.6608
**
7.8568 55.2998
***
-1.3001
***
1.9813
**
-0.3158
***
ma 1312.14 -0.2292
***
0.5362
***
15.6756
*
-10.5060
**
-12.9059
**
7.3770 54.7732
***
-1.2861
***
1.9123
**
-0.3113
***
mi 1118.65 -0.2365
***
0.5340
***
14.2182
*
-9.8116
**
-11.9062
**
8.9259 56.0627
***
-1.2807
***
1.9816
**
-0.3188
***
mn 715.76 -0.2513
***
0.5252
***
11.7094 -8.0472
*
-10.0766
**
12.6523 59.0941
***
-1.2702
***
2.2034
***
-0.3384
***
ms 1077.35 -0.2554
***
0.5547
***
12.0898 -8.9911
*
-11.4569
**
9.8749 57.3563
***
-1.3471
***
2.1668
***
-0.3328
***
mo 929.07 -0.2560
***
0.5453
***
11.6457 -8.4868
*
-10.8556
**
11.1615 58.2529
***
-1.3263
***
2.2110
***
-0.3376
***
mt -131.22 -0.2536
***
0.4902
***
9.2294 -4.2250 -7.4401
*
19.9006
**
69.4634
***
-1.2626
***
2.5549
***
-0.3811
***
ne 470.56 -0.2662
***
0.5339
***
9.5919 -6.3961 -9.0992
**
15.2236
*
62.5050
***
-1.3220
***
2.4143
***
-0.3601
***
nv -434.30 -0.2544
***
0.5104
***
6.2765 -3.0811 -5.9674 19.0328
**
71.7801
***
-1.3710
***
2.7121
***
-0.3931
***
nh 1315.12 -0.2279
***
0.5351
***
15.7959
*
-10.5502
**
-12.9305
**
7.3137 54.6614
***
-1.2810
***
1.8924
**
-0.3102
***
nj 1289.26 -0.2322
***
0.5380
***
15.3112
*
-10.3539
**
-12.7666
**
7.6591 55.0883
***
-1.2947
***
1.9563
**
-0.3140
***
nm 162.63 -0.2753
***
0.5439
***
7.4484 -4.6222 -7.9974
*
17.1236
*
66.9050
***
-1.3840
***
2.5722
***
-0.3817
***
ny 1281.45 -0.2307
***
0.5355
***
15.4175
*
-10.3731
**
-12.7433
**
7.6951 54.9975
***
-1.2858
***
1.9364
**
-0.3131
***
nc 1255.95 -0.2380
***
0.5440
***
14.6262
*
-10.1257
**
-12.5326
**
7.9966 55.5996
***
-1.3138
***
2.0048
**
-0.3179
***
nd 294.28 -0.2605
***
0.5132
***
10.0178 -6.0273 -8.5609
**
16.6132
**
63.3223
***
-1.2668
***
2.4039
***
-0.3600
***
oh 1197.98 -0.2377
***
0.5393
***
14.4196
*
-9.9564
**
-12.2646
**
8.4916 55.8177
***
-1.3002
***
2.0092
**
-0.3191
***
ok 689.45 -0.2702
***
0.5521
***
9.4853 -7.0344 -9.7661
**
13.4313 60.8098
***
-1.3587
***
2.3780
***
-0.3551
***
or -717.86 -0.2412
***
0.4869
***
6.4916 -2.7951 -5.0314 19.6002
**
73.4101
***
-1.3466
***
2.7969
***
-0.3948
***
pa 1261.03 -0.2336
***
0.5376
***
15.0706
*
-10.2344
**
-12.6156
**
7.9388 55.2897
***
-1.2946
***
1.9733
**
-0.3155
***
ri 1314.70 -0.2293
***
0.5367
***
15.6740
*
-10.5139
**
-12.9166
**
7.3468 54.7755
***
-1.2875
***
1.9118
**
-0.3112
***
sc 1243.83 -0.2404
***
0.5468
***
14.3356
*
-10.0357
**
-12.4388
**
8.1070 55.7872
***
-1.3214
***
2.0193
**
-0.3193
***
sd 390.09 -0.2634
***
0.5242
***
9.8061 -6.2279 -8.8642
**
15.8768
*
62.9379
***
-1.2975
***
2.4117
***
-0.3603
***
tn 1147.09 -0.2461
***
0.5469
***
13.3685 -9.5425
*
-11.9145
**
9.0914 56.5440
***
-1.3245
***
2.0879
**
-0.3255
***
tx 602.86 -0.2793
***
0.5632
***
8.1547 -6.3050 -9.3079
**
14.0750 61.9834
***
-1.3922
***
2.4489
***
-0.3639
***
ut -129.01 -0.2612
***
0.5170
***
7.5564 -3.9109 -7.1829
*
18.6908
**
69.7094
***
-1.3492
***
2.6166
***
-0.3866
***
vt 1306.22 -0.2282
***
0.5346
***
15.7310
*
-10.5131
**
-12.8847
**
7.4082 54.7185
***
-1.2803
***
1.8995
**
-0.3107
***
va 1256.18 -0.2365
***
0.5414
***
14.7759
*
-10.1509
**
-12.5567
**
8.0111 55.5072
***
-1.3068
***
1.9994
**
-0.3173
***
wa -733.40 -0.2340
***
0.4671
***
7.7377 -3.0386 -5.2175 20.3000
**
73.7468
***
-1.2951
***
2.7883
***
-0.3927
***
wv 1233.98 -0.2371
***
0.5406
***
14.6237
*
-10.0667
**
-12.4432
**
8.2164 55.6357
***
-1.3048
***
2.0071
**
-0.3182
***
wi 961.55 -0.2437
***
0.5323
***
13.0142 -9.1135
*
-11.1356
**
10.3777 57.2571
***
-1.2807
***
2.0769
**
-0.3270
***
wy 38.87 -0.2620
***
0.5128
***
8.7354 -4.6107 -7.8430
*
18.5153
**
67.8131
***
-1.3103
***
2.5417
***
-0.3788
***
OLS
-54.42 -0.25
***
0.55
***
6.85 -7.08 -6.77 9.43 66.07
***
-1 .4 1
***
2.58
***
-0 .3 4
***
0.9514 0.9392 2.1642 51

92
93

Table III-4 Results of Geographically Weighted Regression: Case of Agg_USC01-Agge_USC09

93

94

Figure III-1 The Estimated and Actual Domestic Imports for The USC Sector 42, Education Service, by Each State
USC42: Education Service
0
1000
2000
3000
4000
5000
6000
7000
8000
al ak az ar ca co ct de dc fl ga hi id il in ia k s k y la me md ma mi mn ms mo mt ne nv nh nj nm ny nc nd oh ok or pa ri sc sd tn tx ut vt va wa wv wi wy
State
Domestic Import ($M.)
yhat
y

Note: yhat=
h
USC
y
42
and y=
42 USC
y .
94

95

The results of estimated coefficients show factors such as population size (Pop),
disposable income per capita (Disp_inc), related to education consumption induce increases
of the consumption of education service from other states. While general government
expenditures (Gov_exp) increase imports of education services, more Gross State Product of
education services in each state (GSP_USC Sector 42) decreases imports. Also, higher
education (N_pub_high) induces more imports, but complementary education (N_pub_12)
relies upon each individual state. Finally, worse socioeconomic environments (R_crime and
HR_index) reduce the imports of education services.
Another important result from Table III-3 is that more domestic imports of commodities
(mean_agg) negatively affect the domestic imports of education services. Further, Table III-
4 shows that commodity characteristics differently affect education services for each state.
This is different from the general belief that higher commodity trades induce more service
trades..Therefore more cautious approaches are required, for example, when using the
average coefficient of commodities for the coefficient of service sectors. This result might be
helpful to understanding which commodity in domestic import is more appropriate to induce
a targeted service sector for each state.

96

Figure III-2 Estimation of State-by-State Trade Flows: Case of Education Service, USC Sector
42

Note: 1. Order of State follows the order in Table III-3.
2. Exclude the main diagonal trade, that is, intrastate movements from the
estimated trade flows, 42
*
USC T .

Because the adjusted R-square in Table III-4 is higher than that in Table III-3, the
dependent variable is estimated based on the coefficients in Table III-4 and the given
independent variables. Figure III-1 shows the estimated (
h
USC
y
42
) and actual (
42 USC
y )
97

domestic imports for education service (USC Sector 42) by each state. This figure shows that
the GWR regressions reflect the spatial effects.
Based on the optimized bandwidth from Table III-4, 1.4316, the state-by-state trade
flows are obtained as shown in Figure III-2. The trade flows in Figure III-2 do not involve
the main diagonal in the trade matrix, T
ˆ
, to adjust the magnitude of trade into the figure.
Therefore, the net state-by-state trade flows,
*
~
T , for education services sector shows the
amount of net domestic import by each state (State X) and the estimated results of domestic
exports by each state (State Y). All trade flows for USC Sector 42, 42
*
USC T , are obtained
adding the
42
ˆ
USC
T to 42
*
~
USC T .
However, this is only a net domestic imports-based GWR application. It might
suggest the necessity of constraining the estimated exports from the trade matrix according
to State Y coordinates upon the actual net domestic exports available from the IMPLAN
dataset. While the current approach is a one-way constrained GWR model relying on
domestic import, the doubly-constrained GWR estimation might be helpful, if these partial
constrained models are not enough.

5. Conclusions and Remarks
Limited access to data on services trade flows has restricted the ability to estimate the
economic interrelationships of services sectors between regions. The rapid increases in
telecommunications, especially web-based industries, however, compel us to investigate the
amount of such trades between regions. Although there are various suggestions on how to
estimate service trade flows, most have focused on the estimation of non-service sectors
along with strong assumptions on limited service trades.
98

In this essay, I overviewed the studies and methodologies dealing with the estimation
of state-by-state trade flows for the U.S. Still, due to limitations of service sector information
and its own characteristics, and depending on different transport trends from commodities an
alternative methodology is required. To estimate the state-by-state trade flows of service
sectors, I applied Geographically Weighted Regression, which was elaborated by LeSage.
The approach applied here reflects other factors as well as distance with respect to trade
flows, and hence is more appropriate to estimating trade flows for service sectors. In an
application to the education services industry, the GWR estimation results show that it does
a good job explaining the net domestic imports for each state.

However, this approach should be tested more extensively and elaborated in two
ways:
1) based on actual trade data and its sum, and
2) constrained by net domestic exports.

Furthermore, it should be noted that because the bandwidth, θ , reflects all the
independent effects, more independent variables would increase bandwidth and decease the
distance effects.
In spite of the possible problems, the application of this approach answers many key
issues addressed in regional science. Here, at least three applications can be discussed,
beyond service sector estimation. GWR can be used to estimate the economic
interrelationships between sub-state regions. In fact, Lindall et al (2005) estimate trade flows
at the county level; GWR can provide alternative results for trade flows at the same or at
99

lower levels, based on secondary data, but only if there are net in- or out-bound data. This
makes it possible for a new MRIO-type model at the sub-state level to be constructed.
Second, the application of GWR supports the estimation of trade flows for services sectors
without resorting to severe assumptions, but in the same way as non-service sector
estimation. That is, the GWR approach can be consistently extended to other estimates of
industry sectors. Third, because the GWR includes statistical probabilities for the results,
these can support discussions of reasonable criteria to determine which model should be
selected. Finally, due to the nature of econometrics, the approach can be used to predict
potential trade flow. In this case, the GWR can be applied to adjust the four- or five-year
based CFS data to a one-year base. Furthermore, with appropriate changes in independent
variables for a targeted regional economy, the GWR can be used to forecast changes in
various trade flows and hence key changes within an MRIO. Therefore, a dynamic MRIO
model also can be constructed and run according to reasonable scenarios. This can support
more plausible results on long-term effects as well as for short-term effects. Therefore, the
application of GWR to the estimation trade flows has many more implications than simply
as an alternative trade flows’ methodology. These should be elaborated and investigated.
100

VI. CHAPTER III REFERENCES

Anselin, T., 1988, Spatial Econometrics: Methods and Models. Dorddrecht: Kluwer
Academic Publishers.

Anselin, T. and D.A. Griffith, 1988 Do Spatial Effects Really matter in Regression
Analysis? Papers in Regional Science 65(1): 11-34.

Batten, D.F., 1982 The Interregional Linkages Between National and Regional Input-Output
Models. International Regional Science Review 7:53-67.

Brundson, C.A., S. Fotheringham, and M. Charlton, 1996, Geographical Weighted
Regression: A Method for Exploring Spatial Nonstationarity. Geographical Analysis 28:
281-298.

Canning, P. and Z. Wang, 2005, A Flexible Mathematical Programming Model to Estimate
Interregional Input-Output Accounts. Journal of Regional Science 45(3): 539-563.

Erlbaum, N. and J. Holguin-Veras, 2005, Some Suggestions for Improving CFS Data
Products. Proceedings of Commodity Flow Survey (CFS) Conference, Boston Seaport
Hotel & World Trade Center, Boston, Massachusetts, Jul. 8~9.

Giuliano, G., P. Gordon, Q. Pan, J.Y. Park, and L. Wang, 2007, Estimating Freight Flows for
Metropolitan Area Highway Networks Using Secondary Data Sources, Networks and
Spatial Economics (forthcoming).

Harrigan, F., J.W. McGilvray, and I.H. McNicoll, 1981, The Estimation of Interregional
Trade Flows. Journal of Regional Science 21(1): 65-78.

Jack Faucett Associates INC., 1983, The Multiregional Input-Output Accounts, 1977:
Introduction and Summary, Vol. I (Final Report), prepared for the U.S. Department of
Health and Human Services, Washington.

Jackson, R.W., W.R. Schwarm, Y. Okuyama, and S. Islam, 2006, A Method for
Constructing Commodity by Industry Flow Matrices. The Annals of Regional Science
40:909-920.

Lahr, M.L., 1993, A Review of the Literature Supporting the Hybrid Approach to
Constructing Regional Input-Output Models. Economic Systems Research 5: 277-293.

LeSage, J.P., 1999, The Theory and Practice of Spatial Econometrics. University of Toledo,
MN, Available at the http://www.spatial-econometrics.com/html/sbook.pdf.

LeSage, J.P., and R.K. Pace, 2006, Spatial Econometric Modeling of Origin-Destination
flows. Available at SSRN: http://ssrn.com/abstract=924609.
101

Lindall, S., D. Olsen, and G. Alward, 2005, Deriving Multi-Regional Models Using the
IMPLAN National Trade Flows Model. Proceedings of 2005 MCRSA/SRSA Annual
Meeting, April 7-9, Arlington, VA.

Liu, L.N. and P. Vilain P., 2004, Estimating Commodity Inflows to a Substate Region Using
Input-Output Data: Commodity Flow Survey Accuracy Tests. Journal of Transportation
and Statistics 7(1): 23-37.

McMillen, D.P., 1996, One Hundred Fifty Years of Land Values in Chicago: A
Nonparametric Approach. Journal of Urban Economics 40: 100-124.

Park, J.Y., P. Gordon, J.E. Moore II, H.W. Richardson, and L. Wang, 2007, Simulating the
State-by-State Effects of Terrorist Attacks on Three Major U.S. Ports: Applying NIEMO
(National Interstate Economic Model): 208-234, in H.W. Richardson, P. Gordon and J.E.
Moore II, eds., The Economic Costs and Consequences of Terrorism. Cheltenham:
Edward Elgar.

Polenske, K.R., 1980, The U.S. Multiregional Input-Output Accounts and Model. DC Health,
Lexington, MA.

Porojan, A., 2001, Trade Flows and Spatial Effects: The Gravity Model Revisited. Open
Economies Review 12(3): 265-280.

Wilson, A.G., 1970, Inter-regional Commodity Flows: Entropy Maximizing Approaches.
Geographical Analysis 2:255-282.

102

VII. CHAPTER IV: ESSAY FOUR

An Evaluation of Input-Output Aggregation Errors Using a New MRIO Model

1. Introduction
National economic models involve high degrees of data aggregation. It is well known
that this may cause an ‘ecological fallacy’ as suggested by Robison (1950). For the case of
national input-output models, they include both significant sectoral and spatial aggregation.
Ever since Isard (1951) proposed the “ideal” interregional input-output approach,
regional scientists and others have considered ways in which spatial disaggregation could be
operationalized. The various approaches to developing multi-regional input-output models
(MRIOs; see Chenery (1953) and Moses (1955) for the general framework, and see Polenske
(1980) and Faucett Associates (1983) for empirical applications) complicate the problem
because disaggregation is obtained at the price of various simplifying assumptions that
introduce further complex trade-offs for consideration.
Since Leontief (1951) had pointed to the “comparative goodness” of the sectoral
aggregation errors, theoretical and empirical studies of sectoral aggregation error have been
conducted with traditional regional input-output models by various scholars (Hatanaka,
1952; McManus, 1956; Theil, 1957; Ara, 1959; Morimoto, 1970; Gibbons et al, 1982;
Dietzenbacher, 1992). However, as Miller and Blair (1985: p.175) noted, the spatial
aggregation problem may be more difficult. Comparable interindustry matrices between
regions require the same aggregated trade sectors in practical uses but generally they may
103

rely on separate data sources and classification systems. Furthermore, complete trade data
between the states are generally unavailable or difficult to use. These reasons have made
empirical studies on spatial aggregation error difficult. In spite of the importance of spatial
aggregation error, empirical studies for the U.S. that test the errors have been unavailable
with the exception of Blair and Miller (1983) and Miller and Shao (1990).
As a new MRIO of the fifty states and the District of Columbia (and the rest of the
world), I introduce the National Interstate Economic Model (NIEMO) to conduct
aggregation error tests. NIEMO is constructed from 2001 IMPLAN data sets for the 51 U.S.
areas and (updated) data from the 1997 U.S. Commodity Flow Survey. To make these
sources compatible, I aggregated IMPLAN’s 509 sectors to 47 sectors which are labeled the
“USC Sectors”. Based on this effort, I examined the trade-offs involved between sectoral
aggregation and spatial disaggregation.

2. Aggregation Errors and NIEMO
Blair and Miller (1983) tested spatial aggregation error associated with an MRIO-type
model based on the 1963 U.S. data constructed by Polenske (1980). Based on their three
types of tests, they argue that the multiregional aggregations are “acceptable, not large”, and
accept the errors. These findings are consistent with their previous theoretical and
hypothetical tests (Miller and Blair, 1981) for interregional aggregation error. From the
results, they conclude that one can get satisfactory results from a “spatially-aggregated
model” instead of a detailed NIEMO-type model (Blair and Miller, 1983: p.196). More
recently, Miller and Shao (1990) reported further empirical tests for multiregional
aggregation errors using the 1977 U.S. MRIO data set compiled by Jack Faucett Associates
104

(1983).
3
They concluded that, “spatial aggregation generates less inaccuracy than sectoral
aggregation.” (p. 1652).
The new MRIO combines state-level input-output models from IMPLAN with
interregional trade flow data from the U.S. Commodity Flow Survey (CFS) aggregated to 47
economic sectors over 52 regions (50 States, Washington, D.C., and the rest of the world),
resulting in a matrix with almost 6 million cells (Park et al., 2007). Construction of the
model involved substantial data assembly and considerable data manipulation, resulting in
two basic tables: industrial trade coefficients tables and regional interindustry coefficients
tables. While trade tables by industry are difficult to construct because of incomplete
information in the CFS data, the interindustry tables present no serious problems because
reliable data are available from IMPLAN at the state and industry levels. For estimating
commodity trade flows, this study adopts a new approach initially developed by Park et al.
(2004) for estimating interstate trade flows for non-service sectors, based on the 1997
Commodity Flow Survey data via an adjusted flow model (AFM) and a doubly-constrained
Fratar model (DFM).
4

The initial version of NIEMO was developed as a demand-side model and was used to
estimate the economic impacts of hypothetical terrorist attacks on major three ports of U.S.
(Park et al., 2007) and the economic impacts due to a mad-cow-disease outbreak in the state

3
Actually, Miller and Shao used data updated and reported in 1988 by various Boston College
researchers.
4
Very recently, Freight Analysis Framework (FAF) released the 2002 FAF
2
commodity origin-
destination database (http://ops.fhwa.dot.gov/freight/freight_analysis/faf/faf2_tech_document.htm,
2006). Because the trade data for commodity sectors include service values used up in commodity
processing and shipping, however, it is hard to say that the values released are comparable to CFS
commodity flows.
105

of Washington (Park et al., 2006).
5
Although the latter tested sectoral aggregation from 509
IMPLAN sectors to 47 USC Sectors and revealed only minor errors in the sectoral
aggregation, it still leaves out thorough tests that would help in evaluating the trade-offs
when contemplating whether NIEMO can be substituted for IMPLAN.
This study, therefore, involves a comparison of sectoral and spatial aggregation, because
NIEMO was aggregated sectorally but disaggregated spatially. To focus on the trade-offs
between NIEMO’s sectoral aggregation and spatial disaggregation, this study of aggregation
and disaggregation tests relies upon a different approach than in the previous studies of
aggregation error tests. This study starts by assuming an ideal interstate input-output model.

3. Data and Model
My approach depends on 2001 IMPLAN data. Before testing the sectoral and spatial
aggregation error tests, tests of data compatibility are required. First, two types of input
output model with 47 USC Sectors at the national level, USIO and USIO_U, are developed.
The former is a traditional Leontief industry-by-industry type IO for the aggregation error
tests, and the latter is a sectoral aggregation IO model, the “use matrix” of IMPLAN. The
USIO_U is used to verify the compatibility of IMPLAN data and the newly constructed
USIO and NIEMO for the aggregation and disaggregation tests. Because the IMPLAN
industry total outputs are based on a use matrix, the results via USIO_U should be the same
as the IMPLAN industry total outputs when value-added data are input.

5
A supply-side NIEMO was developed recently by Park (2006a), and used to test the hypothetical
economic impacts of closures of the twin ports of Los Angeles and Long Beach due to hypothetical
‘dirty-bomb’ attacks.
106

Figure IV-1 shows the plan of this study. With three models in hand, it is possible to
estimate errors of industrial and spatial aggregation so that we can learn about the trade-offs
and possible benefits when moving from a spatially aggregated input-output model with
considerable sectoral detail to a spatially disaggregated model that has less sectoral detail. In
Figure IV-1, the upper-left box indicates the unknown “perfect” interregional input-output
model, IRIO, containing 509 sectors and 52 regions. Just below it and to the right is NIEMO,
aggregated to 47 sectors. In the lower-left is the IMPLAN model for the U.S., but with 509
sectors. In the lower right is an aggregation of IMPLAN sectors to the 47 USC Sectors used
in NIEMO, which I call USIO.

Figure IV-1 Various Types of IO Models Specified
USIO
(1x47)
2
Newly constructed
IMPLAN IO
(1x509)
2
Available from
IMPLAN progra m
Perfect IRIO
(52x509)
2
Unknown
NIEMO
(52x47)
2
Newly constructed

Further, instead of BEA GSP (Gross State Product) data, which are some of the most
credible data, IMPLAN value added data were used as the exogenous inputs for two reasons.
107

First, IMPLAN data reveal greater accuracy when the 51 areas are aggregated to the U.S.
than the BEA GSP data, as shown in Table A.IV-1 of Appendix C. Also, The BEA GSP data
do not include total output for each state and therefore NIEMO cannot input the value added
vector for every state. In Table A.IV-2 of Appendix C, there is a comparison table of total
output between IMPLAN and BEA data and between the sum of states for IMPLAN and
national IMPLAN for 47 USC Sectors.

To address the issue of data compatibility, a transposed IO matrix of A is first
suggested as in equations (1.) to (4.).

V XA X
T
+ = (1.)

where, total output row vector [ ]
n
x x , ,
1
L = X , total value-added row vector
[]
n
v v , ,
1
L = V , and technical coefficients matrix
1
X Z A
−
=
ˆ
when there are n industry
sectors, where matrix
⎥
⎥
⎥
⎦
⎤
⎢
⎢
⎢
⎣
⎡
=
nn n
n
z z
z z
L
M O M
L
1
1 11
Z is a matrix of the element
ij
z , denoting the
deliveries in dollars from industry sector i to j .
6

Hence, to estimate new total output based on current value added, equation (1.) converts
to equation (2.), and shown as,

6
The hat of X
ˆ
notes a diagonalized matrix from a vector X .
108

1 T
) A V(I X
−
− = (2.)

Similar to equation (1.), the technical coefficients matrix of USIO_U,
A
U , is
constructed based on the use matrix, U , shown as equation (3.),

V ) (U X X
T A S S
+ = (3.)

where
A
U =
1 S
) X U
−
ˆ
( . Note U is a use matrix from IMPLAN data, where
S
X a vector of
IMPLAN industry total outputs, that is
∑
=
i
ij
s
j
u x .
ij
u denotes deliveries in dollar values
from input commodity i to industry sector j .

Therefore, the total outputs of USIO_U are estimated as,

1 T A S
] ) (U V[I X
−
− = (4.)

109

Table IV-1 Model estimates of NIEMO and two USIOs using value added and error
comparisons with aggregations of IMPLAN and BEA total output (units: $m.)
Total Output Aggregation Errors
USC Sec. NIEMO
1
USIO
2
USIO_U
2
IMPLAN
3
BEA
4
ANIEMO
5
AUSIO
6
AUSIO_U
7
USC 1 153,313 150,992 172,661 173,097 195,922 -11.43% -12.77% -0.25%
USC 2 110,960 108,985 118,664 118,853 117,442 -6.64% -8.30% -0.16%
USC 3 39,587 39,741 44,678 44,785 50,631 -11.61% -11.26% -0.24%
USC 4 79,370 78,920 84,758 84,932 105,455 -6.55% -7.08% -0.20%
USC 5 258,648 255,814 285,412 286,070 309,327 -9.59% -10.58% -0.23%
USC 6 55,643 53,332 61,415 61,546 42,350 -9.59% -13.35% -0.21%
USC 7 51,122 49,874 52,563 52,637 17,100 -2.88% -5.25% -0.14%
USC 8 17,923 17,410 19,023 19,049 18,364 -5.91% -8.60% -0.13%
USC 9 8,359 7,825 9,115 9,129 8,813 -8.43% -14.28% -0.15%
USC 10 310,773 276,341 370,696 371,603 376,181 -16.37% -25.64% -0.24%
USC 11 67,975 64,587 75,862 76,034 75,787 -10.60% -15.06% -0.23%
USC 12 126,365 120,135 134,193 134,457 134,295 -6.02% -10.65% -0.20%
USC 13 14,712 13,864 16,177 16,209 16,157 -9.24% -14.47% -0.20%
USC 14 128,487 122,630 141,830 142,133 142,483 -9.60% -13.72% -0.21%
USC 15 182,180 173,396 203,203 203,666 200,586 -10.55% -14.86% -0.23%
USC 16 91,709 87,159 101,490 101,676 103,276 -9.80% -14.28% -0.18%
USC 17 127,697 120,112 142,028 142,353 141,106 -10.30% -15.62% -0.23%
USC 18 193,074 188,166 203,515 203,883 213,869 -5.30% -7.71% -0.18%
USC 19 148,667 145,164 172,635 172,998 130,168 -14.06% -16.09% -0.21%
USC 20 90,946 88,630 97,633 97,801 97,201 -7.01% -9.38% -0.17%
USC 21 102,573 91,573 121,217 121,498 118,894 -15.58% -24.63% -0.23%
USC 22 165,613 159,533 184,164 184,519 182,176 -10.25% -13.54% -0.19%
USC 23 294,577 281,213 330,681 331,350 322,670 -11.10% -15.13% -0.20%
USC 24 542,474 510,070 600,089 601,195 608,161 -9.77% -15.16% -0.18%
USC 25 373,244 354,808 445,979 447,184 441,871 -16.53% -20.66% -0.27%
USC 26 100,183 92,634 117,725 118,010 118,170 -15.11% -21.50% -0.24%
USC 27 103,896 97,461 113,919 114,130 112,783 -8.97% -14.60% -0.18%
USC 28 66,855 64,792 73,493 73,637 73,417 -9.21% -12.01% -0.19%
USC 29 99,929 91,526 105,745 105,923 105,886 -5.66% -13.59% -0.17%
USC 30 268,644 254,021 296,155 296,699 343,430 -9.46% -14.38% -0.18%
USC 31 937,997 930,238 1,011,034 1,013,114 899,778 -7.41% -8.18% -0.21%
USC 32 857,192 844,892 873,626 875,258 851,286 -2.06% -3.47% -0.19%
USC 33 473,816 459,999 501,917 502,771 469,464 -5.76% -8.51% -0.17%
USC 34 157,525 155,967 162,091 162,269 101,953 -2.92% -3.88% -0.11%
USC 35 926,980 918,234 941,151 942,803 1,021,032 -1.68% -2.61% -0.18%
USC 36 562,979 553,430 585,257 586,269 710,579 -3.97% -5.60% -0.17%
USC 37 1,275,606 1,261,626 1,285,741 1,287,273 1,361,671 -0.91% -1.99% -0.12%
USC 38 1,661,115 1,650,566 1,679,422 1,681,503 1,775,397 -1.21% -1.84% -0.12%
USC 39 994,981 988,131 1,006,962 1,008,257 1,105,607 -1.32% -2.00% -0.13%
USC 40 207,801 206,390 209,967 210,209 290,414 -1.15% -1.82% -0.12%
USC 41 435,826 432,495 443,272 443,881 481,024 -1.81% -2.57% -0.14%
USC 42 84,076 83,462 85,545 85,680 107,147 -1.87% -2.59% -0.16%
USC 43 1,147,180 1,133,271 1,186,485 1,188,873 1,094,726 -3.51% -4.68% -0.20%
USC 44 151,333 150,017 154,024 154,279 154,140 -1.91% -2.76% -0.17%
USC 45 475,583 470,460 497,804 498,852 500,925 -4.66% -5.69% -0.21%
USC 46 1,279,201 1,276,548 1,288,675 1,288,980 2,019,174 -0.76% -0.96% -0.02%
USC 47 704,222 691,941 754,487 755,883 534,941 -6.83% -8.46% -0.18%
TOTAL 16,708,912 16,368,374 17,564,179 17,593,207 18,403,229 -5.03% -6.96% -0.16%
Note: 1. After being estimated by NIEMO, total outputs are aggregated spatially from 51 states to U.S.
2. Two types of USIOs consisting of 47 USC sectors and U.S. are newly constructed
3. 509 IMPLAN total outputs in use matrix are simply aggregated to 47 USC sectors using an IMPLAN_USC bridge
table, where I dropped IMPLAN 5 sectors, or 501-503 and 507-508, due to nonexistence of coefficients for those sectors.
4. NAICS sectors of BEA suggested in
‘http://www.bea.doc.gov/bea/industry/gpotables/gpo_action.cfm?anon=315&table_id=10984&format_type=0’ are
aggregated to 47 USC sectors
5. ANIEMO=(NIEMO-IMPLAN)/IMPLAN
6. AUSIO =(USIO -IMPLAN)/IMPLAN
7. AUSIO_U=(USIO_U-IMPLAN)/IMPLAN
110

Two USIO models used the aggregated value-added vector of 47 USC Sectors from
IMPLAN’s 509 sectors; those estimates via equations (2.) and (4.) are compared with the
actual 509 IMPLAN industry total outputs aggregated to 47 USC Sectors. The third through
fifth columns in Table IV-1 show comparisons between the estimated results via USIO and
USIO_U and the aggregated IMPLAN industry total outputs. Similarly, for the use of
aggregation and disaggregation tests, the total output of the demand-driven NIEMO via the
transposed matrix of W (= CA ) will provide useful comparisons with those aggregate
models and the aggregate IMPLAN industry total output. The equation is shown in (5.),
where new total outputs
N
X via NIEMO based on the value added vector for 51 states are
summed spatially to the U.S.

1 T
N N
) W (I V X
−
− = (5.)

where,
N
X and
N
V is a (52x47)x1 row vector, and
T
W (52x47)x(52x47)

is the transposed
matrix of W , where W = CA described in the NIEMO construction discussion.
7
Note
matrix C is a (52x47)x(52x47) diagonal block matrix of interstate trade flow of each USC
industry sector and A is a (52x47)x(52x47) block diagonal interindustry matrix of each
state.

Three estimated results and aggregated IMPLAN total outputs are shown in Table IV-1,
where all USC sectors via USIO_U do not show more than 0.3 percentage sectoral

7
However, all USC Sectors in the Rest of World (=1x47) are zero elements, because there is no
valued-added information for this region.
111

aggregation errors from the aggregated IMPLAN total output.
8
However, because the
aggregation errors by total output via USIO stem from adjusted coefficients using make and
use tables, it is hard to say that there are sizable sectoral aggregation errors from those USIO
models built with the IMPLAN data.
Similarly, the total outputs via NIEMO have aggregation and disaggregation errors when
comparing the aggregated IMPLAN data. However, this includes the errors when combining
inter-industry matrices with trade matrices. Therefore, apart from the direct comparison of
IMPLAN industry total output, it is still necessary to estimate new total outputs via the
IMPLAN IO model in order to compare sectoral aggregation and spatial disaggregation
errors between NIEMO and the IMPLAN IO.
Nevertheless, because these small errors via USIO_U denote that USIO and NIEMO are
constructed based on plausible bridges and aggregation processes
9
, this study invoked further
tests, by estimating new total outputs with new final demands for the demand-side IMPLAN
IO, USIO, and NIEMO models. These models are expressed as,

m m m
Γ ) B (I X
1 −
− = (6.)

where,
m
X is total output vector via IO model m , which denotes a type of IO model,
IMPLAN IO (I), USIO (UI or UN), or NIEMO (N),

m
B is the coefficients matrix for IO model m , where B is a 509x509 industry-by-
industry technical IO coefficients if m is IMPLAN IO model, a 47x47 aggregated

8
The definitions of 47 USC Sectors are shown in Appendix 3.
9
These errors might arise from rounding decimal points when creating the USIO_U matrix and
combining the data.
112

industry-by-industry technical IO coefficients if m is USIO model, or a
(52x47)x(52x47) NIEMO coefficients if m is NIEMO, and

m
Γ is the vector of inputs for IO model m . For the comparison of national
IMPLAN IO and USIO; Γ is 509 elements of one dollar if m is the IMPLAN IO
model and dollar values aggregated to 47 USC sectors shown in the second column
of Table IV-2 if m is the USIO model. For the comparison of USIO and NIEMO,
Γ is $51 for each element if m is the USIO model, and $1 for each element if m is
NIEMO, as shown in the second column of Table IV-3.

Based on the three models introduced in equation (6.), Figure IV-2 elaborates the
approach used in this study, based on Figure IV-1. One contrast is IMPLAN’s IO vs. USIO
for comparison of sectoral aggregation errors; another is NIEMO vs. USIO for comparison
of spatial aggregation errors. Figure IV-2 highlights errors that are plausibly encountered as I
attempt to estimate any one of the models from the others. Except for the disaggregation
error
N UN _
ε and the unknown mixed error
N I _
β , all other errors shown in the Figure IV-2
are aggregation errors.

113

Figure IV-2 Errors Resulting from Various Types of IO models
UI PI _
α
USIO
(1x47)
2
Newly constructed
NIEMO
(52x47)
2
Newly constructed
IMPLAN IO
(1x509)
2
Available from
IMPLAN program
Perfect IRIO
(52x509)
2
Unknown
UI I _
ε
UN N _
ε
UN PN _
α
N PN _
α
N UN _
ε
N I _
β
I PI _
α
PI
X
PN
X
I
X
N
X
UI
X
UN
X

To understand the nature of the various errors, however, it should be pointed out that
some of them, the ε s, can be estimated from the models depicted. The others ( α s and β )
are conceptual. The following shows that the four unknown errors can be inferred from the
estimated errors, but the notations necessary to calculate the errors will be explained first.
114

Notations of Total Outputs
Pi
X : Total output vector aggregated from 509 IMPLAN sectors and 52 regions via
the unknown perfect IRIO to 47 USC Sectors and the whole U.S., where N i = (for
the case of the same total input as NIEMO) or I (for the case of the same total input
as IMPLAN IO).

N
X : Total output vector, spatially aggregated from 47 USC Sectors and 52 regions
via NIEMO.

I
X : Total output vector, sectorally aggregated from 509 IMPLAN sectors and the
U.S., via the available IMPLAN IO for the U.S., from the IMPLAN program

Ui
X : Total output vector of 47 USC Sectors and the U.S. via USIO, where N i =
(for the case of the same total input as NIEMO) or I (for the case of the same total
input as IMPLAN IO).

Assumptions
I PI _
α =
UN N _
ε : Unknown vector of spatial aggregation errors are the same as
calculated vector of spatial aggregation errors. (7.1)

115

N PN _
α =
UI I _
ε : Unknown vector of sectoral aggregation errors are the same as
calculated vector of sectoral aggregation errors. (7.2)

Based on the total output notations and the basic assumptions, a discussion of the errors
can be constructed, as in equations (8.).

Unknown Aggregation Errors
j
PI
j
I
j
PI j
PI_I
x
x x
α
−
= : Unknown element j in vector
I PI _
α of spatial aggregation
errors of IMPLAN IO stemming from the perfect IRIO. (8.1)

j
PN
j
N
j
PN j
PN_N
x
x x
α
−
= : Unknown element j in vector
N PN _
α of sectoral aggregation
errors of NIEMO stemming from the perfect IRIO. (8.2)

j
Pi
j
Ui
j
Pi j
Pi_Ui
x
x x
α
−
= : Unknown element j in vector
Ui Pi _
α of sectoral and spatial
aggregation errors of USIO stemming from the perfect IRIO, where N i = (for the
case of the same total input as NIEMO) or I (for the case of the same total input as
IMPLAN IO). (8.3)

Unknown Mixed Errors
N I _
β : Unknown vector including errors by sectoral aggregation and information
by disaggregation spatially from IMPLAN IO to NIEMO.
116

Calculated Aggregation Errors
j
N
j
UN
j
N j
N_UN
x
x x
ε
−
= : Calculated element j in vector
UN N _
ε of spatial
aggregation errors of USIO stemming from NIEMO. (9.1)

j
I
j
UI
j
I j
I_UI
x
x x
ε
−
= : Calculated element j in vector
UI I _
ε of sectoral aggregation
errors of USIO stemming from the IMPLAN IO. (9.2)

Calculated Disaggregation Errors
j
UN
j
N
j
UN j
UN_N
x
x x
ε
−
= : Calculated element j in vector
N UN _
ε of spatial
disaggregation errors of NIEMO stemming from USIO. (10.)

First, based on the assumptions of equations (8.), the total output vector of the perfect
IRIO,
Pi
X (where N i = or I ), can be estimated as,

j
PI
j
I
j
PI j
PI_I
x
x x
α
−
= (11.1)

=
j
PI
j
I
x
x
− 1 , (11.2)

and by assumption of equation (7.1),
117

j
I PI
α
_
=
j
UN N
ε
_
=
j
N
j
UN
x
x
− 1 . (12.)

Hence,

j
PI
x =
j
UN
j
I
j
N
x
x x
. (13.)

Similarly, based on the assumption of equation (7.2),

j
PN
x =
j
UI
j
I
j
N
x
x x
. (14.)

Further, vector ε s are calculated from the relation of the models, two estimated error
vectors
I PI _
ˆ α and
N PN _
ˆ α can be obtained as,

I PI _
ˆ α =
UN N _
ε , and (15.1)

N PN _
ˆ α =
UI I _
ε . (15.2)

Based on the defined vectors from equations (15.), the rest are estimated with equations
(13.) and (14.) showing that
UN PN _
ˆ α =
UI PI _
ˆ α
U P _
ˆ α ≡ , as shown below.
118

j
UN PN
α
_
ˆ =
j
PN
j
UN
x
x
− 1 (16.1)

=
j
I
j
N
j
UN
j
UI
x x
x x
− 1 (16.2)

=
j
N
j
I
j
UI
j
UN
x x
x x
− 1 (16.3)

=
j
UI PI
α
_
ˆ (16.4)

≡
j
U P
α
_
ˆ (16.5)

Although the vector
N I _
β can be defined as,

j
I
j
N
j
I
j
N
j
I j
I_N
x
x
x
x x
β − =
−
= 1 , (17.)

if
j
I
x and
j
N
x are total outputs for element j with the same inputs from national IMPLAN
IO model and NIEMO, respectively, equation (17.) would not be available because different
inputs for the two models are used in this study.

119

Nevertheless, because the USIO model is based on the same data as both IO models,
j
UI
j
I
j
UN
j
N
x x
x x
/
/
can be substituted for the
j
I
j
N
x
x
shown in equation (17.), where
j
N
x and
j
I
x are
adjusted to the same scales of values based on USIO total outputs. Therefore,

j
N I
β
_
ˆ
=
j
UI
j
I
j
UN
j
N
x x
x x
/
/
1 − (18.1)

=
) 1 (
) 1 (
1
_
_
j
I UI
j
N UN
ε
ε
−
−
− (18.2)

= ) 1 )( 1 ( 1
_ _
j
UI I
j
N UN
ε ε − − − (18.3)

=
j
UI I
ε
_
+
j
N UN
ε
_
-
j
UI I
ε
_
*
j
N UN
ε
_
(18.4)

=
j
UI I
ε
_
+
j
N UN _
ε -
j
N I _
Ξ (18.5)

where,
j
N I _
Ξ ≡
j
UI I _
ε *
j
N UN _
ε , indicates the degree of j sector accuracy in the model.

Further, instead of BEA GSP (Gross State Product) data, which are some of the most
credible data, IMPLAN value added data were used as the exogenous inputs for two reasons.
First, IMPLAN data reveal greater accuracy when the 51 areas are aggregated to the U.S.
120

than the BEA GSP data, as shown in Table A.IV-1 of Appendix C. Also, The BEA GSP data
do not include total output for each state and therefore NIEMO cannot input the value added
vector for every state. In Table A.IV-2 of Appendix C, there is a comparison table of total
output between IMPLAN and BEA data and between the sum of states for IMPLAN and
national IMPLAN for 47 USC Sectors.

4. Results
Figure IV-3 shows the overall estimated and calculated errors resulting from various
types of IO models matched to the errors suggested in Figure IV-2. I find that the vector of
N I _
ˆ
β (column 8 of Table IV-4) denotes an overall error of 5.2 percent. Yet, this error is
difficult to put into perspective because it is not known whether IMPLAN or NIEMO
provides the better estimate of the ideal model. My approach suggests an answer in columns
4 and 5 of Table IV-4, which reveal that spatial aggregation introduces an overall larger error
(=3.87 percent) than sectoral aggregation (=-1.13 percent). Tables IV-2 and IV-3 include the
corresponding absolute values of errors which have the same relationship, respectively 3.45
as sectoral aggregation errors and 6.31 percent as spatial aggregation errors. This indicates
spatial aggregation might produce more severe aggregation problems, that is, the cost of
choosing the IMPLAN national model is higher than that to use NIEMO if the perfect IRIO
is accurate.

121

Figure IV-3 Estimated Errors Resulting from Various Types of IO Models
% 20 . 5
ˆ
_
− =
N I
β
% 78 . 2 ˆ ˆ ˆ
_ _ _
= ≡ =
U P UN PN UI PI
α α α
USIO
(1x47)
2
Newly constructed
NIEMO
(52x47)
2
Newly constructed
IMPLAN IO
(1x 509)
2
Available from
IMPLAN program
Perfect IRIO
(52x509)
2
Unknown
% 13 . 1
_
− =
UI I
ε
% 13 . 1 ˆ
_
− =
N PN
α
% 02 . 4
_
− =
N UN
ε
% 87 . 3
ˆ
_
=
I PI
α
% 87 . 3
_
=
UN N
ε
% 31 . 6
_
=
UN N
ε
% 45 . 3
_
=
UI I
ε

Note: The hat on the vectors α and β indicates estimates of the errors.

122

Table IV-2 Sectoral aggregation errors between 47 USIO and 509 IMPLAN IO (units of X s: $)
Total Impacts Sectoral Aggregation Errors
USC
Sectors
Direct
Impacts
*
I
X
**

UI
X
UI I
X X −
UI I _
ε
UI I
X X −
UI I _
ε
USC 1 8.25 12.25 14.94 -2.69 -21.98% 2.69 21.98%
USC 2 10.50 18.67 16.70 1.97 10.57% 1.97 10.57%
USC 3 2.25 4.80 5.61 -0.81 -16.82% 0.81 16.82%
USC 4 10.33 11.66 11.68 -0.02 -0.16% 0.02 0.16%
USC 5 22.08 29.87 31.38 -1.51 -5.06% 1.51 5.06%
USC 6 4.00 6.43 6.47 -0.04 -0.59% 0.04 0.59%
USC 7 3.00 3.22 3.11 0.12 3.60% 0.12 3.60%
USC 8 3.33 5.56 5.39 0.17 3.03% 0.17 3.03%
USC 9 3.33 5.45 5.24 0.21 3.80% 0.21 3.80%
USC 10 6.50 25.08 24.59 0.49 1.95% 0.49 1.95%
USC 11 6.33 11.72 11.25 0.47 4.00% 0.47 4.00%
USC 12 2.00 2.47 2.57 -0.10 -4.02% 0.10 4.02%
USC 13 2.50 3.97 3.67 0.30 7.57% 0.30 7.57%
USC 14 11.33 19.25 18.91 0.35 1.81% 0.35 1.81%
USC 15 14.50 26.52 26.53 0.00 -0.01% 0.00 0.01%
USC 16 13.00 22.63 23.27 -0.64 -2.84% 0.64 2.84%
USC 17 9.83 21.77 20.91 0.86 3.96% 0.86 3.96%
USC 18 7.00 13.30 13.52 -0.22 -1.69% 0.22 1.69%
USC 19 22.58 31.21 31.42 -0.21 -0.67% 0.21 0.67%
USC 20 21.17 25.83 26.36 -0.53 -2.05% 0.53 2.05%
USC 21 19.50 37.14 36.02 1.12 3.00% 1.12 3.00%
USC 22 25.75 38.86 39.68 -0.83 -2.13% 0.83 2.13%
USC 23 50.83 62.09 62.37 -0.28 -0.45% 0.28 0.45%
USC 24 33.58 46.97 49.12 -2.15 -4.58% 2.15 4.58%
USC 25 10.83 16.22 18.38 -2.16 -13.30% 2.16 13.30%
USC 26 5.00 5.83 5.83 0.00 0.02% 0.00 0.02%
USC 27 12.50 13.50 13.66 -0.16 -1.18% 0.16 1.18%
USC 28 12.00 12.80 12.75 0.05 0.42% 0.05 0.42%
USC 29 20.17 23.34 23.33 0.00 0.02% 0.00 0.02%
USC 30 3.00 13.79 12.95 0.84 6.09% 0.84 6.09%
USC 31 13.00 16.45 16.14 0.31 1.87% 0.31 1.87%
USC 32 1.00 42.77 42.62 0.15 0.35% 0.15 0.35%
USC 33 7.00 28.49 27.03 1.46 5.13% 1.46 5.13%
USC 34 3.00 9.53 9.39 0.14 1.42% 0.14 1.42%
USC 35 12.00 16.41 17.03 -0.62 -3.77% 0.62 3.77%
USC 36 6.00 19.31 20.06 -0.75 -3.87% 0.75 3.87%
USC 37 6.00 25.48 27.29 -1.81 -7.11% 1.81 7.11%
USC 38 7.00 33.48 33.96 -0.48 -1.42% 0.48 1.42%
USC 39 14.00 41.73 41.46 0.27 0.65% 0.27 0.65%
USC 40 1.00 14.39 13.91 0.48 3.36% 0.48 3.36%
USC 41 9.00 22.25 22.84 -0.58 -2.62% 0.58 2.62%
USC 42 2.00 2.24 2.28 -0.05 -2.05% 0.05 2.05%
USC 43 8.00 9.56 9.25 0.31 3.27% 0.31 3.27%
USC 44 8.00 10.74 10.13 0.61 5.67% 0.61 5.67%
USC 45 3.00 7.11 7.38 -0.27 -3.83% 0.27 3.83%
USC 46 9.00 8.94 12.60 -3.66 -40.98% 3.66 40.98%
USC 47 18.00 39.99 40.52 -0.53 -1.32% 0.53 1.32%
TOTAL 504.00 921 931 -10.42 -1.13% 31.78 3.45%
* 509 IMPLAN IO model actually uses $1 inputs for 504 sectors, excluding 5 sectors, or 501-503 and 507-508,
due to the nonexistence of coefficients for those sectors
** Results of IMPLAN IO are aggregated to 47 USC Sectors, based on the bridge linking IMPLAN and USC
after being calculated from 509-sector IMPLAN IO models
123

Also, based on these aggregation errors, the perfect IRIO vector
PI
X and
PN
X can be
estimated easily, as shown in column 2 and column 3 of Table IV-4. The estimates for total
sum value of the vector
PI
X (=
∑
j
j
PI
x ), and that of
PN
X (=
∑
j
j
PN
x ) are $958 and $4,469
respectively, showing 3.87 percent for the overall aggregation error between IRIO and
IMPLAN IO and –1.13 percent for that between IRIO and NIEMO. Therefore, NIEMO is
better at reflecting the perfect IRIO than the IMPLAN IO, with smaller exaggeration, by
1.13 percent (column 2 of Table IV-4). The aggregation errors from IRIO to USIO (=
U P _
ˆ α )
are close to the sum of
N PN _
ˆ α and
UN N _
ε or the sum of
I PI _
ˆ α and
UI I _
ε , or
N PN _
ˆ α +
I PI _
ˆ α , because it is derived as their products. The following discussion elaborates the
approach and findings further.

4.1. NIEMO Benefits from Spatial Information:
N UN _
ε
Figure IV-2 highlighted the fact that disaggregation errors
N UN _
ε add spatial
information to the sectorally aggregated IO model (USIO), making the simple one-region
USIO model spatially detailed. Because the simple USIO model has no detailed information
for states or subregions, the overall estimates are smaller than the spatially distributed
NIEMO, although the simple USIO model may have the possibility of under- (36 sectors) or
over- (11 sectors) estimating the various NIEMO sectors.
When total outputs via NIEMO and via USIO are compared in Table IV-3 and IV-4, for
instance, USIO underestimates $174.71 overall, showing 3.87 percent as total aggregation
error from NIEMO and -4.02 percent as total disaggregation error for NIEMO, based on the
124

same inputs. Therefore, the total disaggregation error of
N UN _
ε indicates the overall
percentage of spatial information required to adjust a simplified one-region model and also
that NIEMO adds benefits by 4.02 percent from disaggregating spatially.

125

Table IV-3 Spatial aggregation errors between NIEMO and USIO (units of X s: $)
Total Impacts Spatial Aggregation Errors
USC
Sectors
Direct
Impacts
*

N
X
**

UN
X
UN N
X X −
UN N _
ε
UN N
X X −
UN N _
ε
USC 1 51.00 84.70 86.40 -1.70 -2.00% 1.70 2.00%
USC 2 51.00 83.45 85.26 -1.80 -2.16% 1.80 2.16%
USC 3 51.00 70.92 70.77 0.15 0.21% 0.15 0.21%
USC 4 51.00 59.19 58.95 0.25 0.41% 0.25 0.41%
USC 5 51.00 99.87 100.57 -0.70 -0.70% 0.70 0.70%
USC 6 51.00 65.84 63.29 2.55 3.87% 2.55 3.87%
USC 7 51.00 52.34 52.50 -0.15 -0.29% 0.15 0.29%
USC 8 51.00 63.14 60.35 2.78 4.41% 2.78 4.41%
USC 9 51.00 62.26 62.07 0.19 0.31% 0.19 0.31%
USC 10 51.00 180.71 161.33 19.38 10.73% 19.38 10.73%
USC 11 51.00 87.16 76.38 10.78 12.37% 10.78 12.37%
USC 12 51.00 57.97 57.82 0.15 0.25% 0.15 0.25%
USC 13 51.00 61.12 59.19 1.93 3.17% 1.93 3.17%
USC 14 51.00 85.84 82.34 3.50 4.08% 3.50 4.08%
USC 15 51.00 103.78 99.83 3.95 3.81% 3.95 3.81%
USC 16 51.00 99.09 93.54 5.54 5.60% 5.54 5.60%
USC 17 51.00 104.22 102.30 1.92 1.84% 1.92 1.84%
USC 18 51.00 67.47 85.51 -18.05 -26.75% 18.05 26.75%
USC 19 51.00 72.29 78.44 -6.15 -8.51% 6.15 8.51%
USC 20 51.00 72.22 70.17 2.05 2.84% 2.05 2.84%
USC 21 51.00 111.02 93.95 17.07 15.38% 17.07 15.38%
USC 22 51.00 98.22 95.14 3.08 3.14% 3.08 3.14%
USC 23 51.00 93.71 89.73 3.97 4.24% 3.97 4.24%
USC 24 51.00 118.55 101.64 16.91 14.26% 16.91 14.26%
USC 25 51.00 78.71 81.59 -2.87 -3.65% 2.87 3.65%
USC 26 51.00 56.00 56.53 -0.53 -0.94% 0.53 0.94%
USC 27 51.00 56.99 56.56 0.43 0.76% 0.43 0.76%
USC 28 51.00 53.41 54.27 -0.86 -1.61% 0.86 1.61%
USC 29 51.00 79.69 61.69 17.99 22.58% 17.99 22.58%
USC 30 51.00 107.31 97.23 10.09 9.40% 10.09 9.40%
USC 31 51.00 69.69 67.98 1.72 2.47% 1.72 2.47%
USC 32 51.00 227.75 218.81 8.95 3.93% 8.95 3.93%
USC 33 51.00 155.65 144.63 11.02 7.08% 11.02 7.08%
USC 34 51.00 84.00 82.15 1.85 2.20% 1.85 2.20%
USC 35 51.00 76.02 74.29 1.72 2.27% 1.72 2.27%
USC 36 51.00 111.89 122.17 -10.28 -9.19% 10.28 9.19%
USC 37 51.00 165.11 159.25 5.86 3.55% 5.86 3.55%
USC 38 51.00 213.64 198.72 14.92 6.98% 14.92 6.98%
USC 39 51.00 225.20 184.29 40.91 18.17% 40.91 18.17%
USC 40 51.00 110.82 109.70 1.12 1.01% 1.12 1.01%
USC 41 51.00 121.02 117.62 3.40 2.81% 3.40 2.81%
USC 42 51.00 53.92 52.56 1.36 2.52% 1.36 2.52%
USC 43 51.00 62.94 60.64 2.29 3.65% 2.29 3.65%
USC 44 51.00 63.09 62.77 0.32 0.51% 0.32 0.51%
USC 45 51.00 72.63 71.27 1.36 1.87% 1.36 1.87%
USC 46 51.00 56.08 68.26 -12.18 -21.72% 12.18 21.72%
USC 47 51.00 162.50 153.99 8.51 5.24% 8.51 5.24%
TOTAL 2397.00 4,519 4,344 174.71 3.87% 285.25 6.31%
*NIEMO actually uses $1 inputs for 51 states and 47 sectors, then aggregates the inputs by USC Sector, with $51
excluding rest of world.
**Results of NIEMO aggregate those of 51 states by USC Sector.
126

4.2. NIEMO costs:
N I _
ˆ
β
The estimated vector
N I _
ˆ
β (47x1) includes mixed information: benefits from
information gained from disaggregating spatially (
N UN _
ε ) vs. costs from sectoral
aggregations (
UI I _
ε ). Via the definitions and calculations of various errors ( ε ) and IO
models shown in Figure IV-2, vector
N I _
ˆ
β includes the sum of two errors (
UI I _
ε
and
N UN _
ε ) adjusted by their product. Although there might be spatial aggregation errors in
the IMPLAN U.S. national IO model, compared to the perfect IRIO-type model with 52
regions, IMPLAN has been widely used for more than twenty years to in various U.S.
impact studies. The vector
N I _
ˆ
β indicates, in this sense, how close the results of NIEMO
and IMPLAN U.S. IO are, even though NIEMO, derived from state-level coefficients of
IMPLAN IO models, simultaneously includes both information to improve model accuracy
and errors to aggravate this. This means that if the overall difference of
N I _
ˆ
β , is large
enough, then a spatially disaggregated model such as NIEMO might be an unacceptable
model.
A good indicator to reflect the credibility of NIEMO, therefore, can be the absolute
values of total percentage errors shown in Table IV-4. Because the results reveal that the
absolute overall difference of
N I _
ˆ
β is 5.20 percent, I can obtain detailed information for
each state by USC Sector, with a loss vis-à-vis the widely accepted IMPLAN model of 5.2
percent, denoting the costs for adopting this MRIO-type model.
127

Furthermore, if NIEMO is true, then each sectoral error in vector
N I _
ˆ
β denotes that the
corresponding result for IMPLAN national IO can be adjusted by NIEMO by the amount of
j
N UN _
ε -
j
UI I _
ε *
j
N UN _
ε , which is the spatial dissaggregation of sector j and the model
accuracy term.

128

Table IV-4 Estimates of
PI
X and
PN
X , and various other errors (units of X s: $)
USC Sectors
PI
X
PN
X
N PN _
ˆ α
I PI _
ˆ α
U P _
ˆ α
N UN _
ε
N I _
ˆ
β
N I _
Ξ
USC 1 12.01 69.44 -21.98% -2.00% -24.43% 1.96% -19.59% -0.43%
USC 2 18.27 93.32 10.57% -2.16% 8.64% 2.11% 12.46% 0.22%
USC 3 4.81 60.71 -16.82% 0.21% -16.58% -0.21% -17.07% 0.04%
USC 4 11.71 59.10 -0.16% 0.41% 0.25% -0.42% -0.58% 0.00%
USC 5 29.66 95.06 -5.06% -0.70% -5.80% 0.70% -4.33% -0.04%
USC 6 6.69 65.45 -0.59% 3.87% 3.30% -4.02% -4.64% 0.02%
USC 7 3.21 54.30 3.60% -0.29% 3.32% 0.29% 3.88% 0.01%
USC 8 5.82 65.11 3.03% 4.41% 7.31% -4.61% -1.44% -0.14%
USC 9 5.47 64.72 3.80% 0.31% 4.10% -0.31% 3.49% -0.01%
USC 10 28.10 184.30 1.95% 10.73% 12.46% -12.01% -9.83% -0.23%
USC 11 13.38 90.79 4.00% 12.37% 15.88% -14.11% -9.54% -0.56%
USC 12 2.47 55.73 -4.02% 0.25% -3.76% -0.25% -4.28% 0.01%
USC 13 4.10 66.12 7.57% 3.17% 10.49% -3.27% 4.54% -0.25%
USC 14 20.07 87.43 1.81% 4.08% 5.82% -4.25% -2.36% -0.08%
USC 15 27.57 103.77 -0.01% 3.81% 3.80% -3.96% -3.97% 0.00%
USC 16 23.97 96.35 -2.84% 5.60% 2.92% -5.93% -8.93% 0.17%
USC 17 22.18 108.52 3.96% 1.84% 5.73% -1.87% 2.17% -0.07%
USC 18 10.49 66.35 -1.69% -26.75% -28.89% 21.10% 19.77% -0.36%
USC 19 28.76 71.80 -0.67% -8.51% -9.24% 7.84% 7.22% -0.05%
USC 20 26.59 70.77 -2.05% 2.84% 0.85% -2.93% -5.04% 0.06%
USC 21 43.89 114.46 3.00% 15.38% 17.92% -18.18% -14.63% -0.55%
USC 22 40.11 96.18 -2.13% 3.14% 1.07% -3.24% -5.44% 0.07%
USC 23 64.83 93.28 -0.45% 4.24% 3.81% -4.43% -4.90% 0.02%
USC 24 54.79 113.36 -4.58% 14.26% 10.34% -16.64% -21.98% 0.76%
USC 25 15.65 69.47 -13.30% -3.65% -17.44% 3.52% -9.32% -0.47%
USC 26 5.77 56.01 0.02% -0.94% -0.93% 0.93% 0.95% 0.00%
USC 27 13.60 56.32 -1.18% 0.76% -0.41% -0.77% -1.96% 0.01%
USC 28 12.60 53.64 0.42% -1.61% -1.19% 1.58% 1.99% 0.01%
USC 29 30.15 79.70 0.02% 22.58% 22.59% -29.17% -29.14% -0.01%
USC 30 15.22 114.27 6.09% 9.40% 14.91% -10.37% -3.66% -0.63%
USC 31 16.87 71.02 1.87% 2.47% 4.29% -2.53% -0.61% -0.05%
USC 32 44.52 228.55 0.35% 3.93% 4.26% -4.09% -3.73% -0.01%
USC 33 30.66 164.07 5.13% 7.08% 11.85% -7.62% -2.09% -0.39%
USC 34 9.74 85.21 1.42% 2.20% 3.59% -2.25% -0.79% -0.03%
USC 35 16.79 73.26 -3.77% 2.27% -1.42% -2.32% -6.17% 0.09%
USC 36 17.69 107.72 -3.87% -9.19% -13.42% 8.42% 4.87% -0.33%
USC 37 26.42 154.16 -7.11% 3.55% -3.30% -3.68% -11.05% 0.26%
USC 38 36.00 210.64 -1.42% 6.98% 5.66% -7.51% -9.04% 0.11%
USC 39 50.99 226.66 0.65% 18.17% 18.69% -22.20% -21.41% -0.14%
USC 40 14.54 114.67 3.36% 1.01% 4.33% -1.02% 2.37% -0.03%
USC 41 22.90 117.92 -2.62% 2.81% 0.26% -2.89% -5.59% 0.08%
USC 42 2.30 52.84 -2.05% 2.52% 0.53% -2.59% -4.69% 0.05%
USC 43 9.92 65.07 3.27% 3.65% 6.80% -3.78% -0.39% -0.12%
USC 44 10.79 66.88 5.67% 0.51% 6.16% -0.51% 5.19% -0.03%
USC 45 7.24 69.95 -3.83% 1.87% -1.89% -1.90% -5.80% 0.07%
USC 46 7.34 39.78 -40.98% -21.72% -71.59% 17.84% -15.83% -7.31%
USC 47 42.20 160.38 -1.32% 5.24% 3.99% -5.53% -6.92% 0.07%
TOTAL 958 4,469 -1.13% 3.87% 2.78% -4.02% -5.20% 0.05%
Note: The hat on the vector α s and β indicates the estimates of errors.

129

4.3. Model Accuracy:
N I _
Ξ
Finally, model accuracy can also be represented by the product of two errors,
or
j
UI I _
ε *
j
N UN _
ε defined as
j
N I _
Ξ . In the previous discussion, I noted that
N I _
ˆ
β might be
equal to the sum of
UI I _
ε and
N UN _
ε . However, if both errors are exceptionally large, then
the sector or model itself will be far different from the normal sums due to the error product.
Therefore, the product can be used as another indicator to monitor model accuracy, that is,
whether the case of mixed aggregation contains severely distorted sectors (or their
aggregate) or not. In the calculated results, there are only minor distortions because all
j
N I _
Ξ are smaller than 1 percent in absolute value, except the USC 46 (Public
Administration) Sector.

5. Conclusions
This study, based on the newly constructed NIEMO and USIO models, reports on tests
of the accuracy of the approach. I use as a benchmark the widely used IMPLAN IO model.
While the IMPLAN IO models have sectorally detailed information but lack spatially
disaggregated detail, NIEMO develops spatially detailed information at the cost of
aggregated industrial sectors. As a common reference model, I constructed USIO, which has
aggregated sectors corresponding to those used for NIEMO but only one aggregate national
area as does the IMPLAN national IO model. This was constructed to compare spatial and
sectoral aggregation errors. Also, with respect to spatial disaggregation errors, I tested
NIEMO’s sectoral aggregation cost in terms of model accuracy incurred by switching from
the IMPLAN IO model to NIEMO.
130

I have shown that a reasonably accurate multiregional input-output model for the U.S.
containing approximately six-million multipliers can be constructed at low cost given the
fact that IMPLAN’s input-output matrices are plausible. With respect to the estimation of
overall model accuracy, I found only relatively small errors when comparing the aggregates
of all sectors and also only minor errors on an individual sector-by-sector comparison basis.
I have also demonstrated that the sectoral aggregation required to go from IMPLAN to
NIEMO imparts only minor errors. I conclude that, all things considered, it is more useful to
construct an MRIO-type model instead of a general one-region IO model, especially for a
large region that includes many economically diverse subregions.
Nevertheless, NIEMO requires further work. The next steps in the research program
involve relaxing the assumption of no intestate trade in services and, instead, estimating
interstate trade flows for the 18 service sectors of the USC-Sector system. This may address
some of the underestimated impacts and spatial aggregation errors vis-à-vis the perfect IRIO
model. In this sense, the recent suggestions by Park (2006b) will help to estimate origin-
destination matrices for the service industries. This requires using the state-level domestic
imports and exports vector available for each state and applying Geographically Weighted
Regressions (GWR) along the lines suggested by LeSage (1999).
131

VIII. CHAPTER IV REFERENCES

Ara, K., 1959, The Aggregation Problem in Input-Output Analysis, Econometrica 27: 257-
262.

Blair, P. and R.E. Miller, 1983, Spatial Aggregation of Multiregional Input-Output Models,
Environmental Planning A 15: 187-206.

Chenery, H.B., 1953, Regional Analysis, in The Structure and Growth of the Italian
Economy, edited by H.B. Chenery, P.G. Clark and V.C. Pinna, U.S. Mutual Security
Agency, Rome: 98-139.

Dietzenbacher, E., 1992, Aggregation in Multisector Models: Using the Perron Vector,
Economic Systems Research 4: 3-24.

Gibbons, J.C., A. Wolsky, and G. Tolley, 1982, Approximate Aggregation and Error in
Input-Output Models, Resource and Energy 4: 203-230.

Hatanaka, M., 1952, Note on Consolidation within a Leontief System, Econometrica 20:
301-303.

Isard, W., 1951, Interregional and Regional Input-Output Analysis: A Model of a Space
Economy, Review of Economics and Statistics 33: 318-328.

Jack Faucett Associates, INC, 1983, The Multiregional Input-Output Accounts, 1977:
Introduction and Summary, Vol. I (Final Report), prepared for the US Department of
Health and Human Services, Washington.

Leontief, W., 1951, The Structure of American Economy, 1919-1939: an Empirical
Application of Equilibrium Analysis, Oxford Univ. Press, NY.

LeSage, J.P., 1999, The Theory and Practice of Spatial Econometrics, University of Toledo,
MN, which is available at the http://www.spatial-econometrics.com/html/sbook.pdf.

McManus, M., 1956, General Consistent Aggregation in Leontief Models, Yorkshire Bulletin
of Economic and Social Research 8: 28-48.

Miller, R.E. and G. Shao, 1990, Spatial and Sectoral Aggregation in the Commodity-
Industry Multiregional Input-Output Model, Environment and Planning A 22: 1637-
1656.

Miller, R.E. and P. Blair, 1981, Spatial Aggregation in Interregional Input-Output Models,
Papers of the Regional Science Association 48: 149-164.

132

Miller, R.E. and P. Blair, 1985, Input-Output Analysis: Foundations and Extensions, New
Jersey: Prentice-Hall.

Morimoto, Y., 1970, On Aggregation Problems in Input-Output Analysis, Review of
Economic Studies 37: 119-126.

Moses, L.N., 1955, The Stability of Interregional Trading Patterns and Input-Output
Analysis, American Economic Review, 45: 803-832.

Park, J.Y., 2006a, The Economic Impacts of a Dirty-Bomb Attack on the Los Angeles and
Long Beach Port: Applying Supply-driven NIEMO, Paper presented at 17
th
Annual
Meeting of the Association of Collegiate Schools of Planning, Fort Worth, TX, USA,
November 9-12.

Park, J.Y., 2006b, Estimation of State-by-State Trade Flows for Service Industries, Paper
presented at North American Meetings of the Regional Science Association International
53rd Annual Conference, Fairmont Royal York Hotel, Toronto, Canada, November 16-
18.

Park, J.Y., P. Gordon, J. E. Moore II, and H. W. Richardson, 2004, Construction of a U.S.
Multiregional Input Output Model Using IMPLAN, Paper presented at 2004 National
IMPLAN User’s Conference, Shepherdstown, West Virginia, USA, October 6-8.

Park, J.Y., P. Gordon, J.E. Moore II, H.W. Richardson, and L. Wang, 2007, Simulating the
State-by-State Effects of Terrorist Attacks on Three Major U.S. Ports: Applying NIEMO
(National Interstate Economic Model): 208-234, in H.W. Richardson, P. Gordon and J.E.
Moore II, eds., The Economic Costs and Consequences of Terrorism. Cheltenham:
Edward Elgar.

Park, J.Y., C.K. Park, and P. Gordon, 2007, The State-by-State Effects of Mad Cow Disease
using a new MRIO model, revised to resubmit to Journal of Agricultural and Resource
Economics.

Polenske, K.L., 1980, The US Multiregional Input-Output Accounts and Model, DC Health,
Lexington, MA.

Robison, W.S., 1950, Ecological Correlations and the Behavior of Individuals, American
Sociological Review 15: 351-357.

Theil, H., 1957, Linear Aggregation in Input-Output Analysis, Econometrica 25 (1): 111-
122.

133

IX. CHAPTER V: ESSAY FIVE

Constructing a Flexible National Interstate Economic Model (FlexNIEMO)

1. Introduction
There are two standard models of the classic input-output (IO) system. The first, the
Leontief, demand-driven IO model, follows Leontief’s early contributions (1936, 1941) on
how to generalize interdependences between industries in an economy. The second, the
Ghoshian, supply-driven IO model was introduced by Ghosh in 1958, and suggested an
alternative way to understand the interrelations between industries. Inter-industry linkages in
the demand-driven IO model account for technical relationships in an economy via
production functions. In contrast, the supply-driven model is less transparent, suggesting
monopolistic markets or a centralized, planned market in which all resources are scarce
except for one, and considers the best use of this non-scarce input in combination with
scarce resources. This best use is derived from a known social welfare function (Ghosh,
1964).
Spatial extensions of the classic IO model include interregional or multiregional input
output (IRIO or MRIO) models (Isard, 1951; Chenery, 1953; Moses, 1955); and empirical
models of the U.S. constructed in the late 1970s (Polenske, 1980) and early 1980s (Jack
Faucett Associates, 1983). Recently, Park et al. (2007) constructed a new demand-driven
MRIO model, the National Interstate Economic Model (NIEMO). Park (2006) and Park et al.
(2006a) elaborated a supply-driven version of NIEMO, including empirical tests.
134

As many have noted, a key limitation of IO models is that the coefficients in the models
are fixed, and the models ignore substitution opportunities that should be prompted by
market signals. Gordon et al (2006) suggested an approach to constructing new IO
coefficients that captures substitution effects actually experienced in the labor sector. This
new approach builds on both the demand- and supply- driven models. Their flexible
approach relaxed the assumption of fixed coefficients in IO models by applying the RAS
method to adjust coefficient matrices to account for empirical changes in value added and
final demand. They demonstrated their approach via an example consisting of a two-by-two
matrix of intersectoral flows. An important implication of this approach is that it can be
applied to extend the classic IO model, making the standard model a useful tool for studying
economic resiliency. The coefficients in the resulting IO model can be adjusted across time
periods to account for substitution-driven adjustments resulting from exogenous events such
a natural disaster or a terrorist attack. This study suggests an approach for constructing such
a resiliency IO model that reflects actual substitution effects.
The rest of this essay deals with constructing a Flexible IO model (FlexIO) without
spatial disaggregation. Then, the approach is extended to the current National Interstate
Economic Model (NIEMO), a multi-regional input output (MRIO) model that has constant
coefficients. The result is a Flexible NIEMO (FlexNIEMO). An example is shown
comparing the results provided by FlexNIEMO and by NIEMO. Conclusions and remarks
follow.

135

2. Flexible Input Output Models (FlexIO)
This section describes construction of a Flexible Input Output (FlexIO) model based on
the classic demand- and supply-driven input output models. It is helpful to define the matrix
of inter-industry flows to include the final demand and value added sectors. Table V-1 shows
such a national, expanded set of economic transaction flows. Matrix notation appears in
parentheses.

Figure V-1 Expanded Matrix of National Economic Transaction Flows
NN N
ij
N
z z
z
z
z z z
1
21
1 12 11
O
M
O
L
NK N
ik
K
f f
f
f
f f f
1
21
1 12 11
O
M
O
L
) ) ( (
) ) ( (
) ) ( (
2 2 2
1 1 1
∑∑
∑∑
∑ ∑
+ =
+ =
+ =
jk
Nk Nj
d
N
jk
k j
d
jk
k j
d
f z x
f z x
f z x
M
LN L
lj
N
v v
v
v
v v v
1
21
1 12 11
O
M
O
L

-
) (
) (
) (
2 2
1 1
∑
∑
∑
=
=
=
j
Lj L
j
j
j
j
v v
v v
v v
M
s
N
s s
x x x L
2 1

K
f f f L
2 1

-
Notes: Z is an NxN matrix of intermediate interindustry flows. Its elements
ij
z denote
deliveries of dollar values from industry sector i to j .
( Z ) ( F )
(
d
X )
( V )
( V )
(
s
X )
( F )
136

F is an NxK matrix showing various kinds of final demands. Its elements
ik
y
denote deliveries of dollar values from industry sector i to final users k .
Generally, k includes private consumers, governments, investments, and exports.
V is an LxN matrix showing various kinds of value added factors. Its elements
lj
v denote the dollar values added to product sector j by factor inputs l . Generally,
l contains various kinds of labor, capital, taxes by government, and imports.
d
X is the monetary value column vector of total outputs for each sector. Its
elements
d
i
x are the column sums of intermediate flows from and final demands
for sector i .
s
X is the monetary value row vector of total inputs for each sector. Its elements
i
j
x are the row sums of intermediate flows into and value added factors for sector
j ,

∑∑
+ =
i
lj
l
ij
s
j
v z x ) ( . (1.)

V is a column vector consisting of the column sum across l value added factors.

V =
N
Vu (2.)

where
T
N
u is the N -element unit row vector, i.e., (1, …, 1).

137

F is a row vector consisting of the row sum across k final demands. Its elements

∑
=
i
ik k
f f , (3.)

and

F = F u
T
N
. (4.)

In the general IO model, it is assumed that

T
s d
X X = . (5.)

Given the flows in Table V-1, consider an IO matrix that is open with respect to the
household sector. The input coefficient matrix
0
A for some initial time period capturing the
backwards linkages between industries is defined as

0
A =
1 d
) X Z(
−
ˆ
=
1 s
) X Z(
−
ˆ
. (6.)

where
s
X
ˆ
denotes the diagonalization of the vector
s
X into a square matrix, and
hence

138

d
T
s s
X X X
ˆ ˆ ˆ
= = . (7.)

The standard Leontief inverse matrix
1
0
) A (I
−
− is obtained via equations (8.),

d
X =
N
Zu +
k
Fu . (8.1)

T
s
X =
N
T
s
0
u X A
ˆ
+
k
Fu (8.2)

=
T
s
0
X A +
k
Fu (8.3)

=
k
1
0
Fu ) A (I
−
− (8.4)

Obtaining the corresponding Ghoshian supply-driven model requires a matrix of
allocation (output) coefficients, B , that allocates the sales of total inputs to each sector. Here,
the allocation matrix for an initial time period,
0
B , that captures forward linkages between
industries defined as

0
B = Z ) X (
1 s −
ˆ
= Z ) X (
1 d −
ˆ
(9.)

The Ghoshian inverse allocation matrix is obtained via equations (10.).

139

s
X = Z u
T
N
+ V u
T
l
(10.1)

T
d
X = Z u
T
N
+ V u
T
l
(10.2)

=
0
T
d T
N
B X u
ˆ
+ V u
T
l
(10.3)

=
0
T
d
B X + V u
T
l
(10.4)

=
1
0
T
l
) B V(I u
−
− (10.5)

Note that

s
0
1 s d
0
1 d
0
X A ) X ( X A ) X ( B
ˆ ˆ ˆ ˆ
− −
= = , (11.)

because

d
0 0
d
X A B X Z
ˆ ˆ
= = . (12.)

Table V-1 shows provides a matrix that shows the normal, pre-event economy in period
0
t . An exogenous shock such a natural disaster or a terrorist attack is imposed on the
economy. As a result of this shock, changes in production of economic outputs occur week-
140

to-week or month-to-month. Total output changes can be obtained from monthly economic
data. These changes capture a number of adjustments. Given that data is available, a two-
period data set for impacted industries can be prepared. Designate the total output for an
industry interest (such as the oil industry) just prior to the shock as the “nominal
output,”
s(d)
t
X
0
, and total output of the industry one month after the shock as the “decreased
output,”
s(d)
t
X
1
.
I assume that the relationship between value added and total input and between final
demand and total output is unchanged across all periods. The observed, pre-event value
added and final demand vectors can be calculated as

V
0
t
=
s
t
X
0
s
X
R
ˆ
, (13.)

where the proportionate row vector relating valued added to total input defined as

s
X
R = V
1 s
) X (
−
ˆ
; (14.)

and

F
0
t
=
d
X
R
ˆ
d
t
X
0
, (15.)

where the proportionate column vector relating final demand to total output defined as

141

d
X
R =
1 d
) X (
−
ˆ
F . (16.)

Similarly, new post-event final demand and value added vectors for the first month of
the post-event period are computed by combining the decreased total output matrix and the
vectors of constant proportions,
s
X
R
ˆ
and
d
X
R
ˆ
,

F
t
1
=
d
X
R
ˆ
d
t
X
1
(17.1)

V
1
t
=
s
t
X
1
s
X
R
ˆ
(17.2)

The reductions in final demand and value added during the first month following the
shock are calculated, respectively, as

F ∆
1
t
= F
1
t
- F
0
t
(18.1)

V ∆
1
t
= V
1
t
- V
0
t
(18.2)

Given the changes F ∆
1
t
and V ∆
1
t
in equations (18.), the total output and total input
vectors
d
t
DX
1
and
s
t
DX
1
for the first post-event period
1
t are obtained via equations (19.)
and (20.),

142

d
t
DX
1
=
d
t
X
0
+
d
t
DX ∆
1
(19.1)

=
d
t
X
0
+
T 1
0 t
} ) B V(I { ∆
1
−
− , (19.2)

and

s
t
DX
1
=
s
t
X
0
+
s
t
DX ∆
1
(20.1)

=
s
t
X
0
+
T
t
1
0
F} ∆ ) A {(I
1
−
− . (20.2)

The updated final demand vector DF
1
t
and updated value added vector DV
1
t
are
calculated as in equations (17.), using the vectors of constant proportions
s
X
R
ˆ
and
d
X
R
ˆ
,

DF
1
t
=
d
X
R
ˆ
d
t
DX
1
(21.1)

DV
1
t
=
s
t
DX
1
s
X
R
ˆ
. (21.2)

The standard RAS technique (Hewings, 1985; Miller and Blair, 1985) can be applied to
created new coefficient matrices for first post-event period
1
t based on the four vectors F
1
t
,
V
1
t
, DF
1
t
, and DV
1
t
. First, the diagonal matrix R(1) for the period
1
t is calculated as
143

R(1) = ] F X [
1 1
t
d
t
ˆ ˆ
−
1
^
t
d
t ] DF DX [
1 1
−
−
^
(22.1)

= [] ] R [I X
d
1 X
d
t
ˆ ˆ
−
1
X
^
d
t ] R [I DX
d 1
−
⎥
⎦
⎤
⎢
⎣
⎡
−
ˆ
(22.2)

= ] X [
d
t
1
ˆ
1 d
t ] DX [
1
−
^
(22.3)

= []
d
1 X
d
t
R X
ˆ ˆ
1
X
d
t d 1
R DX
−
⎥
⎦
⎤
⎢
⎣
⎡
ˆ
^
(22.4)

= ] F [
1
t
ˆ
1
^
t ] DF [
1
−
. (22.5)

Similarly, the diagonal matrixS(1) for period
1
t is calculated as

S(1) = ] V [
1
t
ˆ
1
^
t ] DV [
1
−
(23.)

Updated technical coefficient matrices
1
A and
1
B for the demand-side and supply-side
models for period
1
t are obtained as

144

1
A = R(1)
0
A S(1) (24.)

and

1
B = S(1)
0
B R(1) (25.)

These updated matrices include the substitution effects actually occurring during the
first post-event month.
This approach makes it possible to estimate total impacts due to decreases in value
added flows following an exogenous shock. These impacts can be calculated relative to
the pre-event conditions for any month for which post-event data is available. Ideally,
these impacts should be calculated relative to the economy’s expected state had no
exogenous shock occurred. Such estimates of normal economic status can be produced via
a time-series approach (Park et al., 2006; 2007; Gordon et al., 2007; Richardson et al.,
2007). The difference between this estimated normal value added and the estimate of
decreased value added the exogenous direct impact for the supply-side IO model. The total
direct and indirect impacts in period
1
t ,
d
t
X ∆
1
~
, are

d
t1
X ∆
~
=
T 1
1 t1
} ) B (I V { ∆
−
−
~
(26.)

where
d
t1
X ∆
~
is the total direct and indirect impacts in period
1
t , and
145

V ∆
t1
~
is the difference between the estimate of normal valued added and the
estimate of decreased value added in period
1
t .

Similarly,
s
t1
X ∆
~
is estimated based on F ∆
t1
~
. Equations (6.) through (26.) provide
demand-side and supply-side coefficient matrices
i
A and
i
B , respectively, for each
period i. These reflect the equilibrium of demand and supply for each period. These
matrices are used to estimate total direct and indirect impacts (induced impacts if the IO
model is the closed version) in each period i based on the exogenous changes V ∆
i
t
~
or
F ∆
i
t
~
. Generally, actual total output would be a V-shaped function over time. After a short
period of rapid losses following a natural disaster or terrorist attack, losses which may
continue for more than one period, total output begins to increase toward pre-event levels.
Numerous trade-offs and substitutions are being made during this period, and this
approach captures these changes.

Simple Example

The following simple, two-by-two example of an economy with two economic sectors
illustrates the flexible IO model. Consider Miller and Blair’s (1985, p. 15) simplified 2x2
example. The 2x2 inter-industry flow matrix

Z =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
100 200
500 150
. (27.)
146

The total output vector

T
s d
X X = =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
2000
1000
. (28.)

The total value added row vector

T
V =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
1400
650
. (29.)

The total final demand column vector

F =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
1700
350
. (30.)

From the given vectors and matrices and equation (6.), the demand-side coefficients
matrix for the pre-event economy is

0
A =
1 d
) X Z(
−
ˆ
=
1 s
) X Z(
−
ˆ
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
05 . 0 20 . 0
25 . 0 15 . 0
, (31.)

and hence

147

1
0
) A (I
−
− =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
122 . 1 264 . 0
330 . 0 254 . 1
. (32.)

Similarly, from equation (8.), matrix

0
B =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
05 . 0 10 . 0
50 . 0 15 . 0
(33.)

and

1
0
) B (I
−
− =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
122 . 1 132 . 0
660 . 0 254 . 1
. (34.)

Scaling to some time period of interest, set the total output vector for the pre-event
economy as 10 percent of the given vector. Consequently,

d
t
X
0
ˆ
=
s
t
X
0
ˆ
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
200 0
0 100
. (35.)

From equations (13.) and (15.), the vector

V
0
t
=() 140 65 (36.)

148

and the vector

F
0
t
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
170
35
. (37.)
Due to some exogenous shock, the total output only for the second sector decreases from
200 to 130; and hence

d
t
X
1
ˆ
=
s
t
X
1
ˆ
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
130 0
0 100
. (38.)

From equations (17.),

F
1
t
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
111
35
(39.)

and

V
1
t
=() 91 65 . (40.)

Equations (18.) give direct losses for final demand,

F ∆
1
t
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
− 60
0
, (41.)
149

and direct losses for value added

V ∆
1
t
=() 49 0 − . (42.)

These values are used in equations (19.) and (20.) to update the total output and total
input vectors for the first post-event month,

d
t
DX
1
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
145
94
(43.)

and

s
t
DX
1
=() 93 52 . (44.)

These values are used in equations (21.) to update the final demand vector

DF
1
t
=
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
123
33
(45.)

and the value added vector

DV
1
t
=() 93 52 . (46.)
150

From equations (22.) and these four vectors F
t1
, V
t1
, DF
t1
, and DV
t1
, the diagonal
matrix

R(1) =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
896 . 0 0
0 069 . 1
. (47.)

From equation (23.), the diagonal matrix

S(1) =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
976 . 0 0
0 244 . 1
. (48.)

From equation (24.), the updated technical coefficient matrix for the demand-side model
is

1
A = R(1)
0
A S(1) =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
044 . 0 223 . 0
261 . 0 200 . 0
; (49.)

and from equation (25.), the updated technical coefficient matrix for the supply-
side model is

1
B = S(1)
0
B R(1) =
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
044 . 0 104 . 0
558 . 0 200 . 0
. (50.)
151

These reflect substitution effects during the month defining period
1
t . If there are
empirical data available by period, the flexible input output matrices
i
A and
i
B can be
obtained for each period i, and the resilient, total impacts for each period are as estimated in
equation (26.).

3. Flexible National Interstate Economic Model (FlexNIEMO)
The construction of a demand-driven NIEMO is described in detail in Park et al. (2007).
The supply-side version of NIEMO is described in Park (2006). Based on those studies, I
suggest two final equations to estimate total impacts. Let
s N
X be the total input row vector
for the various commodity and service sectors, 47 ,..., 1 = m USC economic sectors, and
52 ,..., 1 = n regions (50 states, plus the District of Columbia, plus the rest of the world). Let
d N
X be total output column vector for the same 47 ,..., 1 = m USC-Sectors and
52 ,..., 1 = n regions. If Z
N
is an nm nm × block diagonal matrix of direct technical flows
between industries within a region, and C is an nm nm × block diagonal matrix of
interregional trade flows, then the total output for a demand-driven version of NIEMO is

F ) A C (I X
N 1 N d d N −
− =
0
; (51.1)

F ) NA (I
N 1 −
− =
0
; (51.2)

152

where

1 s N N N
) X Z( A
−
=
ˆ
0
; (52.1)

0
NA =
0
A C
N d
(52.2)

s N
X
ˆ
is an nm nm × matrix resulting from the block diagonalization of
vector
s N
X , in which all elements off the various regional diagonals are
zero;

d
C =
1 m
j
) C C(
−
ˆ
; (53.)

m
j
C
ˆ
is an nm nm × matrix resulting from the diagonalization of nm × 1 row vector

∑
=
i
m
ij
c
m
j
C , (54.)
in which the off-diagonals elements for a specific regional block are zero and
m
ij
c is an element giving trade flows for USC Sector m from region i to
region j ; and F
N
is a column vector of region specific final demand losses.

Based on the same set of matrices, the total output for the supply-driven version of
NIEMO is
153

1 s
0
N N s N
) C B V(I X
−
− = ; (55.1)

1 N
) NB V(I
−
− =
0
; (55.2)

where

Z ) X ( B
N 1 d N
0
N −
=
ˆ
; (56.1)

0
NB =
s N
C B
0
(56.2)

d N
X
ˆ
is an nm nm × matrix resulting from the block diagonalization of vector
d N
X , in which all elements off the various regional diagonals are zero;

s
C = C ) C (
1 m
i
−
ˆ
; (57.)

m
i
C
ˆ
is an nm nm × matrix resulting from the diagonalization of 1 × nm column
vector

∑
=
j
m
ij
c
m
i
C , (58.)

154

in which the off-diagonals elements for a specific regional block are zero; and
V
N
is a row vector of region specific reductions in value added factors.

Given
s(d) N
t
X
0
, the output for pre-event period
0
t ; and
s(d) N
t
X
1
, the decreased total output
for post-event period
1
t ; equations (13.) through (21.) give four vectors F
N
t
1
, V
N
t
1
, DF
N
t
1
,
and DV
N
t
1
. These are used to compute the RAS matrices

R(1)
N
= ] F [
N
t
1
ˆ
1
^
N
t
] DF [
1
−
(59.)

and

S(1)
N
= ] V [
N
t
1
ˆ
1
^
N
t
] DV [
1
−
, (60.)
which are used to update the technical coefficient matrix in period
1
t for the
demand-driven version of NIEMO

1
NA = R(1)
N
0
NA S(1)
N
(61.)

and for the supply-driven version of NIEMO

1
NB = S(1)
N
0
NB R(1)
N
. (62.)

155

Updated technical coefficient matrices
i
NA and
i
NB can be estimated similarly for
each period i for which data are available. However, one important distinction from the
above approaches should be mentioned. Because only actual region-specific final demand
and value added changes are the focus, it is unnecessary to obtain data on all the other
regions’ final demand and value added changes to update the technical coefficient matrices
by period. I assume these unspecified sectors and states should be indifferent to changes
between pre- and post-event magnitudes. Therefore, unless the actual information for all
industries and states on the changes by period are known, the initial values of other
industries and states in vectors F
N
t
0
and V
N
t
0
cannot be specified. To resolve these problems,
I modified the difference approaches with proportional approaches, constraining untargeted
sectors and states to one, to describe the relationship between initial period
0
t and the next
period
1
t . This implies that there are no significant changes between the pre-event period
0
t
and post-period
1
t for the untargeted sectors, and therefore, the proportional final demand
and value-added changes between the first two periods are defined as,

ρF
N
t
1
=
1
)
ˆ
(
−
F
N
t
1
F
N
t
0
(63.)

and,

ρV
N
t
1
= V
N
t
0
1
) (
−
V
N
t
1
(64.)

156

Also, the proportional changes of the derived final demand and value added are the same
as total output and total input changes, under the constant proportional values of
s
X
R and
d
X
R .

s N
t1
ρX ρF ) NA (I
N
t1
1
0
−
− = ; (65.1)

ρDF
N
t1
= ; (65.2)

and hence,

d N
t1
ρX
1 N
t1
) NB ρV(I
−
− =
0
(66.1)

ρDV
N
t1
= (66.2)

The two equations (65.) and (66.) substitute for equations (59.) and (60.) to calculate
R(1)
N
and S(1)
N
as,

R(1)
N
= ] ρF [
^
N
t
1
1
^
N
t
] ρDF [
1
−
(67.)

and

157

S(1)
N
= ] ρV [
^
N
t
1
1
^
N
t
] ρDV [
1
−
, (68.)

Therefore, equations (61.) and (62.) can be applied with the two newly obtained
diagonalized matrices, equations (67.) and (68.), which are based on the proportioned final
demand and value-added changes.
Unless direct final demand and value-added losses are available, the exogenous direct
losses V ∆
i
t
~
and F ∆
i
t
~
for period i are calculated using the difference between estimated
normal total output and observed reduced total output.

V ∆
N
t
i
~
= (
s N
t
EX
i
-
s N
t
X
i
)
s N
X
R
ˆ
(69.1)

and,

F ∆
N
t
i
~
=
d N
X
R
ˆ
(
d N
t
EX
i
-
d N
t
X
i
); (69.2)

where
s N
t
EX
i
is the estimated total output vector representing normal (baseline)
economic conditions for period i ,

s N
t
EX
i
=
T d N
t
) EX (
i
; (70.)

s N
t
X
i
is the observed reduced total output vector for period i ,
158

s N
t
X
i
=
T d N
t
) X (
i
; (71.)

and
s N
X
R and
d N
X
R are, respectively, the row vector of constant proportions
between value added and total input and the column vector of constant proportions
between final demand and total output across 52 regions and 47 USC Sectors.

Total impacts in the supply-driven model for period i ,
d N
t
X ∆
i
~
, and total impacts in the
demand-driven model for period i ,
s N
t
X ∆
i
~
, are estimated as

d N
t
X ∆
i
~
=
T 1
i
N
t
} ) NB (I V { ∆
i
−
−
~
(72.)

and

s N
t
X ∆
i
~
=
T N
t
1
i
} F ∆ ) NA {(I
i
~
−
− (73.)

Finally, the sum of the total impacts across all post-event periods,
∑
i
d N
ti
X ∆
~
or
∑
i
s N
ti
X ∆
~
, can be obtained and compared against the values computed with from the
constant coefficient versions of NIEMO.
Table V-2 compares the outputs of the FlexNIEMO and NIEMO formulations, based on
the previous paper by Park et al. (2007). Because this example includes final demand and
159

value added losses directly, the constant proportional adjustment of
s
X
R and
d
X
R shown in
equations (69.) are not necessary.
To demonstrate an application, we were able to assemble monthly trade data for the
Louisiana Customs District for all the months of 2003 to 2005. We used these data to test the
FlexNIEMO model. Data applications are in two directions:

Updating technical coefficients based on the last five months of 2005 and
first quarter of 2006,
Estimating direct losses for the same periods, based on the monthly data of
January 2003 to August 2005. The direct losses are obtained from the
differences between forecasts and actual volume, which determined
changes in final demand and changes in value-added resulting from
perturbed foreign imports and exports, from the data.

From the direct losses, we only applied the direct losses of USC Sector 10 (Coal and
petroleum products) to the input data for FlexNIEMO which has monthly updated technical
coefficients for this application. I found that the model is able to estimate the changes in all
of the model’s multipliers. Table V-2 shows comparisons between results obtained from an
application of just the final demand changes to NIEMO along with the application of both
vectors to FlexNIEMO.

160

Table V-1 . Comparison State-by-State Effects, NIEMO and FlexNIEMO.
State D irect Im pacts
To ta l Im p a cts:
NIEM O
T o ta l Im p a cts:
FLEX N IEM O
AL 0.0000 -1.1215 -0.0071
AK 0.0000 -0.0379 -0.0002
AZ 0.0000 -0.5650 -0.0013
AR 0.0000 -0.7722 -0.0052
C A 0.0000 -76.9350 -0.3833
C O 0.0000 -0.5722 -0.0014
C T 0.0000 -0.3026 -0.0016
D E 0.0000 -0.0732 -0.0005
D C 0.0000 -0.0356 -0.0002
FL 0.0000 -4.7969 -0.0443
GA 0.0000 -1.0020 -0.0053
H I 0.0000 -0.0315 -0.0001
ID 0.0000 -0.1203 -0.0010
IL 0.0000 -1.4019 -0.0044
IN 0.0000 -1.6162 -0.0079
IA 0.0000 -0.3004 -0.0010
KS 0.0000 -0.5839 -0.0023
KY 0.0000 -0.6168 -0.0024
LA -345.0733 -708.7323 -348.1264
M E 0.0000 -0.0647 -0.0003
M D 0.0000 -0.2382 -0.0013
M A 0.0000 -0.2737 -0.0009
M I 0.0000 -1.5320 -0.0044
M N 0.0000 -0.5852 -0.0023
M S 0.0000 -4.3134 -0.0294
M O 0.0000 -0.6270 -0.0028
M T 0.0000 -0.2970 -0.0003
N E 0.0000 -0.1571 -0.0005
N V 0.0000 -0.1075 -0.0003
N H 0.0000 -0.1237 -0.0008
N J 0.0000 -0.8444 -0.0036
N M 0.0000 -0.4224 -0.0007
N Y 0.0000 -0.9803 -0.0031
N C 0.0000 -0.7675 -0.0031
N D 0.0000 -0.0449 -0.0002
O H 0.0000 -1.4619 -0.0038
O K 0.0000 -2.0785 -0.0030
O R 0.0000 -0.2579 -0.0009
PA 0.0000 -1.6926 -0.0085
R I 0.0000 -0.0810 -0.0005
SC 0.0000 -0.4270 -0.0025
SD 0.0000 -0.0496 -0.0002
TN 0.0000 -0.9685 -0.0057
TX 0.0000 -79.2055 -0.3066
U T 0.0000 -0.6304 -0.0011
V M 0.0000 -0.0386 -0.0002
V A 0.0000 -0.5299 -0.0026
W A 0.0000 -0.5678 -0.0019
W V 0.0000 -0.2774 -0.0015
W I 0.0000 -0.8766 -0.0032
W Y 0.0000 -0.2943 -0.0006
U S T otal -345.0733 -900.4341 -3 4 8 .9 928
Rest of W orld 0.0000 -89.2166 -0.6239
W orld T otal -345.0733 -989.6507 -3 4 9 .6 167

Note: 1. The Case of Customs District of Louisiana.
2. Foreign Export Losses of USC Sector 10 (Coal and petroleum products) during Last Four Months of
2005 ($M.).
161

Figure V-2 shows the change of total multipliers by each month resulted from
FlexNIEMO, where all multipliers are below 1.1. The monthly multipliers can be compared
to the constant total multiplier of 2.8670 obtained from NIEMO. When comparing the
multipliers of the two models in Figure V-2, the results of FlexNIEMO demonstrate that we
are able to estimate significantly reduced impact multipliers, reflecting the many
substitutions forced by the cataclysmic event.

Figure V-2 Monthly Multipliers Changes for FlexNIEMO
FlexNIEMO
1.00
1.02
1.04
1.06
1.08
1.10
Aug. 05 Sep. 05 Oct. 05 Nov. 05 Dec. 05 Jan. 06 Feb. 06 Mar. 06

The same results can be verified from the comparison results between FlexNIEMO
and NIEMO are mapped in Figures A.V-1 to A.V-10 of Chapter V Appendix. Figures A.V-1
to A.V-8 indicate month-to-month proportional changes to total losses, which show the
spatial differences of economic losses by month. Figure A.V-9 shows the proportional state-
by-state economic impact losses to total economic losses resulted from FlexNIEMO, while
Figure A.V-10 indicates the proportional economic losses stemming from NIEMO. The
comparison clear shows the spatially distributed impacts of FlexNIEMO are significantly
162

different from those of NIEMO. Also, it is especially interesting that some states swung
from experiencing negative impacts to garnering positive impacts.

4. Conclusions and Remarks
As we become a global economy and as changes of all sorts occur at a faster and faster
pace via radical innovations in transportation and communication systems, we also expect
facilities disruptions (such as from terrorist attacks and natural disasters) as well as
significant facilities expansions. Evironmental impacts from climate change are also hard to
predict To provide more useful and important information to policy makers because they
highlight local economic impacts, input-output models are routinely used.
Unfortunately, analysts trade-off convenience for plausibility if the direct impacts are
large because the fixed-coefficients assumptions of IO are necessarily violated. Questions
about the effects of time, distance and industry linkage can be posed and studied. I suggest
that it is worthwhile and that it is feasible to study how such coefficients reflect actual
adaptations. The heart of economics is all about how agents respond to new facts of life, as
reported to them by price changes. The sum total of very large numbers of adjustments to a
major event can be summarized in the perturbed technological coefficients.
This essay suggests coefficient adaptations by period and by industry and by region can
be plotted. I have suggested and illustrated a straightforward way to identify changing IO
coefficients. A simple “toy” (2 x 2) example and an actual application, both, show that much
smaller multipliers result, as expected. From those results, I can identify the economic
resilience that standard IO analyses miss. For future study, structural decomposition analysis
163

(see Rose and Casler, 1996) may be necessary to analyze the temporal and spatial
coefficients obtained from FlexNIEMO.
164

X. CHAPTER V REFERENCES

Chenery, H.B., 1953, Regional Analysis, in H.B. Chenery, P.G. Clark and V.C. Pinna, eds.,
The Structure and Growth of the Italian Economy, U.S. Mutual Security Agency, Rome:
98-139.

Ghosh, A., 1958, Input-Output Approach to an Allocative System, Economica, Vol. XXV:
58-64.

Ghosh, A., 1964, Experiments with Input-Output Models: An Application to the Economy of
the United Kingdom, 1948-55, London, Cambridge University Press.

Gordon, P., J.Y. Park, and H. W. Richardson, 2006, Modeling Economic Impacts in Light of
Substitutions in Household Sector Final Demand, submitted to Economic Modelling.

Gordon, P., H. W. Richardson, J. E. Moore, II, and J.Y. Park, 2007, The Economic Impacts
of a Terrorist Attack on the U.S. Commercial Aviation System, Forthcoming in Risk
Analysis.

Hewings, G.J.D., 1985, Regional Input-Output Analysis. Beverly Hills, CA: Sage
Publications, Inc.

Isard, W., 1951, Interregional and Regional Input-Output Analysis: A Model of a Space
Economy, Review of Economics and Statistics, 33: 318-328.

Jack Faucett Associates, INC, 1983, The Multiregional Input-Output Accounts, 1977:
Introduction and Summary, Vol. I (Final Report), prepared for the US Department of
Health and Human Services, Washington.

Leontief, W., 1936, Quantitative Input and Output Relations in the Economic System of the
United States, Review of Economic Statistics, XVIII (3): 105-125.

Leontief, W., 1941, The Structure of American Economy, 1919-1929: An Empirical of
Equilibrium Analysis, Cambridge, MA, Harvard University Press.

Miller, R. E. and P. D. Blair, 1985, Input-Output Analysis: Foundations and Extensions.
Englewood Cliffs, NJ: Prentice-Hall, Inc.

Moses, L.N., 1955, The Stability of Interregional Trading Patterns and Input-Output
Analysis, American Economic Review, 45: 803-832

165

Park, J. Y., 2006, The Economic Impacts of Dirty- Bomb Attacks on the Los Angeles and
Lon Beach Port: Applying Supply-driven NIEMO, Paper will be presented at 17
rd
Annual Meeting of the Association of Collegiate Schools of Planning, Fort Worth, TX,
USA, November 9-12.

Park, J.Y., P. Gordon, S.J. Kim, Y.K. Kim, J.E. Moore II, and H.W. Richardson, 2006,
Estimating the State-by-State Economic Impacts of Hurricane Katrina, Paper presented
at CREATE symposium: Economic and Risk Assessment of Hurricane Katrina,
University of Southern California, California, USA, August 18-19.

Park, J.Y., P. Gordon, J.E. Moore II, H.W. Richardson, and L. Wang, 2007, Simulating the
State-by-State Effects of Terrorist Attacks on Three Major U.S. Ports: Applying NIEMO
(National Interstate Economic Model): 208-234, in H.W. Richardson, P. Gordon and J.E.
Moore II, eds., The Economic Costs and Consequences of Terrorism. Cheltenham:
Edward Elgar.

Park, J.Y., C.K. Park, and P. Gordon, 2007, The State-by-State Effects of Mad Cow Disease
using a new MRIO model, revised to resubmit to Journal of Agricultural and Resource
Economics.

Polenske, K.R., 1980, The US Multiregional Input-Output Accounts and Model, DC Health,
Lexington, MA.

Richardson, H.W., P. Gordon, J.E. Moore, II, S.J. Kim, J.Y. Park, and Q. Pan, 2007,
Tourism and Terrorism: The National and Interregional Economic Impacts of Attacks on
Major U.S. Theme Parks: 235-253, in H.W. Richardson, P. Gordon and J.E. Moore II,
eds., The Economic Costs and Consequences of Terrorism. Cheltenham: Edward Elgar.
166

REMARKS

This dissertation develops extensions of the typical one-region Leontief input-output
(IO) model. These elaborations contribute to a better and more detailed understanding of
economic impacts from man-made or natural disasters.
As a useful tool for estimating socioeconomic impacts, IO approaches have been widely
utilized, and the applied analyses to the study of economic impacts (actual and hypothetical)
have led to a number of discoveries and innovations. Although the applications using IO
models are widely conducted, one consistently raised and unavoidable question is over the
constant fixed coefficients production function assumption. While we now have better data,
better software and hardware than ever, many impact analyses, still depend on the traditional
version of the IO model. While many trials advancing the IO approach such as the non-linear
version (West and Jackson, 2004) or the econometric+IO model (Rey, 2000) have been
introduced, many new demands on the computable general equilibrium (CGE) approaches to
economic impacts analysis have been made over the past thirty years (Dietzenbacher and
Lahr, 2001).
In this sense, this dissertation provides two alternatives for the standard CGE model. The
price-sensitive supply-driven IO approach can be extended to a more simplified non-linear
IO approach, combining exogenous price elasticities of demand for specific sectors with the
supply-side IO model. The newly obtained coefficients would reflect price effects from
observed economic behavior to consume due to the changed economic conditions. Also, the
flexible input-output approaches address endogenous coefficient changes based on actually
167

observed economic conditions. The approaches require both demand- and supply-driven IO
models, and hence can be understood as a new type of CGE model, one that does not depend
on data on price-elasticities. This is because the supply-side IO model has been shown to be
an absolute price IO model. In light of difficulties in obtaining price elasticities when using
the traditional CGE approach, therefore, the flexible IO versions revitalize the usefulness of
IO models.
Spatial modeling and disaggregation have been hampered by difficulties obtaining
plausible, complete spatial data sets. The data incompatibility problem between different
industrial code systems, incomplete commodity trade flows between states or sub-state areas,
or no information on service trade flows have raised serious questions related to making
spatially, temporally extended IO model operable. However, because this dissertation
includes separate methodologies necessary to constructing spatially, temporally extended IO
model, its contribution is not restricted to developing extended IO models, but also involves
processes that remedy the problems run into when constructing operational IO models. The
various types of operational IO models, which are suggested in Figure R-1 include 12 types
that can be used to analyze economic impacts according to the characteristics of disasters.
Based on the suggested IO models, several empirical research projects have been
conducted, as summarized in Table R-1. Most of them depend on time-series analyses for
the calculation of direct losses, and the IO models address the indirect and total economic
impacts based on their coefficients.
Also, based on the methodologies used in this dissertation, further research directions are
already being explores and projects that involve various applications are now underway.
168

V-2 Figure R-1 Various Input-Output Models with Suggested Methodologies
Price sensitive
supply-driven
USIO
Price sensitive
supply-driven
NIEMO
Non-linear
USIO
-
Non-linear
NIEMO
-
Demand -
driven
FlexUSIO
Supply - driven
FlexUSIO
Demand -
dirven
FlexNIEMO
Supply - dirven
FlexNIEMO
Price elasticity of demand Temporal extension method
Demand-driven
NIEMO
Supply-driven
NIEMO
Demand-driven
USIO
Supply-driven
USIO
Spatial extension
method

First, although many applications using IO models report induced effects as well as
direct and indirect impacts, all the versions of IO models introduced in this dissertation are
open to the household sector. Therefore, to capture the induced effects, it is necessary to
create closed IO models.
169

Table R-1 Empirical Applications of the Various IO models
Empirical Studies Models
Researchers Title
Demand-
driven
NIEMO
Supply-
driven
NIEMO
Price sensitive
supply-driven
USIO
Supply-
driven
FlexNIEMO
Park et al.
(2006)
The State-by-State Economic Impacts
of Mad Cow Disease on the United
States
x
Park (2007a)
The Economic Impacts of a Dirty-
Bomb Attack on the Los Angeles and
Long Beach Port: Applying Supply-
driven NIEMO
x
Park (2007b)
Application of a Price-Sensitive
Supply-Side Input-Output Model to an
Examination of the Economic Impacts
of Hurricane Katrina and Rita
Disruptions of the U.S. Oil-Industry
x
Park et al.
(2007b)
Simulating The State-by-State Effects
of Terrorist Attacks on Three Major
U.S. Ports: Applying NIEMO
(National Interstate Economic Model)
x x
Richardson et
al. (2007)
Tourism and Terrorism: The National
and Interregional Economic Impacts of
Attacks on Major U.S. Theme Parks
x
Park et al.
(2007c)
Estimating the State-by-State
Economic Impacts of Hurricane
Katrina
x x
Lee et al.
(2007)
Estimating the State-by-State
Economic Impacts of Bio-terrors: The
Case Study of FMD
x
Gordon et al.
(2008)
U.S. Border Closing Economic Impact
Simulations
x x
Park et al.
(2007d)
The Regional Economic Impacts of
Hurricanes Katrina and Rita on Oil and
Gas Refinery Operations in the Gulf of
Mexico: Applying a Flexible Multi-
regional Input-Output Model
x
Gordon et al.
(2007)
The State-By-State Economic Impacts
of an Attack on the Los Angeles-Long
Beach Harbors in Light of Trade
Diversion to Other West Coast Ports
x x
V-3 Table R-1 Empirical Applications using the Various IO models

170

Second, the suggested models and methods can be applied to the construction of
international IO models. For example, NIEMO can be combined with the 2000 Asian
International IO models published by the Institute of Developing Economics (IDE). This
project would construct an Inter-Countries Economic Model of Asian Pacific Rim
(ICEMAP). This effort can be applied to analyze e.g. the economic effects on the U.S. from
Free Trade Agreements (FTA) between the U.S. and various Asian countries, or to address
the international economic impacts from climate change in the Pacific Rim.
Third, the current IO models can be combined with the Hazards U.S. Multi-Hazard
(HAZUS-MH) model released by the Federal Emergency Management Agency (FEMA).
The HAZUS-MH provides historical direct losses resulting from various hypothetical and
actual U.S. disasters, including earthquakes, floods, and hurricanes. However, because
HAZUS-MH assumes constant recapture factors, it requires lessening the strict assumptions
by developing recapture factor functions (Park et al., 2007a). Based on this proposed remedy,
various combinations of NIEMO and HAZUS-MH would provide policy-makers with useful
information on their prevention programs with respect to future natural disasters (Park and
Cho, 2007).
Finally, decomposition of NIEMO-type IO models with respect to transport mode types
will provide useful policy information for transportation systems planning. Most usefully,
the origin-destination (OD) trade flows of the truck freight mode can be distributed onto the
national highway system, providing estimates of economic losses resulting from new
burdens on the highway system from disruptions that would follow from e.g. earthquakes or
hypothetical terrorist attacks on the many vulnerable bridges that serve interstate and
international trade.
171

While this dissertation only includes methodologies for economic modeling but, as seen
in this remarks section, the newly developed IO approaches and model construction
processes can be applied to many empirical challenges associated with various ecological
disruptions. Also, these approaches can be developed and applied to advance yet
unidentified methodologies which also should eventually be addressed in the area of regional
science.

172

REMARKS REFERENCES

Gordon, P., J.E. Moore II, J.Y. Park, and H.W. Richardson, 2007, The State-By-State
Economic Impacts of An Attack on the Los Angeles-Long Beach Harbors In Light of
Trade Diversion to Other West Coast Ports, Paper is invited to Growth and Change, and
will be presented at 54th Annual North American Meetings of the Regional Science
Association International, Savannah, GA, November 7-10.

Gordon, P., J.E. Moore II, J.Y. Park, and H.W. Richardson, 2008, U.S. Border Closing
Economic Impact Simulations, in H.W. Richardson, P. Gordon and J.E. Moore II, eds.,
Global Business and Terrorism. Cheltenham: Edward Elgar (Forthcoming).

Lee, B., P. Gordon, J. E. Moore II, H. J. Y. Park, W. Richardson, 2007, Estimating the State-
by-State Economic Impacts of Bio-terrors: The Case Study of FMD, Paper will be
presented at 54th Annual North American Meetings of the Regional Science Association
International, Savannah, GA, November 7-10.

Park, J.Y., 2007a, The Economic Impacts of a Dirty- Bomb Attack on the Los Angeles and
Long Beach Port: Applying Supply-driven NIEMO, submitted to Journal of Homeland
Security and Emergency Management.

Park, J.Y., 2007b, Application of a Price-Sensitive Supply-Side Input-Output Model to an
Examination of the Economic Impacts of Hurricane Katrina and Rita Disruptions of the
U.S. Oil-Industry, submitted to Ecological Economics.

Park, J.Y. and J.K. Cho, 2007, Combining the National Interstate Economic Model
(NIEMO) with HAZUS: The Case of Hurricane Andrew, Paper will be submitted to the
47
th
Annual Meeting of the Western Regional Science Association, Hilton Waikoloa
Resort, The Big Island of Hawaii, February 17-20.

Park, J.Y., J.K. Cho, and A. Rose, 2007a, Modeling a Major Source of Economic Resilience
to Disasters: Recapturing Lost Production, Paper submitted to the 2007 Annual
Conference of Risk Analysis Society, Marriott Rivercenter, San Antonio, Texas,
December 9-12.

Park, J.Y., P. Gordon, J. E. Moore II, and H. W. Richardson, L. Wang, 2007b, Simulating
The State-by-State Effects of Terrorist Attacks on Three Major U.S. Ports: Applying
NIEMO (National Interstate Economic Model): 208-234, in H.W. Richardson, P.
Gordon and J.E. Moore II, eds., The Economic Costs and Consequences of Terrorism.
Cheltenham: Edward Elgar.

Park, J.Y., P. Gordon, J. E. Moore II, and H. W. Richardson, 2007c, Estimating the State-by-
State Economic Impacts of Hurricane Katrina, in H.W. Richardson, P. Gordon and J.E.
Moore II, eds., The Socioeconomic Costs and Consequences of Hurricane Katrina.
Cheltenham: Edward Elgar (Forthcoming).
173

Park, J.Y., P. Gordon, Y.K. Kim, J.E. Moore II, and H.W. Richardson, 2007d, The Regional
Economic Impacts of Hurricanes Katrina and Rita on Oil and Gas Refinery Operations
in the Gulf of Mexico: Applying a Flexible Multi-regional Input-Output Model. Paper
will be submitted to Journal of Economic Geography.

Park, J.Y., C.K. Park, and P. Gordon, 2006, The State-by-State Economic Impacts of Mad
Cow Disease on the United States, revised to resubmit to Journal of Agricultural and
Resource Economics.

Richardson, H. W., P. Gordon, J. E. Moore, II, S.J. Kim, J.Y. Park, and Q. Pan, 2007,
Tourism and Terrorism: The National and Interregional Economic Impacts of Attacks on
Major U.S. Theme Parks: 235-253, in H.W. Richardson, P. Gordon and J.E. Moore II,
eds., The Economic Costs and Consequences of Terrorism. Cheltenham: Edward Elgar.

Sergio J. Rey, 2000, Integrated Regional Econometric+Input-output Modeling: Issues and
Opportunities, Papers in Regional Science, 79 (3): 271-292.

174

REFERENCES

Anselin, T. and D.A. Griffith, 1988 Do Spatial Effects Really matter in Regression
Analysis? Papers in Regional Science 65(1): 11-34.
Anselin, T., 1988, Spatial Econometrics: Methods and Models. Dorddrecht: Kluwer
Academic Publishers.
Ara, K., 1959, The Aggregation Problem in Input-Output Analysis, Econometrica 27: 257-
262.
Batten D.F., 1982, The Interregional Linkages Between National and Regional Input-Output
Models. International Regional Science Review 7:53-67.
Blair, P. and R.E. Miller, 1983, Spatial Aggregation of Multiregional Input-Output Models,
Environmental Planning A 15: 187-206.
Bon, R. and T. Yashiro, 1995, Comparative Stability Analysis of Demand-side and Supply-
side Input-Output Models: The Case of Japan, 1960-90, Applied Economics Letters, 3:
349-354.
Bon, R., 1988, Supply-side Multiregional Input-Output Models, Journal of Regional Science,
28 (1): 41-50.
Bon, R., 2001, Comparative Stability Analysis of Demand-side and Supply-side Input-
Output Models: Toward Index of Economic “Maturity”, in Lahr, M.L. and E.
Dietzenbacher eds., Input-Output Analysis: Frontiers and Extensions, NY, Palgrave.
Brundson, C.A., S. Fotheringham, and M. Charlton, 1996, Geographical Weighted
Regression: A Method for Exploring Spatial Nonstationarity, Geographical Analysis 28:
281-298.
Cai, J., P. Leung, and J. Mak, 2006, Tourism’s Forward and Backward Linkages, Journal of
Travel Research, 45 (1): 36-52.
Cai, J., P. Leung, M. Pan, and S. Pooley, 2005, Linkage of Fisheries Sectors to Hawaii's
Economy and Economic Impacts of Longline Fishing Regulations, University of Hawaii
- NOAA, Joint Institute for Marine and Atmospheric Research, Honolulu, Hawaii.
Canning, P. and Z. Wang, 2005, A Flexible Mathematical Programming Model to Estimate
Interregional Input-Output Accounts, Journal of Regional Science 45(3): 539-563.
175

Chen, C.Y. and A. Rose, 1991, The Absolute and Relative Joint Stability of Input-Output
Production and Allocation Coefficients, in Peterson, W., ed., Advances in Input-Output
Analysis. NY, Oxford University Press.
Chenery, H.B., 1953, Regional Analysis, in H.B. Chenery, P.G. Clark and V.C. Pinna, eds.,
The Structure and Growth of the Italian Economy, U.S. Mutual Security Agency, Rome:
98-139.
Davis, H.C. and E.L. Salkin, 1984, Alternative Approaches to the Estimation of Economic
Impacts Resulting from Supply Constraints, The Annals of Regional Science, 18: 25-34.
Dietzenbacher, E., 1992, Aggregation in Multisector Models: Using the Perron Vector,
Economic Systems Research, 4: 3-24.
Dietzenbacher, E., 2002, Interregional Multipliers: Looking Backward, Looking Forward,
Regional Studies, 36 (2):125-136.
Erlbaum, N. and J. Holguin-Veras, 2005, Some Suggestions for Improving CFS Data
Products, Paper presented at Commodity Flow Survey (CFS) Conference, Boston Seaport
Hotel & World Trade Center, Boston, Massachusetts, Jul. 8~9.
Ghosh, A., 1958, Input-Output Approach to an Allocative System, Economica, XXV: 58-64.
Ghosh, A., 1964, Experiments with Input-Output Models: An Application to the Economy of
the United Kingdom, 1948-55, London, Cambridge University Press.
Giarratani, F., 1976, Application of an Interindustry Supply Model to Energy Issues,
Environmental Planning A, 8: 447-454.
Gibbons, J.C., A. Wolsky, and G. Tolley, 1982, Approximate Aggregation and Error in
Input-Output Models, Resource and Energy 4: 203-230.
Giuliano, G., P. Gordon, Q. Pan, J.Y. Park, and L. Wang, 2007, Estimating Freight Flows for
Metropolitan Area Highway Networks Using Secondary Data Sources, Networks and
Spatial Economics (forthcoming).
Gordon, P., H. W. Richardson, J. E. Moore, II, and J.Y. Park, 2007, The Economic Impacts
of a Terrorist Attack on the U.S. Commercial Aviation System, Forthcoming in Risk
Analysis.
Gordon, P., J.E. Moore II, J.Y. Park, and H.W. Richardson, 2007, The State-By-State
Economic Impacts of An Attack on the Los Angeles-Long Beach Harbors in Light of
Trade Diversion to Other West Coast Ports, Paper is invited to Growth and Change, and
will be presented at 54th Annual North American Meetings of the Regional Science
Association International, Savannah, GA, November 7-10.
176

Gordon, P., J.E. Moore II, J.Y. Park, and H.W. Richardson, 2008, U.S. Border Closing
Economic Impact Simulations, in H.W. Richardson, P. Gordon and J.E. Moore II, eds.,
Global Business and Terrorism. Cheltenham: Edward Elgar (Forthcoming).
Gordon, P., J.Y. Park, and H. W. Richardson, 2006, Modeling Economic Impacts in Light of
Substitutions in Household Sector Final Demand, submitted to Economic Modeling.
Gruver, G.W., 1989, On the Plausibility of the Supply-driven Input-Output Model: A
Theoretical Basis for Input-coefficient Change, Journal of Regional Science, 29, 441-
450.
Harrigan, F., J.W. McGilvray, and I.H. McNicoll, 1981, The Estimation of Interregional
Trade Flows, Journal of Regional Science, 21(1): 65-78.
Hatanaka, M., 1952, Note on Consolidation within a Leontief System, Econometrica 20:
301-303.
Hewings, G.J.D., 1985, Regional Input-Output Analysis, Beverly Hills, CA: Sage
Publications, Inc.
Howe, D., 2004, Planning Scenarios, available at the
(http://132.160.230.113:8080/revize/repository/CSSPrototype/simplelist/Planning_Scena
rios_Exec_Summary_.pdf).
Isard, W., 1951, Interregional and Regional Input-Output Analysis: A Model of a Space
Economy, Review of Economics and Statistics, 33: 318-328.
Jack Faucett Associates INC., 1983, The Multiregional Input-Output Accounts, 1977:
Introduction and Summary, Vol. I (Final Report), prepared for the U.S. Department of
Health and Human Services, Washington.
Jackson, R.W., W.R. Schwarm, Y. Okuyama, and S. Islam, 2006, A Method for
Constructing Commodity by Industry Flow Matrices. The Annals of Regional Science
40:909-920.
Lahr, M.L., 1993, A Review of the Literature Supporting the Hybrid Approach to
Constructing Regional Input-Output Models, Economic Systems Research 5: 277-293.
Lee, B., P. Gordon, J. E. Moore II, H. J. Y. Park, W. Richardson, 2007, Estimating the State-
by-State Economic Impacts of Bio-terrors: The Case Study of FMD, Paper will be
presented at 54th Annual North American Meetings of the Regional Science Association
International, Savannah, GA, November 7-10.
Lee, C., 1973, Models in Planning: An Introduction to the Use of Quantitative Models in
Planning, NY, Pergamon Press Inc.
177

Leontief, W., 1936, Quantitative Input and Output Relations in the Economic System of the
United States, Review of Economic Statistics, XVIII (3): 105-125.
Leontief, W., 1941, The Structure of American Economy, 1919-1929: An Empirical of
Equilibrium Analysis, Cambridge, MA, Harvard University Press.
Leontief, W., 1951, The Structure of American Economy, 1919-1939: an Empirical
Application of Equilibrium Analysis, Oxford Univ. Press, NY.
Leontief, W., 1976, The Structure of American Economy, 1919-1939: An Empirical of
Equilibrium Analysis, White Plains, NY, International Arts and Science Press, Inc.
LeSage, J.P., 1999, The Theory and Practice of Spatial Econometrics, University of Toledo,
MN, Available at the http://www.spatial-econometrics.com/html/sbook.pdf.
LeSage, J.P., and R.K. Pace, 2006, Spatial Econometric Modeling of Origin-Destination
flows, Available at SSRN: http://ssrn.com/abstract=924609.
Lindall, S., D. Olsen, and G. Alward, 2005, Deriving Multi-Regional Models Using the
IMPLAN National Trade Flows Model, Paper presented at the 2005 MCRSA/SRSA
Annual Meeting, April 7-9, Arlington, VA.
Liu, L.N. and P. Vilain P., 2004, Estimating Commodity Inflows to a Substate Region Using
Input-Output Data: Commodity Flow Survey Accuracy Tests, Journal of Transportation
and Statistics 7(1): 23-37.
McManus, M., 1956, General Consistent Aggregation in Leontief Models, Yorkshire Bulletin
of Economic and Social Research 8: 28-48.
McMillen, D.P., 1996, One Hundred Fifty Years of Land Values in Chicago: A
Nonparametric Approach. Journal of Urban Economics 40: 100-124.
Miller, R.E. and G. Shao, 1990, Spatial and Sectoral Aggregation in the Commodity-
Industry Multiregional Input-Output Model, Environment and Planning A 22: 1637-
1656.
Miller, R.E. and P. Blair, 1981, Spatial Aggregation in Interregional Input-Output Models,
Papers of the Regional Science Association 48: 149-164.
Miller, R.E. and P. Blair, 1985, Input-Output Analysis: Foundations and Extensions, New
Jersey: Prentice-Hall.
Morimoto, Y., 1970, On Aggregation Problems in Input-Output Analysis, Review of
Economic Studies 37: 119-126.
178

Moses, L.N., 1955, The Stability of Interregional Trading Patterns and Input-Output
Analysis, American Economic Review, 45: 803-832.
Oosterhaven, J., 1988, On the Plausibility of the Supply-driven Input-Output Model, Journal
of Regional Science, 28: 203-217.
Oosterhaven, J., 1989, The Supply-driven Input-Output Model: A New Interpretation but
Still Implausible, Journal of Regional Science, 29: 459-465.
Oosterhaven, J., 1996, Leontief versus Ghoshian Price and Quantity Models, Southern
Economic Journal, 62(3): 750-759.
Park, J. Y., 2006, The Economic Impacts of Dirty- Bomb Attacks on the Los Angeles and
Lon Beach Port: Applying Supply-driven NIEMO, Paper presented at 17rd Annual
Meeting of the Association of Collegiate Schools of Planning, Fort Worth, TX, USA,
November 9-12.
Park, J.Y. and J.K. Cho, 2007, Combining the National Interstate Economic Model
(NIEMO) with HAZUS: The Case of Hurricane Andrew, Paper will be submitted to the
47th Annual Meeting of the Western Regional Science Association, Hilton Waikoloa
Resort, The Big Island of Hawaii, February 17-20.
Park, J.Y., 2006, Estimation of State-by-State Trade Flows for Service Industries, Paper
presented at North American Meetings of the Regional Science Association International
53rd Annual Conference, Fairmont Royal York Hotel, Toronto, Canada, November 16-
18.
Park, J.Y., 2007, Application of a Price-Sensitive Supply-Side Input-Output Model to an
Examination of the Economic Impacts of Hurricane Katrina and Rita Disruptions of the
U.S. Oil-Industry, submitted to Ecological Economics.
Park, J.Y., C.K. Park, and P. Gordon, 2006, The State-by-State Economic Impacts of Mad
Cow Disease on the United States, revised to resubmit to Journal of Agricultural and
Resource Economics.
Park, J.Y., J.K. Cho, and A. Rose, 2007, Modeling a Major Source of Economic Resilience
to Disasters: Recapturing Lost Production, Paper submitted to the 2007 Annual
Conference of Risk Analysis Society, Marriott Rivercenter, San Antonio, Texas,
December 9-12.
Park, J.Y., P. Gordon, J. E. Moore II, and H. W. Richardson, 2004, Construction of a U.S.
Multiregional Input Output Model Using IMPLAN, Paper presented at 2004 National
IMPLAN User’s Conference, Shepherdstown, West Virginia, USA, October 6-8.
179

Park, J.Y., P. Gordon, J. E. Moore II, and H. W. Richardson, 2007, Estimating the State-by-
State Economic Impacts of Hurricane Katrina, in H.W. Richardson, P. Gordon and J.E.
Moore II, eds., The Socioeconomic Costs and Consequences of Hurricane Katrina.
Cheltenham: Edward Elgar (Forthcoming).
Park, J.Y., P. Gordon, J. E. Moore II, and H. W. Richardson, L. Wang, 2007, Simulating The
State-by-State Effects of Terrorist Attacks on Three Major U.S. Ports: Applying NIEMO
(National Interstate Economic Model): 208-234, in H.W. Richardson, P. Gordon and J.E.
Moore II, eds., The Economic Costs and Consequences of Terrorism. Cheltenham:
Edward Elgar.
Park, J.Y., P. Gordon, J.E. Moore II, and H.W. Richardson, 2006, Simulating the State-by-
State Effects of Terrorist Attacks on Three Major U.S. Ports: Applying NIEMO
(National Interstate Economic Model), available at the
http://www.metrans.org/nuf/documents/MooreII.pdf.
Park, J.Y., P. Gordon, Y.K. Kim, J.E. Moore II, and H.W. Richardson, 2007, The Regional
Economic Impacts of Hurricanes Katrina and Rita on Oil and Gas Refinery Operations
in the Gulf of Mexico: Applying a Flexible Multi-regional Input-Output Model. Paper
will be submitted to Journal of Economic Geography.
Pindyck, R.S. and D.L. Rubinfeld, 1998, Microeconomics, NJ. Prentice-Hall, Inc.
Polenske, K.L., 1980, The US Multiregional Input-Output Accounts and Model, DC Health,
Lexington, MA.
Porojan, A., 2001, Trade Flows and Spatial Effects: The Gravity Model Revisited. Open
Economies Review 12(3): 265-280.
Richardson, H. W., P. Gordon, J. E. Moore, II, S.J. Kim, J.Y. Park, and Q. Pan, 2007,
Tourism and Terrorism: The National and Interregional Economic Impacts of Attacks on
Major U.S. Theme Parks: 235-253, in H.W. Richardson, P. Gordon and J.E. Moore II,
eds., The Economic Costs and Consequences of Terrorism. Cheltenham: Edward Elgar.
Robison, W.S., 1950, Ecological Correlations and the Behavior of Individuals, American
Sociological Review 15: 351-357.
Rose, A. and T. Allison, 1989, On the Plausibility of the Supply-driven Input-Output Model:
Empirical Evidence on Joint Stability, Journal of Regional Science, 29: 451-458.
Samuelson, P.A. and W.D. Nordhaus, 1995, Economics, McGraw-Hill, Inc.
Sergio J. Rey, 2000, Integrated Regional Econometric+Input-output Modeling: Issues and
Opportunities, Papers in Regional Science, 79 (3): 271-292.
180

Shao, G. and R.E. Miller, 1990, Demand-side and Supply-side Commodity-Industry
Multiregional Input-Output Models and Spatial Linkages in the US regional Economy,
Economic Systems Research, 2 (4): 385-405.
Theil, H., 1957, Linear Aggregation in Input-Output Analysis, Econometrica 25 (1): 111-
122.
Wang, E.C., 1997, Structural Change and Industrial Policy in Taiwan, 1966-91: An
Extended Input-Output Analysis, Asian Economic Journal, 11 (2): 187-206.
Wilson, A.G., 1970, Inter-regional Commodity Flows: Entropy Maximizing Approaches.
Geographical Analysis 2:255-282.

181

APPENDICES
Appendix A: Chapter II Appendices
Appendix A1. Data Reconciliation Steps, SCTG and IMPLAN
Figure A. II-1. Data Reconciliation Steps, SCTG and IMPLAN

Figure 4. Data Reconciliation Steps, SCTG and IMPLAN

Notes:
Bold: Used as Reconciliation Code
1: Sector type
2: One = One sector, Many = Multiple Sectors
3: Quality of Reconciled Data
4: Sources and Abbreviations:
IMPLAN
BEA: Bureau of Economic Analysis (http://www.bea.doc.gov)
SCTG : Standard Classification of Transported Goods (http://www.bts.gov/cfs/sctg/welcome.htm)
HS : Harmonized System (http://www.statcan.ca/trade/htdocs/hsinfo.html)
5. Source: Park et al. (2007)
Step 1-1
IMPLAN
(2001)
BEA Code
(1997)
1. Industry.-to-Commodity.
2. One-to-One
3. Perfect
4. IMPLAN
Step 1-2
BEA code
(1997)
HS Code
(1997)
1. Commodity-to-Commodity.
2. One-to-Many
3. Very Good
4. BEA web

Step 1-3
HS Code
(1997)
SCTG code
(1997)
1. Commodity-to-Commodity
2. Many-to-One
3. Perfect
4. HS web
Step 2
BEA Code
(1997)
SCTG code
(1997)
1. Commodity-to-Commodity
2. Almost Many-to- One
3. Very good
Step 3
BEA Code
(1997)
SCTG code
(1997)
IMPLAN
(2001)
1. Industry-to-Commodity
2. Almost Many-to-One
3. Very good
182

Appendix A2. Data Reconciliation and Definitions of USC Sectors

The aggregation of 2001 IMPLAN sectors was in light of the fundamental characteristics
of sectors: commodity sectors vs. non-commodity (service) sectors. IMPLAN’s commodity
sectors are based on NAICS and BEA sectors and were aggregated into 43 SCTG sectors
used in the 1997 CFS, as shown in Appendix A1. Although the 43 (1997) CFS SCTG
commodity sectors were based on SIC industry codes, and not the NAICS industry codes,
SCTG commodity classification names in 1997 CFS and 2002 CFS remain the same.
Because at the time of our work, the 2002 CFS full data sets are not available, we worked
with the 1997 CFS data. To test our converted results, we compared aggregates to the
currently available 2002 CFS values (Table A.II-2f). After noting the reasonability of our
sector aggregations at the national level, the final USC Sectors and various conversion
bridges were used in NIEMO construction which required the same data conversion
processes at the state level.
Starting in 2001, IMPLAN adopted the NAICS industry codes, while also maintaining
matches to the BEA commodity codes. Owing to its basis in the NAICS codes, the remaining
IMPLAN sectors are relatively easily aggregated into 19 NAICS two-digit service sectors
added to the commodity aggregations and redefined as aggregation codes in Table A.II-2a,
which were combined with the 19 USC service Sectors.
In addition to the IMPLAN service sector aggregation, the reconciliation to the final 29
commodity USC sectors was accomplished by further manual adjustments,
producer/purchaser dollar value adjustments, and minor sector (SCTG 16 and 43)
corrections at the national level. Minor manual adjustments were based on judgments and
183

using sector names. A detailed sector bridge table for IMPLAN->BEA->SCTG->USC for all
commodity and service sectors is shown in Appendix 2 of Park et al (2006).

184

Table A. II-2a Selected IMPLAN Sector Aggregation to Two-Digit NAICS Codes
IMPLAN
2001 (509)
Aggregation
Codes
Aggregation Descriptions NAICS 2 digit codes
19 16 Oil Extraction -
33-45 53 Construction Construction
390 54 Wholesale Trade Wholesale Trade
401-412 57 Retail Trade Retail Trade
30-32 52 Utility Utility
391-397 55 Transportation Transportation
398-400 56 Warehousing Warehousing
416,420-424 58
Broadcasting and information
services
Part Information (Publishing,
Motion pictures, and Recording
(IMPLAN 413-415, 417-419) are
excluded in this sector and
included in Commodity Flows)
425-430 59 Finance and Insurance Finance and Insurance
431-436, 509 60 Real estate and rental and leasing Real estate and rental and leasing
437-450 61
Professional, Scientific, and
Technical services
Professional, Scientific, and
Technical services
451 62
Management of companies and
enterprises
Management of companies and
enterprises
452-460 63
Administrative support and waste
management
Administrative support and waste
management
461-462 64 Education Services Education Services
463-470 65 Health Care and Social Assistances
Health Care and Social
Assistances
471-478 66 Arts, Entertainment, and Recreation
Arts, Entertainment, and
Recreation
479-481 67 Accommodation and Food services
Accommodation and Food
services
495-499,
503-506
68 Public administration Public administration
482-494 69
Other services (except public
administration)
Other services (except public
administration)
18, 27-29,
243
69
Support activities (18=Agriculture
and forestry, 27-29=Mining) and
Etc. (243=Machine shops)
-
500, 507-508 99 Unknown commodity Unknown
All other
IMPLAN
sectors
1

1-15, 17-41,
43
SCTG 1-15, 17-41, 43 -
0-1 Table A. II-1a Selected IMPLAN Sector Aggregation to Two-Digit NAICS Codes
Notes: Detail sector bridge between IMPLAN and SCTG in this Aggregation Codes is shown in
Appendix 2 of Park et al. (2006)
185

Table A. II-2b: Aggregated 1997 BEA Benchmark: Producer/Purchaser Values and Ratios
SCTG V_PRO P_PRO V_PUR P_PUR P-Ratios (=V_PRO/V_PUR)
1 15,217 0.35% 15,346 0.27% 0.99
2 44,068 1.01% 57,901 1.01% 0.76
3 93,060 2.13% 135,909 2.37% 0.68
4 33,075 0.76% 44,105 0.77% 0.75
5 157,516 3.60% 195,765 3.41% 0.80
6 72,776 1.67% 101,044 1.76% 0.72
7 174,908 4.00% 240,070 4.18% 0.73
8 77,799 1.78% 117,648 2.05% 0.66
9 40,018 0.92% 62,357 1.09% 0.64
10 2,686 0.06% 3,914 0.07% 0.69
11 1,967 0.05% 3,181 0.06% 0.62
12 260 0.01% 395 0.01% 0.66
13 4,290 0.10% 8,313 0.14% 0.52
14 9,375 0.21% 10,019 0.17% 0.94
15 23,597 0.54% 35,128 0.61% 0.67
17 83,541 1.91% 146,500 2.55% 0.57
18 338 0.01% 593 0.01% 0.57
19 26,510 0.61% 33,865 0.59% 0.78
20 56,732 1.30% 71,607 1.25% 0.79
21 121,089 2.77% 171,889 2.99% 0.70
22 369 0.01% 427 0.01% 0.86
23 130,938 3.00% 177,939 3.10% 0.74
24 185,554 4.25% 256,787 4.47% 0.72
25 20,189 0.46% 22,050 0.38% 0.92
26 100,980 2.31% 128,085 2.23% 0.79
27 57,525 1.32% 70,262 1.22% 0.82
28 20,104 0.46% 27,944 0.49% 0.72
29 116,818 2.67% 150,008 2.61% 0.78
30 264,847 6.06% 424,046 7.39% 0.62
31 94,800 2.17% 118,661 2.07% 0.80
32 188,172 4.31% 223,302 3.89% 0.84
33 173,656 3.97% 208,604 3.63% 0.83
34 361,183 8.26% 433,471 7.55% 0.83
35 691,944 15.83% 820,069 14.28% 0.84
36 558,530 12.78% 661,763 11.53% 0.84
37 107,897 2.47% 111,758 1.95% 0.97
38 80,347 1.84% 107,100 1.87% 0.75
39 60,851 1.39% 99,511 1.73% 0.61
40 100,010 2.29% 213,248 3.71% 0.47
41 7,432 0.17% 21,146 0.37% 0.35
43 -- -- -- -- --
99 9,251 0.21% 9,251 0.16% 1.00
ALL 4,370,221 100% 5,740,983 100% 0.76
0-2 Table A. II-1b: Aggregated 1997 BEA Benchmark: Producer/Purchaser Values and Ratios
Data source: BEA NDN-0307 data (http://www.bea.gov/bea/dn2/iedguide.htm#IO)
Note: V_Pro=Producer’s Value, P_Pro=Proportions of V_Pro,
V_Pur=Purchaser’s Value, P_Pro=Proportions of V_Pur
186

Dollar-value comparisons by aggregated sectors make it easier for data reconciliation to
be tested. Producer/purchaser dollar value adjustments were conducted because the
IMPLAN data uses producer values, while the CFS data are based on purchaser values
which include transportation costs, wholesale markups, and retail markups besides the
producer values. Table A.II-2b shows dollar value adjustments of all the CFS data using
price ratios (= producer prices/purchaser prices) at the sector-level. These were adjusted by
calibrating producer/purchaser ratios aggregated to the two-digit SCTG sectors following the
conversion steps shown in Appendix A1, utilizing producer and purchaser values at the BEA
five-digit level from BEA NDN-0307 data at the BEA website. This step allows the
estimated commodity flows in terms of producer values to be converted to flows in terms of
purchaser values consistent with the CFS reports. Any estimated flows can be compared with
CFS flows using these conversion ratios (the P-ratios; for SCTG 43 is assumed as equal to
one due to its unavailability).
Special adjustments were required for two of the SCTG sectors, CFS ‘Mixed Freight’
(SCTG 43), and ‘Oil and Gas Extraction’ (SCTG 16). The CFS Mixed Freight sector has no
corresponding BEA or IMPLAN commodity sectors. Using the labels and definitions that
accompany the CFS Mixed Freight sector, we assumed that the national value for SCTG 43
from the 2002 CFS preliminary version to be the same as similarly named subsectors' values
of Wholesale Trade in 2002 the Economic Census. Table A.II-2c shows the subsectors of the
Economic Census whose names roughly correspond to SCTG Sector 43.
Based on this definition, our adjusted subsector value of the wholesale value for 2001
IMPLAN is shown in Table A.II-2d. Instead of using $733 billions shown in Table A.II-2c
we substituted SCTG Sector 43 for the wholesale subsector value of the 2002 Economic
187

Census with sector values from the 2002 CFS. It was better to use 2002 CFS values as the
subsector values, because the subsectors of the 2002 Economic Census wholesale still do not
reflect the SCTG 43 sector entirely. Therefore, by showing that this SCTG Sector 43 is made
up of subsectors of wholesale, we assume that the value of 2002 SCTG 43 can be that
subsector’s value of wholesale trade. In Table A.II-2d, the relevant subsectors’ value in the
2002 Economic Census is estimated at 19.6 percent. This was used to adjust 2001 IMPLAN
Wholesale total value ($875.3 million) for the following results.
Table A. II-2c. Sales Values Matched to SCTG 43 from 2002 Wholesale Economic Census
2002 NAICS code and
Description
CFS Mixed Freight Code and
Description Sales Value($1,000)
4244. Grocery and related
products merchant
wholesalers
43991. Items(including food)
for Grocery and Convenience
stores

43992. Supplies and food for
restaurant and fast food chains
616,389,515

4237. Hardware, and
plumbing and heating
equipment and supplies
merchant wholesalers

43992. Hardware or plumbing
supplies
82,578,288

42412. Stationery and
office supplies merchant
wholesalers
43994. Office Supplies 34,218,647
-- 43999. Miscellaneous --
Total 733,186,450
0-3 Table A. II-1c. Sales Values Matched to SCTG 43 from 2002 Wholesale Economic Census
*Source: 2002 Economic Census, Industry Series Reports, Wholesale Trade from
“http://www.census.gov/econ/census02/guide/INDRPT42.HTM”
188

Table A. II-2d. Calculation of SCTG 43 Value From 2001 IMPLAN Using Economic Census

2002 Economic
Census
2002 CFS 2001 IMPLAN
Wholesale (Sales) Value

4,376,337,051

Purchaser
Values
Subsectors of Wholesale
Trade or SCTG 43

858,320,000
858,320,000
Adjustment
Ratio**

0.2074

Adjusted Wholesale
Value

907,457,995
875,318,813
Derived
Producer
Values
Adjusted Subsectors’
Value of Wholesale
value

177,977,459
171,674,082
Subsectors’
Proportion

0.196
0.196
0-4Table A. II--1d. Calculation of SCTG 43 Value From 2001 IMPLAN Using Economic Census
* Unit: $1,000
**Source: 1987-1995 average (Gross Margin/Sales price) Ratio from "Annual Benchmark
Report for Wholesale Trade: January 1987 through February 1997"

Here follows a summary of the steps followed to derive the adjusted producer value of
subsector of wholesale trade matched to SCTG 43: (i) adjust dollar value to producer value
by the average of the 1987-1995 ratio of (gross margin)/(sales price) = 20.7percent from the
"Annual Benchmark Report For Wholesale Trade" (U.S. Bureau of the Census, 1997); (ii)
calculate subsectors ratio (Adjusted Wholesale Value/ Adjusted Subsectors of Wholesale
Value=19.6 percent) from the derived producer values cell, and (iii) multiply the calculated
ratio by the wholesale sector output value from 2001 IMPLAN (875.3 million dollars). From
all these steps, we get $178 million as our estimate of CFS Sector 43 and $172 million
correspondently estimated to be our estimate IMPLAN's mixed freight component of the
wholesale sector.
189

Two of the findings in Table A.II-2e and Table A.II-2f show the results of aggregating
the 2001 IMPLAN sectors to the 43 SCTG sectors for 1997 and 2002. In order to improve
the correspondence of IMPLAN sectors to SCTG sectors, we aggregated to 29 USC sectors
from the 43 SCTG sectors. During aggregation, the SCTG ‘Oil and Gas Extraction’ sector
(#16) which was removed from CFS due to the problem of an overwhelming number of
shipments; we were able to include it as USC Sector 10, from IMPLAN data. Based on the
all the itemized procedures, the final USC Sectors are shown in Table A.II-2g.

190

Table A. II-2e. Comparison of Aggregated 2001 IMPLAN with 1997_CFS: U.S. Total, Including SCTG16
2001_IMPLAN_SCTG 1997_CFS BEA Revised_1997_CFS Ratio
SCTG V1* P1** V2 P2 P_Ratio V5(=V2xP_Ratio) P5 V1/V5 P1/P5
1 16,884 0.279% 6,173 0.089% 0.99 6,121 0.120% 2.758 2.398
2 39,472 0.653% 59,642 0.859% 0.76 45,393 0.893% 0.870 0.756
3 91,064 1.506% 102,344 1.474% 0.68 70,078 1.378% 1.299 1.130
4 45,911 0.759% 66,848 0.963% 0.75 50,130 0.986% 0.916 0.796
5 175,594 2.903% 183,784 2.647% 0.80 147,876 2.908% 1.187 1.032
6 86,329 1.427% 109,854 1.582% 0.72 79,122 1.556% 1.091 0.949
7 302,706 5.005% 346,379 4.988% 0.73 252,361 4.962% 1.199 1.043
8 80,602 1.333% 87,932 1.266% 0.66 58,148 1.143% 1.386 1.205
9 54,172 0.896% 56,394 0.812% 0.64 36,191 0.712% 1.497 1.301
10 2,818 0.047% 2,726 0.039% 0.69 1,871 0.037% 1.506 1.309
11 2,374 0.039% 4,279 0.062% 0.62 2,646 0.052% 0.897 0.780
12 5,191 0.086% 11,508 0.166% 0.66 7,572 0.149% 0.686 0.596
13 9,758 0.161% 11,329 0.163% 0.52 5,847 0.115% 1.669 1.451
14 11,054 0.183% 12,605 0.182% 0.94 11,794 0.232% 0.937 0.815
15 24,862 0.411% 25,486 0.367% 0.67 17,120 0.337% 1.452 1.263
16 197,809 3.271% -- -- -- -- -- -- --
17 114,753 1.897% 217,051 3.126% 0.57 123,772 2.434% 0.927 0.806
18 114,753 1.897% 94,309 1.358% 0.57 53,779 1.057% 2.134 1.855
19 27,996 0.463% 74,900 1.079% 0.78 58,633 1.153% 0.477 0.415
20 104,099 1.721% 159,623 2.299% 0.79 126,464 2.487% 0.823 0.716
21 174,086 2.878% 224,448 3.232% 0.70 158,114 3.109% 1.101 0.957
22 22,231 0.368% 27,334 0.394% 0.86 23,606 0.464% 0.942 0.819
23 159,819 2.643% 209,487 3.017% 0.74 154,153 3.031% 1.037 0.901
24 231,896 3.834% 278,832 4.015% 0.72 201,484 3.962% 1.151 1.001
25 15,593 0.258% 15,129 0.218% 0.92 13,852 0.272% 1.126 0.979
26 106,688 1.764% 126,426 1.821% 0.79 99,672 1.960% 1.070 0.931
27 74,409 1.230% 106,578 1.535% 0.82 87,257 1.716% 0.853 0.741
28 81,685 1.351% 98,347 1.416% 0.72 70,753 1.391% 1.155 1.004
29 133,501 2.207% 260,327 3.749% 0.78 202,729 3.986% 0.659 0.573
30 292,878 4.843% 379,161 5.460% 0.62 236,813 4.656% 1.237 1.075
31 113,064 1.869% 109,197 1.573% 0.80 87,240 1.715% 1.296 1.127
32 169,411 2.801% 285,690 4.114% 0.84 240,745 4.734% 0.704 0.612
33 200,391 3.313% 227,182 3.272% 0.85 193,294 3.801% 1.037 0.901
34 433,014 7.160% 417,103 6.007% 0.83 347,545 6.833% 1.246 1.083
35 844,544 13.964% 869,675 12.524% 0.84 733,800 14.428% 1.151 1.001
36 654,570 10.823% 570,981 8.223% 0.84 481,910 9.475% 1.358 1.181
37 143,113 2.366% 129,185 1.860% 0.97 124,723 2.452% 1.147 0.998
38 160,050 2.646% 157,946 2.275% 0.75 118,491 2.330% 1.351 1.174
39 92,277 1.526% 97,255 1.401% 0.61 59,471 1.169% 1.552 1.349
40 225,430 3.727% 420,883 6.061% 0.47 197,389 3.881% 1.142 0.993
41 18,578 0.307% 32,714 0.471% 0.35 11,498 0.226% 1.616 1.405
43 171,674 2.839% 230,415 3.318% 0.20 49,947 0.982% 3.437 2.988
99 20,735 0.343% 36,524 0.526% 1.00 36,524 0.718% 0.568 0.494
ALL 6,047,838 100% 6,943,985 100% 0.77 5,085,927 100% 1.150 1
0-5 Table A. II-1e. Comparison of Aggregated 2001 IMPLAN with 1997_CFS: U.S. Total, Including SCTG16
*Units: (million$),
**(Each SCTG sector value)x100/ (ALL value).
191

Table A. II-2f. Comparison of Aggregated 2001 IMPLAN with 2002_CFS: U.S. Total, Including SCTG16
2001_IMPLAN_SCTG 2002_CFS BEA Revised_2002_CFS Ratio
SCTG V1* P1** V3 P3 P_Ratio V4(=V3xP_Ratio) P4 V1/V4 P1/P4
1 16,884 0.279% 7,200 0.085% 0.99 7,139 0.121% 2.365 2.299
2 39,472 0.653% 55,927 0.659% 0.76 42,565 0.724% 0.927 0.902
3 91,064 1.506% 129,890 1.531% 0.68 88,939 1.513% 1.024 0.996
4 45,911 0.759% 55,251 0.651% 0.75 41,433 0.705% 1.108 1.077
5 175,594 2.903% 204,869 2.415% 0.80 164,841 2.803% 1.065 1.036
6 86,329 1.427% 119,718 1.411% 0.72 86,226 1.466% 1.001 0.973
7 302,706 5.005% 362,312 4.271% 0.73 263,970 4.489% 1.147 1.115
8 80,602 1.333% 115,772 1.365% 0.66 76,558 1.302% 1.053 1.024
9 54,172 0.896% 77,163 0.910% 0.64 49,519 0.842% 1.094 1.064
10 2,818 0.047% 2,451 0.029% 0.69 1,682 0.029% 1.675 1.629
11 2,374 0.039% 4,611 0.054% 0.62 2,851 0.048% 0.832 0.809
12 5,191 0.086% 12,643 0.149% 0.66 8,319 0.141% 0.624 0.607
13 9,758 0.161% 12,680 0.149% 0.52 6,544 0.111% 1.491 1.450
14 11,054 0.183% 15,741 0.186% 0.94 14,729 0.250% 0.751 0.730
15 24,862 0.411% 24,085 0.284% 0.67 16,179 0.275% 1.537 1.494
16 197,809 3.271% -- -- -- -- -- -- --
17 114,753 1.897% 233,563 2.753% 0.57 133,188 2.265% 0.862 0.838
18 114,753 1.897% 109,618 1.292% 0.57 62,509 1.063% 1.836 1.785
19 27,996 0.463% 74,693 0.880% 0.78 58,471 0.994% 0.479 0.466
20 104,099 1.721% 152,069 1.792% 0.79 120,479 2.049% 0.864 0.840
21 174,086 2.878% 426,753 5.030% 0.70 300,630 5.113% 0.579 0.563
22 22,231 0.368% 34,079 0.402% 0.86 29,431 0.501% 0.755 0.734
23 159,819 2.643% 234,355 2.762% 0.74 172,452 2.933% 0.927 0.901
24 231,896 3.834% 343,386 4.048% 0.72 248,130 4.220% 0.935 0.909
25 15,593 0.258% 5,718 0.067% 0.92 5,235 0.089% 2.978 2.896
26 106,688 1.764% 140,006 1.650% 0.79 110,379 1.877% 0.967 0.940
27 74,409 1.230% 102,406 1.207% 0.82 83,842 1.426% 0.887 0.863
28 81,685 1.351% 105,890 1.248% 0.72 76,180 1.296% 1.072 1.043
29 133,501 2.207% 136,886 1.614% 0.78 106,600 1.813% 1.252 1.218
30 292,878 4.843% 506,992 5.976% 0.62 316,653 5.385% 0.925 0.899
31 113,064 1.869% 143,106 1.687% 0.80 114,330 1.944% 0.989 0.962
32 169,411 2.801% 253,678 2.990% 0.84 213,769 3.635% 0.792 0.771
33 200,391 3.313% 234,922 2.769% 0.85 199,880 3.399% 1.003 0.975
34 433,014 7.160% 509,477 6.005% 0.83 424,514 7.219% 1.020 0.992
35 844,544 13.964% 948,049 11.175% 0.84 799,929 13.604% 1.056 1.027
36 654,570 10.823% 735,730 8.672% 0.84 620,959 10.560% 1.054 1.025
37 143,113 2.366% 162,984 1.921% 0.97 157,354 2.676% 0.909 0.884
38 160,050 2.646% 222,042 2.617% 0.75 166,576 2.833% 0.961 0.934
39 92,277 1.526% 135,049 1.592% 0.61 82,582 1.404% 1.117 1.086
40 225,430 3.727% 404,683 4.770% 0.47 189,791 3.228% 1.188 1.155
41 18,578 0.307% 49,307 0.581% 0.35 17,330 0.295% 1.072 1.042
43 171,674 2.839% 858,320 10.117% 0.20 177,977 3.027% 0.965 0.938
99 20,735 0.343% 19,588 0.231% 1.00 19,588 0.333% 1.059 1.029
ALL 6,047,838 100% 8,483,662 100% 0.77 5,880,253 100% 1.028 1
Table A. II-1f. Comparison of Aggregated 2001 IMPLAN with 2002_CFS: U.S. Total Including SCTG16
*Units: (million$),
**(Each SCTG sector value)x100/ (ALL value).
192

Table A. II-2g. Definitions of USC Two-Digit Sectors
Classification USC Description SCTG NAICS
USC01 Live animals and live fish & Meat fish seafood and their preparations (1+5) 11 31
USC02 Cereal grains & Other agricultural products except for Animal Feed (2+3) 11,31
USC03 Animal feed and products of animal origin, n.e.c. 4 11,31
USC04 Milled grain products and preparations, and bakery products 631
USC05 Other prepared foodstuffs and fats and oils 7 11,31
USC06 Alcoholic beverages 8 31,32
USC07 Tobacco products 9 11,31
USC08
Nonmetallic minerals (Monumental or building stone, Natural sands, Gravel and crushed stone,
n.e.c.) (10~13) 21,32
USC09 Metallic ores and concentrates 14 21,32
USC10 Coal and petroleum products (Coal and Fuel oils, n.e.c.) (15~19) 21,32
USC11 Basic chemicals 20 32
USC12 Pharmaceutical products 21 32,33
USC13 Fertilizers 22 32
USC14 Chemical products and preparations, n.e.c. 23 31,32
USC15 Plastics and rubber 24 31,32,33
USC16 Logs and other wood in the rough & Wood products (25+26) 11,32
USC17 Pulp, newsprint, paper, and paperboard & Paper or paperboard articles (27+28) 32
USC18 Printed products 29 32,51
USC19 Textiles, leather, and articles of textiles or leather 30 11,31,32,33
USC20 Nonmetallic mineral products 31 32,33
USC21 Base metal in primary or semi-finished forms and in finished basic shapes 32 33
USC22 Articles of base metal 33 33
USC23 Machinery 34 32,33
USC24 Electronic and other electrical equipment and components, and office equipment 35 32,33,51
USC25 Motorized and other vehicles (including parts) 36 32,33
USC26 Transportation equipment, n.e.c. 37 33
USC27 Precision instruments and apparatus 38 33
USC28 Furniture, mattresses and mattress supports, lamps, lighting fittings, and illuminated signs 39 33
Commodity
Sectors
USC29 Miscellaneous manufactured products, Scrap, Mixed freight, and Commodity unknown (40~99) 11,31,32,33
USC30 Utility 22
USC31 Construction 23
USC32 Wholesale Trade 42
USC33 Transportation 48
USC34 Postal and Warehousing 49
USC35 Retail Trade (44+45)
USC36 Broadcasting and information services* (515~519)
USC37 Finance and Insurance 52
USC38 Real estate and rental and leasing 53
USC39 Professional, Scientific, and Technical services 54
USC40 Management of companies and enterprises 55
USC41 Administrative support and waste management 56
USC42 Education Services 61
USC43 Health Care and Social Assistances 62
USC44 Arts, Entertainment, and Recreation 71
USC45 Accommodation and Food services 72
USC46 Public administration 92
Non-
Commodity
(Service)
Sectors
USC47 Other services except public administration** 81
0-6 Table A. II-1g. Definitions of USC Two-Digit Sectors
*Publishing, Motion pictures, and Recording (IMPLAN 413-415, 417-419, or NAICS 511~512) are
excluded in this sector and included in Commodity Sectors
**USC47 includes NAICS 81plus Support activities (18=Agriculture and Forestry, 27-29=Mining)
and Etc. (243=Machine Shops) in IMPLAN
193

Table A. II-3. Trade flows matrix between States for USC sector 15 (=SCTG sector 24) from 1997 CFS
($M.)
AL AK AZ AR CA CO CT DE DC FL
V S V S V S V S V S V S V S V S V S V S
AL 824 - - 1 - 2 - 2 220 - - 2 - 2 - 2 - 1 201 -
AK
- 1 77 - - 1 - 1 - 2 - 1 - 1 - 1 - 1 - 2
AZ
- 2 - 2 1014 - - 2 284 - 25 - - 2 - 2 - 2 9 -
AR
27 - - 2 - 2 785 - 140 - 21 - - 2 - 2 - 1 108 -
CA
76 - - 2 610 - 104 - 12557 - 228 - 92 - 7 - - 2 297 -
CO
- 2 - 1 14 - 11 - 107 - 977 - - 2 - 2 - 2 13 -
CT
6 - - 2 - 2 6 - 262 - - 2 484 - - 2 - 2 87 -
DE
- 2 - 1 6 - - 2 53 - - 2 - 2 14 - - 2 25 -
DC
- 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 2 - 1
FL
109 - - 2 19 - 24 - 122 - 17 - 15 - - 2 - 2 4234 -
GA
475 - - 2 65 - - 2 359 - 19 - 44 - 26 - - 2 785 -
HI
- 1 - 1 - 1 - 1 - 2 - 1 - 1 - 1 - 1 - 2
ID
- 2 - 2 - 2 - 2 44 - - 2 - 2 - 2 - 1 - 2
IL
106 - - 2 82 - 105 - 777 - 50 - - 2 - 2 - 2 212 -
IN
49 - - 2 23 - 73 - 157 - 27 - 26 - 26 - - 2 118 -
IA
- 2 - 2 9 - 34 - 101 - 11 - - 2 - 2 - 1 29 -
KS
- 2 - 1 18 - 59 - 207 - 89 - 1 - 2 - - 2 36 -
KY
77 - - 2 7 - 85 - 102 - 9 - 23 - - 2 - 2 93 -
LA
254 - - 2 15 - 350 - 241 - 12 - - 2 - 2 - 1 114 -
ME
- 2 - 1 - 2 - 2 27 - - 2 9 - - 1 - 1 - 2
MD
- 2 - 1 - 2 - 2 - 2 - 2 4 - 49 - 59 - 25 -
MA
10 - - 2 - 2 22 - 225 - 16 - 214 - - 2 - 2 68 -
MI
39 - - 1 74 - - 2 383 - - 2 9 - - 2 - 2 143 -
MN
9 - - 1 36 - 11 - 269 - 75 - 18 - - 2 - 2 56 -
MS
66 - - 2 - 2 33 - 125 - 8 - - 2 - 2 - 1 75 -
MO
35 - - 2 5 - 101 - 88 - - 2 5 - - 2 - 2 - 2
MT
- 2 - 1 - 2 - 2 - 2 - 2 - 1 - 2 - 1 - 2
NE
6 - - 2 5 - - 2 37 - - 2 - 2 - 2 - 2 25 -
NV
- 2 - 2 19 - - 2 251 - 8 - - 2 - 2 - 1 - 2
NH
- 2 - 1 - 2 - 2 - 2 - 1 - 2 - 2 - 1 - 2
NJ
74 - - 2 - 2 57 - 611 - 71 - 243 - - 2 - 2 284 -
NM
- 1 - 2 - 2 - 2 18 - 11 - - 2 - 1 - 1 - 2
NY
49 - - 2 - 2 36 - 617 - 15 - 248 - 31 - - 2 136 -
NC
132 - - 1 55 - 50 - 422 - 17 - 58 - 79 - - 2 286 -
ND
- 2 - 1 - 2 - 2 - 2 3 - - 2 - 1 - 1 - 2
OH
301 - - 2 152 - 174 - 609 - 80 - 93 - 49 - - 2 366 -
OK
12 - - 2 28 - 45 - 241 - 34 - 6 - - 2 - 2 70 -
OR
1 - - 2 - 2 - 2 280 - 18 - - 2 - 2 - 2 4 -
PA
123 - - 2 48 - 51 - 686 - 42 - 153 - - 2 17 - 340 -
RI
- 2 - 2 - 2 10 - - 2 1 - 26 - - 2 - 1 4 -
SC
126 - - 2 31 - 23 - 115 - - 2 30 - 90 - - 1 99 -
SD
- 2 - 2 - 2 - 2 16 - - 2 - 2 - 1 - 1 10 -
TN
249 - - 2 35 - 213 - 283 - 155 - - 2 - 2 - 1 166 -
TX
300 - - 2 180 - 492 - 1820 - 322 - 47 - - 2 - 1 411 -
UT
- 2 - 2 16 - - 2 123 - - 2 - 2 - 1 - 1 4 -
VM
- 2 - 2 - 2 - 2 5 - - 2 6 - - 2 - 1 - 2
VA
- 2 - 2 10 - 21 - 200 - 39 - 24 - - 2 - 2 83 -
WA
- 2 34 - 28 - - 2 330 - 7 - - 2 - 1 - 1 - 2
WV
- 2 - 1 - 2 - 2 - 2 - 2 - 2 - 2 - 1 5 -
WI
30 - - 2 49 - 93 - 507 - 44 - - 2 - 2 - 2 141 -
WY
- 2 - 1 - 1 - 1 - 2 - 2 - 1 - 1 - 1 - 1
D
T
j
3809 - 148 - 3106 - 3775 - 24374 - 2632 - 2596 - 1715 - 113 - 9291 -
0-7 Table A. II-2. Trade flows matrix between States for USC sector 15 (=SCTG sector 24) from 1997 CFS
($M.)
194

(Table A.II-3:Continued)
GA HI ID IL IN IA KS KY LA ME
V S V S V S V S V S V S V S V S V S V S
AL 305 - - 2 - 2 154 - 75 - - 2 - 2 100 - 42 - - 2
AK
- 1 - 1 - 1 - 1 - 1 - 2 - 2 - 1 - 1 - 1
AZ
- 2 - 2 4 - - 2 9 - 1 - 3 - - 2 - 2 - 2
AR
162 - - 1 - 2 115 - 128 - 20 - - 2 80 - 63 - - 2
CA
148 - 66 - 80 - 416 - 206 - 52 - 53 - 66 - 38 - - 2
CO
26 - - 2 - 2 42 - 5 - - 2 - 2 - 2 - 2 - 2
CT
61 - - 2 - 1 53 - 44 - - 2 - 2 20 - 8 - 11 -
DE
24 - - 1 - 2 6 - 19 - - 2 - 2 - 2 - 2 - 2
DC
- 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1
FL
260 - - 2 - 2 105 - 70 - 15 - 19 - 60 - 25 - - 2
GA
3085 - - 2 - 2 259 - 73 - 34 - 13 - 82 - 98 - - 2
HI
- 1 186 - - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1
ID
- 2 - 2 242 - - 2 - 2 - 2 - 2 - 2 - 2 - 1
IL
753 - - 2 - 2 6895 - - 2 427 - 201 - - 2 61 - - 2
IN
189 - - 1 - 2 750 - 2129 - 118 - 71 - 336 - 19 - 26 -
IA
108 - - 1 4 - 319 - 153 - 813 - 66 - 164 - 37 - - 2
KS
59 - - 1 - 2 154 - 158 - 59 - 750 - 28 - - 2 - 2
KY
249 - - 2 - 2 197 - 274 - - 2 48 - 865 - 37 - - 2
LA
134 - - 1 - 2 333 - 173 - 162 - 22 - 286 - 1386 - - 2
ME
- 2 - 1 - 2 29 - - 2 - 2 - 2 - 2 - 2 135 -
MD
8 - - 1 - 2 50 - - 2 - 2 1 - 6 - - 2 - 2
MA
112 - - 2 - 2 214 - 121 - 7 - 21 - 51 - 8 - 235 -
MI
248 - - 2 - 2 382 - 593 - 33 - 36 - 125 - 46 - - 2
MN
49 - - 2 - 2 326 - 179 - 167 - 29 - - 2 - 2 - 2
MS
197 - - 2 - 1 - 2 88 - 15 - - 2 42 - 75 - - 2
MO
111 - - 2 - 2 380 - 87 - 93 - 164 - 63 - - 2 - 2
MT
- 2 - 1 - 2 - 2 - 1 - 2 - 2 - 1 - 2 - 2
NE
11 - - 1 10 - 76 - - 2 95 - 16 - 52 - - 2 - 2
NV
- 2 - 2 7 - 1 - - 2 - 2 - 2 - 2 - 2 - 2
NH
- 2 - 2 - 1 - 2 - 2 - 2 - 2 - 2 - 2 - 2
NJ
181 - - 2 - 2 356 - 199 - 23 - 98 - 141 - 41 - 29 -
NM
- 2 - 1 - 2 - 2 - 2 - 2 - 2 - 1 - 2 - 1
NY
112 - 3 - - 2 272 - 192 - 15 - 5 - 38 - 17 - - 2
NC
787 - - 2 - 2 184 - 88 - 26 - 32 - 195 - 37 - 15 -
ND
- 2 - 1 - 2 2 - - 2 4 - - 2 - 2 - 2 - 2
OH
710 - - 2 18 - 924 - 935 - 231 - 166 - 714 - 55 - 39 -
OK
150 - - 2 - 2 155 - 82 - - 2 58 - 13 - 10 - - 2
OR
7 - - 2 39 - 12 - - 2 - 2 - 2 - 2 - 2 - 2
PA
265 - - 2 26 - 400 - 693 - 128 - 109 - 105 - 83 - 24 -
RI
- 2 - 1 - 2 21 - - 2 - 2 - 2 1 - - 2 - 2
SC
446 - - 2 - 2 192 - 48 - 5 - - 2 114 - - 2 56 -
SD
- 2 - 2 - 2 7 - 16 - 7 - - 2 - 2 2 - - 2
TN
1152 - - 1 - 2 370 - 144 - 111 - 36 - 267 - 103 - 50 -
TX
986 - - 2 27 - 1292 - 659 - 214 - 262 - 278 - 282 - - 2
UT
- 2 - 2 65 - 14 - - 2 3 - 4 - - 2 - 2 - 2
VM
- 2 - 2 - 2 2 - - 2 - 2 - 2 - 2 - 2 - 2
VA
259 - - 2 - 2 160 - 98 - - 2 54 - 156 - - 2 - 2
WA
17 - 6 - 87 - 43 - - 2 - 2 15 - - 2 - 2 - 2
WV
18 - - 1 - 2 - 2 - 2 - 2 - 2 29 - - 2 - 2
WI
110 - - 2 8 - 864 - 282 - 115 - 46 - 99 - 34 - - 2
WY
- 1 - 1 - 2 - 2 - 2 - 2 - 2 - 2 - 1 - 1
D
T
j
11559 - 320 - 830 - 16948 - 9600 - 3233 - 2687 - 5384 - 2803 - 1004 -
195

(Table A.II-3: Continued)

MD MA MI MN MS MO MT NE NV NH
V S V S V S V S V S V S V S V S V S V S
AL - 2 16 - 30 - - 2 45 - 74 - - 2 - 2 - 2 - 2
AK - 1 - 1 - 2 - 1 - 1 - 1 - 1 - 1 - 1 - 1
AZ 4 - 4 - - 2 15 - - 2 - 2 - 2 - 2 25 - - 2
AR - 2 9 - 78 - 18 - 41 - 64 - - 2 11 - - 2 - 2
CA 40 - 86 - 163 - 128 - 67 - 129 - 30 - 49 - 247 - - 2
CO 6 - 14 - 7 - 37 - 1 - 19 - - 2 - 2 - 2 - 2
CT 5 - 204 - 31 - - 2 - 2 11 - - 2 8 - - 2 84 -
DE 65 - 39 - - 2 - 2 9 - - 2 - 2 - 1 - 2 - 2
DC - 2 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1
FL 24 - 36 - 89 - 25 - 23 - 38 - - 2 15 - 2 - 4 -
GA 59 - 57 - 223 - - 2 152 - 108 - - 2 11 - 10 - - 2
HI - 2 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1
ID - 2 - 2 - 2 - 2 - 2 - 2 7 - - 1 - 2 - 2
IL 563 - 172 - 1463 - 544 - 88 - 599 - 19 - 208 - - 2 11 -
IN 34 - 50 - 648 - 99 - 21 - 281 - - 2 78 - 22 - 4 -
IA 7 - 18 - 228 - 223 - 32 - 285 - 21 - 133 - - 2 - 2
KS - 2 7 - 82 - 48 - - 2 241 - - 2 81 - - 2 - 2
KY - 2 - 2 297 - 36 - 61 - 64 - - 2 2 - 8 - 1 -
LA 6 - 62 - - 2 74 - 258 - 50 - - 2 - 2 - 2 - 2
ME - 2 36 - 13 - 6 - - 2 - 2 - 2 - 1 - 2 12 -
MD 798 - 66 - 35 - - 2 - 2 11 - - 2 - 2 - 1 - 2
MA 88 - 2080 - 116 - 68 - 28 - 50 - - 2 - 2 - 2 - 2
MI 62 - 93 - 4766 - 106 - 18 - 270 - - 2 20 - - 2 20 -
MN 22 - - 2 115 - 2019 - - 2 157 - 9 - 107 - 5 - - 2
MS 3 - 26 - 35 - - 2 668 - 43 - - 2 59 - - 2 - 2
MO 26 - 18 - 169 - 71 - - 2 1482 - - 2 40 - 1 - 6 -
MT - 2 - 2 - 2 - 2 - 2 - 2 87 - - 2 - 1 - 2
NE - 2 10 - 73 - 35 - 2 - 58 - - 2 246 - - 2 - 2
NV - 2 - 2 - 2 - 2 - 2 - 2 - 2 - 2 146 - - 2
NH - 2 - 2 - 2 - 2 - 2 - 2 - 2 1 - - 2 265 -
NJ 404 - 518 - 168 - 197 - 49 - 157 - - 2 - 2 - 2 - 2
NM - 2 - 2 - 2 - 2 - 2 - 1 - 2 - 2 - 2 - 1
NY 117 - 178 - 322 - 113 - 16 - 169 - 5 - - 2 - 2 63 -
NC 156 - - 2 239 - 102 - 47 - 189 - 1 - 23 - - 2 - 2
ND - 2 - 2 - 2 29 - - 2 - 2 15 - - 2 - 2 - 1
OH 190 - 255 - 1436 - 258 - 114 - 344 - 10 - 66 - 36 - 25 -
OK 24 - 8 - 143 - - 2 - 2 105 - - 2 - 2 - 2 - 2
OR - 2 - 2 5 - 9 - - 2 12 - 23 - - 2 7 - - 2
PA 391 - 289 - 434 - 138 - 80 - 144 - - 2 34 - - 2 62 -
RI 5 - - 2 - 2 1 - - 2 3 - - 2 - 2 - 2 13 -
SC - 2 51 - 197 - 61 - 48 - - 2 - 2 10 - - 2 56 -
SD - 2 1 - 21 - - 2 - 1 10 - 2 - 11 - - 2 - 2
TN 41 - - 2 635 - 100 - 206 - 155 - - 2 25 - 15 - 14 -
TX 91 - 183 - 790 - 201 - 280 - 405 - 69 - 122 - 138 - 53 -
UT - 2 - 2 7 - 5 - - 2 - 2 7 - - 2 13 - - 2
VM - 2 55 - 2 - 3 - 2 - - 2 - 2 - 1 - 2 14 -
VA 172 - 96 - 191 - 27 - 7 - 57 - - 2 4 - - 2 67 -
WA - 2 10 - 6 - - 2 6 - - 2 - 2 7 - 10 - - 2
WV - 2 - 2 128 - - 2 - 2 - 2 - 2 - 1 - 1 - 2
WI 62 - 44 - 470 - 463 - - 2 108 - - 2 44 - 9 - - 2
WY - 1 - 1 - 2 - 2 - 1 - 2 - 1 - 2 - 1 - 1
D
T
j
3710 - 5731 - 14241 - 5610 - 2743 - 6054 - 530 - 1639 - 1026 - 1312 -
196

(Table A.II-3: Continued)
NJ NM NY NC ND OH OK OR PA RI
V S V S V S V S V S V S V S V S V S V S
AL - 2 - 2 98 - 74 - - 2 393 - - 2 - 2 149 - - 2
AK - 1 - 1 - 2 - 1 - 1 - 1 - 1 - 2 - 1 - 1
AZ - 2 26 - 10 - 5 - - 2 8 - 5 - 21 - - 2 1 -
AR 46 - - 2 29 - 102 - - 2 95 - 140 - - 2 63 - - 2
CA 265 - 32 - 322 - 303 - - 2 312 - 58 - 395 - 200 - - 2
CO 26 - 16 - 14 - 7 - - 2 - 2 6 - 4 - - 2 - 2
CT 232 - - 2 321 - 118 - - 2 - 2 - 2 - 2 77 - - 2
DE 108 - - 2 64 - 21 - - 1 44 - - 2 7 - 37 - 21 -
DC - 1 - 1 - 1 - 2 - 1 - 1 - 1 - 1 - 1 - 1
FL - 2 1 - 112 - 81 - - 2 118 - - 2 - 2 108 - 2 -
GA 78 - - 2 82 - 497 - - 2 387 - 33 - 12 - 197 - - 2
HI - 2 - 1 - 2 - 2 - 1 - 1 - 2 - 1 - 1 - 1
ID - 2 - 2 - 2 - 2 - 2 - 2 - 2 22 - - 2 - 1
IL 534 - 12 - 374 - 330 - 50 - 1140 - 139 - 57 - 629 - - 2
IN 91 - - 2 154 - 129 - - 2 712 - 32 - 7 - 205 - - 2
IA 35 - - 2 64 - 96 - 73 - 177 - 26 - 34 - 82 - - 2
KS 70 - 4 - - 2 31 - - 2 75 - 58 - 13 - 82 - - 2
KY 136 - - 2 63 - 107 - - 2 378 - 14 - 6 - 124 - - 1
LA 123 - - 2 180 - 235 - 37 - 219 - 215 - 17 - 190 - - 2
ME - 2 - 1 - 2 - 2 - 1 8 - - 2 - 1 6 - - 2
MD 98 - - 2 160 - - 2 - 2 41 - - 2 - 2 199 - - 2
MA 338 - - 2 610 - 192 - - 2 263 - 9 - 13 - 191 - 161 -
MI 89 - - 2 74 - 130 - 3 - 818 - 26 - - 2 172 - 2 -
MN 83 - 6 - 64 - - 2 57 - 43 - - 2 - 2 75 - - 2
MS 41 - - 2 33 - 68 - - 2 200 - - 2 - 2 62 - - 2
MO 21 - - 2 - 2 33 - - 2 133 - 91 - 80 - 65 - - 2
MT - 2 - 2 - 2 - 2 - 2 1 - - 1 - 2 - 2 - 1
NE - 2 - 2 24 - 13 - - 2 48 - 4 - 3 - 45 - - 1
NV - 2 1 - - 2 - 2 - 2 - 2 - 2 14 - - 2 - 2
NH 20 - - 1 92 - - 2 - 1 - 2 - 2 - 2 15 - 12 -
NJ 3689 - - 2 1844 - 480 - - 2 522 - 18 - - 2 902 - - 2
NM - 2 197 - - 2 - 2 - 1 - 2 - 1 - 2 - 2 - 2
NY 851 - - 2 3489 - 181 - - 2 515 - 5 - - 2 1022 - 17 -
NC 165 - - 2 - 2 3294 - - 2 474 - 64 - 82 - 313 - - 2
ND - 2 - 2 - 2 - 2 114 - - 2 - 2 - 2 - 2 - 1
OH 525 - - 2 778 - 671 - 12 - 5821 - 85 - 154 - 958 - - 2
OK 36 - - 2 69 - 18 - - 2 155 - 456 - 33 - 76 - - 2
OR - 2 1 - 4 - 4 - - 2 - 2 - 2 830 - 12 - - 2
PA 985 - 6 - 881 - 372 - 13 - 760 - 47 - 57 - 3970 - 61 -
RI 51 - - 2 - 2 - 2 - 1 21 - - 2 - 2 34 - - 2
SC 67 - - 2 190 - 680 - - 2 290 - - 2 - 2 124 - 6 -
SD - 2 - 2 13 - 8 - - 2 - 2 - 2 2 - 6 - - 2
TN 155 - - 2 313 - 225 - - 2 520 - 37 - - 2 227 - 54 -
TX 503 - 207 - 382 - 628 - - 2 987 - 456 - 166 - 674 - 51 -
UT - 2 3 - 7 - 3 - - 2 13 - 3 - 20 - 3 - - 1
VM 13 - - 1 53 - - 2 - 1 23 - - 2 - 2 27 - - 2
VA 283 - - 2 263 - 243 - - 2 423 - 11 - - 2 317 - - 2
WA 16 - - 2 14 - 10 - - 2 40 - - 2 294 - 25 - - 2
WV - 2 - 2 - 2 36 - - 1 232 - - 1 - 2 166 - - 1
WI 107 - - 2 - 2 102 - 21 - 422 - - 2 11 - 185 - - 2
WY - 2 - 2 - 2 - 1 - 1 - 2 - 2 - 2 - 1 - 1
D
T
j
10216 - 670 - 13047 - 9723 - 626 - 17050 - 2281 - 2752 - 12099 - 1001 -
197

(Table A.II-3: Continued)
SC SD TN TX UT VM VA WA WV WI WY O
T
i
V S V S V S V S V S V S V S V S V S V S V S V S
AL 48 - - 2 205 - 326 - - 2 - 2 22 - 4 - - 2 19 - - 2 3869 -
AK - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 2 - 1 - 1 - 1 79 -
AZ - 2 - 2 4 - - 2 - 2 - 2 - 2 15 - - 2 11 - - 2 1789 -
AR 114 - - 2 79 - 355 - - 2 1 - 3 - 15 - - 2 19 - - 2 3114 -
CA 26 - 8 - 119 - 699 - 204 - 7 - 67 - 585 - 5 - 141 - 21 - 19867 -
CO - 2 - 2 - 2 91 - 41 - - 2 - 2 34 - - 1 - 2 27 - 1911 -
CT 6 - - 2 - 2 68 - - 2 47 - - 2 - 2 - 2 34 - - 2 2768 -
DE - 2 - 1 8 - 29 - - 1 - 2 - 2 4 - - 2 30 - - 1 791 -
DC - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 2
FL 59 - - 2 59 - 191 - - 2 - 2 32 - 21 - 4 - 23 - - 1 6385 -
GA 444 - - 2 434 - 407 - 31 - - 1 108 - 24 - - 2 179 - - 2 9658 -
HI - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 - 1 188 -
ID - 2 - 1 - 2 11 - 49 - - 2 - 2 36 - - 1 - 2 5 - 557 -
IL 67 - 42 - 319 - 1451 - 85 - - 2 127 - 146 - - 2 896 - - 2 22300 -
IN - 2 - 2 180 - 285 - - 2 - 2 - 2 40 - 15 - 224 - - 2 7732 -
IA 29 - 55 - 72 - 194 - 6 - - 2 55 - 54 - 3 - 185 - - 2 4148 -
KS 53 - - 2 35 - 277 - 24 - - 2 11 - - 2 7 - 49 - 4 - 3140 -
KY - 2 - 2 379 - 327 - - 2 - 2 64 - 12 - 11 - 46 - - 1 4436 -
LA 88 - - 2 289 - 967 - - 2 - 2 192 - - 2 - 2 121 - - 2 7343 -
ME - 2 - 1 - 2 - 2 - 2 - 2 - 2 - 2 - 1 14 - - 2 479 -
MD 10 - - 1 - 2 32 - - 2 - 2 148 - 3 - - 2 - 2 - 2 2165 -
MA - 2 - 2 146 - 291 - 8 - 123 - 85 - 34 - - 2 85 - - 2 6765 -
MI 37 - - 2 208 - 335 - 9 - - 2 68 - - 2 5 - 135 - - 2 9923 -
MN 7 - 61 - 45 - 146 - - 2 - 2 22 - 35 - - 2 242 - - 2 5073 -
MS 16 - - 1 143 - 176 - 1 - - 2 32 - 16 - - 2 - 2 - 2 2993 -
MO 44 - 7 - 132 - 305 - 10 - - 2 - 2 35 - - 2 55 - - 1 4329 -
MT - 2 - 2 - 2 - 2 - 2 - 1 - 2 - 2 - 2 - 2 16 - 143 -
NE - 2 14 - - 2 32 - - 2 - 1 - 2 - 2 - 2 48 - - 2 1231 -
NV - 1 - 2 - 2 - 2 13 - - 2 - 2 17 - - 2 - 2 - 2 554 -
NH - 2 - 2 5 - - 2 - 2 100 - - 2 - 2 - 2 9 - - 1 1284 -
NJ - 2 - 2 165 - 647 - 35 - 28 - 225 - 74 - 17 - 108 - - 2 13705 -
NM - 1 - 1 - 2 33 - - 2 - 1 - 1 - 2 - 1 - 2 - 2 311 -
NY - 2 - 2 89 - 599 - 24 - 45 - 142 - 74 - 20 - 148 - - 2 10325 -
NC 1024 - - 2 541 - 299 - 32 - 15 - 479 - - 2 93 - 85 - - 2 11207 -
ND - 1 - 2 - 2 - 2 - 2 - 2 - 2 - 2 - 2 - 2 - 2 216 -
OH 216 - 20 - 524 - 864 - 43 - - 2 261 - 110 - 544 - 354 - 18 - 20427 -
OK - 2 - 2 45 - 292 - 8 - - 2 13 - 34 - - 2 149 - - 2 2775 -
OR - 2 - 2 7 - 35 - - 2 - 2 - 2 443 - - 2 - 2 - 2 1978 -
PA 95 - 6 - 240 - 422 - 30 - 19 - 244 - 51 - 125 - 164 - - 2 14060 -
RI - 2 - 2 - 2 20 - - 2 - 2 - 2 - 2 - 1 - 2 - 1 953 -
SC 1893 - - 2 119 - 248 - - 2 23 - 109 - - 2 - 2 85 - - 2 6012 -
SD 8 - 121 - - 2 21 - - 2 - 1 - 2 - 2 - 2 - 2 7 - 412 -
TN 208 - - 2 1669 - 540 - 31 - 3 - - 2 112 - 17 - 128 - - 2 9947 -
TX 375 - - 2 588 - 13169 - 120 - - 2 373 - 229 - 88 - 210 - - 2 29313 -
UT - 2 - 2 1 - 36 - 545 - - 2 5 - 22 - - 2 7 - 4 - 1014 -
VM - 2 - 1 - 2 9 - - 2 74 - - 2 - 2 - 2 - 2 - 1 377 -
VA 135 - 5 - - 2 254 - - 2 - 2 1524 - 26 - 52 - 95 - - 2 6007 -
WA - 2 - 2 - 2 36 - 49 - - 2 6 - 1749 - - 2 - 2 - 2 3126 -
WV - 2 - 1 23 - - 2 - 2 - 1 60 - 97 - 485 - 19 - - 1 2582 -
WI 46 - - 2 199 - 267 - 33 - - 2 129 - 86 - - 2 2444 - - 2 8769 -
WY - 2 - 2 - 2 - 2 - 2 - 2 - 2 - 1 - 2 - 2 - 2 - 2
D
T
j
6025 - 470 - 7608 - 24996 - 1825 - 703 - 5812 - 4507 - 1832 - 6713 - 352 - 278832
Note: 1. V: Value of Trade 2. S: Symbol. Both are used in CFS
Source: Bureau of Transportation Statistics and U.S. Census Bureau, 2000, Commodity Flow Survey
1997: CD-EC97-CFS, Washington, DC
198

Appendix B: Chapter III Appendix
Table A. III-1. Definition of USC sector
C lassification U SC D escription SC TG N A IC S A gg_U SC *
U SC 01 L ive anim als and live fish & M eat, fish, seafood, and their preparations (1+5)
U SC 02 C ereal grains & O ther agricultural products except for A nim al Feed (2+3)
U SC 03 A nim al feed and products of anim al origin, n.e.c. 4
U SC 04 M illed grain products and preparations, and bakery products 6
U SC 05 O ther prepared foodstuffs and fats and oils 7
U SC 06 A lcoholic beverages 8
U SC 07 Tobacco products 9
U SC 08 N onm etallic m inerals (M onum ental or building stone, N atural sands, G ravel and crushed stone, n.e.c. (10~ 1 3)
U SC 09 M etallic ores and concentrates 14
U SC 10 C oal and petroleum products (C oal and Fuel oils, n.e.c.) (15~ 1 9)
U SC 11 B asic chem icals 20
U SC 12 Pharm aceutical products 21
U SC 13 Fertilizers 22
U SC 14 C hem ical products and preparations, n.e.c. 23
U SC 15 Plastics and rubber 24
C om m odity U SC 16 L ogs and other w ood in the rough & W ood products (25+26)
Sectors U SC 17 Pulp, new sprint, paper, and paperboard & Paper or paperboard articles (27+28)
U SC 18 Printed products 29
U SC 19 Textiles, leather, and articles of textiles or leather 30
U SC 20 N onm etallic m ineral products 31
U SC 21 B ase m etal in prim ary or sem i-finished form s and in finished basic shapes 32
U SC 22 A rticles of base m etal 33
U SC 23 M achinery 34
U SC 24 Electronic and other electrical equipm ent and com ponents, and office equipm ent 35
U SC 25 M otorized and other vehicles (including parts) 36
U SC 26 Transportation equipm ent, n.e.c. 37
U SC 27 Precision instrum ents and apparatus 38
U SC 28 Furniture, m attresses and m attress supports, lam ps, lighting fittings, and illum inated signs 39
U SC 29 M iscellaneous m anufactured products, Scrap, M ixed freight, and C om m odity unknow n (40~ 9 9)
USC30Utility 22
USC31Construction 23
U SC 32 W holesale Trade 42
USC33Transportation 48
U SC 34 Postal and W arehousing 49
N on-C om m odity U SC 35 R etail Trade (44+45)
(Service) U SC 36 B roadcasting and inform ation services** (515~519)
Sectors U SC 37 Finance and Insurance 52
U SC 38 R eal estate and rental and leasing 53
U SC 39 Professional, Scientific, and Technical services 54
U SC 40 M anagem ent of com panies and enterprises 55
U SC 41 A dm inistrative support and w aste m anagem ent 56
U SC 42 Education Services 61
U SC 43 H ealth C are and Social A ssistances 62
U SC 44 A rts, E ntertainm ent, and R ecreation 71
U SC 45 A ccom m odation and Food services 72
U SC 46 Public adm inistration 92
U SC 47 O ther services except public adm inistration*** 81
A gg_U SC 08
A gg_U SC 09
A gg_U SC 01
A gg_U SC 02
A gg_U SC 03
A gg_U SC 04
A gg_U SC 05
A gg_U SC 06
A gg_U SC 07

0-8 Table A. III-1. Definition of USC sector
*Agg_USC sectors are defined basically at the basis of SCTG groups aggregated from the SCTG sectors
suggested by CFS.
**Publishing, Motion pictures, and Recording (IMPLAN 413-415, 417-419, or NAICS 511~512) are excluded in
this sector and included in Commodity Sectors.
***USC47 includes NAICS 81plus Support activities (18=Agriculture and forestry, 27-29=Mining) and Etc.
(243=Machine shops) in IMPLAN.
199

Table A. III-2. Total Domestic Import for each state by USC service sector, 2001
T otal D omestic Import ($M .)
States USC30 USC31 USC32 USC33 USC34 USC35 USC36 USC37 USC38 USC39 USC40 USC41 USC42 USC43 USC44 USC45 USC46 USC47
Alabama 543 264 3863 2800 591 1948 3933 7630 5352 6217 1979 2607 354 3000 1291 1221 77 2723
Alaska 199 392 1069 320 138 377 691 1693 1184 1010 302 713 176 663 107 147 2 516
Arizona 924 1302 2081 2001 568 1808 3886 7309 4684 4470 785 2168 1327 4132 929 960 40 1985
Arkansas 1196 943 2921 1612 444 1262 2156 4762 3818 4245 967 1621 238 1498 630 672 49 1661
California 9822 1535 0 18856 5037 10892 31050 45651 33537 26788 4088 12150 7064 32388 3729 9017 398 14561
Colorado 419 0 26 2307 656 1702 5830 7428 5318 4799 676 2530 1067 4366 537 1041 218 2307
Connecticut 533 17 917 2949 569 1859 4617 11327 3935 3390 558 1208 451 2982 762 2438 392 1957
Delaware 175 0 1221 546 184 366 1117 2450 1142 1655 328 700 240 474 201 279 35 1089
D istrict of Columbia 394 954 1380 955 326 1493 2674 1179 1937 5684 58 999 100 679 314 228 11 1051
Florida 3771 1642 3137 7518 2307 3771 13028 20815 15025 9211 2007 5334 3485 8950 2187 3367 164 6008
Georgia 1743 4171 222 3819 1077 1885 8338 11195 8258 8285 1604 3904 797 9305 1759 1940 422 3923
Hawaii 462 34 679 782 207 399 1078 2199 1477 1070 184 1201 273 989 177 280 4 472
Idaho 719 49 930 599 253 371 1013 2217 1456 1515 265 735 185 1042 219 234 26 706
Illinois 3135 1758 151 5016 1757 5398 13196 18628 15360 8003 1738 4395 1354 7749 1452 5202 218 7008
Indiana 1920 1386 6847 4006 808 1759 5671 11515 7286 9865 2086 3870 621 2607 931 1913 480 4060
Iowa 680 994 2545 2455 493 663 2911 6009 5038 4847 1585 1778 253 1227 420 752 72 2183
Kansas 340 690 1348 1765 598 766 3165 4800 4532 4825 873 1584 351 1192 634 1119 12 2216
Kentucky 1993 3653 4268 3051 1020 1579 3401 7574 7211 6090 1429 2753 379 2083 976 1095 103 5529
Louisiana 1821 65 4393 2376 693 1833 3536 7706 7825 5201 1475 1885 367 3191 708 919 115 2674
Maine 712 0 1402 632 217 365 1082 2587 2206 1779 474 974 109 575 236 285 40 813
Maryland 550 373 3671 3157 812 2283 6284 9867 5572 6321 1891 2304 755 2919 1354 1565 34 2239
M assachusetts 1572 81 736 4873 995 3556 7153 13078 8674 5466 882 3343 782 4002 1495 2881 327 4230
M ichigan 651 585 12032 7707 1549 2722 9062 19065 12850 10110 1901 3920 1131 4134 1180 3402 202 15148
Minnesota 1949 366 75 2943 837 1438 5447 8379 5247 4602 746 2164 547 3054 782 1546 304 2407
M ississippi 971 677 3271 1369 351 1055 2094 4510 4639 4144 969 1704 282 2553 492 479 74 1838
M issouri 2270 1170 1197 2213 670 2064 5584 8217 6374 4739 714 2744 578 2232 785 1679 120 3246
Montana 63 146 1095 492 223 574 653 1686 1943 1212 373 746 108 416 160 167 4 623
Nebraska 434 402 985 1148 392 457 1640 3528 3988 2661 397 860 232 1054 288 514 9 1346
Nevada 597 1099 1549 1078 318 665 1914 4390 2481 2541 482 1429 807 3015 404 503 28 1436
New Hampshire 358 19 1270 1021 149 267 1300 2760 1732 1785 360 749 130 656 188 332 6 817
New Jersey 2201 180 695 4058 1292 3641 10338 19217 9259 6448 1240 2942 2587 4879 1188 6148 173 5328
New M exico 391 223 1397 743 300 575 1205 3012 2350 3034 481 637 251 1536 355 306 5 919
New York 3998 10040 245 10923 4086 16928 19595 30360 22585 12713 2132 6908 2172 12325 4158 12501 986 13114
North Carolina 4235 839 5178 4415 1021 2185 7138 14470 11854 12818 3572 5571 709 5614 1378 2208 225 4320
North Dakota 122 46 811 391 191 240 614 1508 2033 1055 199 524 64 371 149 127 3 694
Ohio 5079 2338 32098 5793 1609 2398 10180 21463 13362 12546 2579 5225 1137 5403 1308 4412 129 7462
Oklahoma 331 216 2559 1655 850 1235 2917 5753 6091 3438 1272 1455 379 1587 887 938 42 2108
Oregon 709 277 5 1531 449 926 2893 5584 3450 2722 483 1381 371 1621 563 816 21 1854
Pennsylvania 2808 150 5997 5546 1526 3301 12600 22469 13757 9357 2744 5079 1216 6828 2181 4308 804 5375
Rhode Island 456 114 1383 911 259 999 938 1509 1264 1805 408 973 100 543 224 385 49 482
South Carolina 1285 91 4575 2238 404 926 3294 7376 4595 5730 1808 3513 316 3966 658 749 142 2565
South D akota 221 298 461 416 222 301 707 1556 1598 1329 264 668 73 478 125 154 9 686
Tennessee 2209 619 2301 2242 1072 1199 4844 8831 5994 7766 2270 2915 496 3079 771 1173 304 4263
Texas 2010 12419 656 8619 4265 4768 19921 30191 31899 15693 11788 8329 2472 21272 4247 7845 885 12304
Utah 559 273 530 869 228 438 1653 3596 2102 2742 430 983 207 873 307 399 62 892
Vermont 185 2 866 359 64 269 537 1225 965 838 249 418 54 322 140 125 10 371
Virginia 2279 2791 7688 3805 1047 3211 12620 13757 8867 16023 1210 6589 771 4114 1737 1909 155 3469
W ashington 1187 744 1368 3120 881 1475 6381 9666 6751 8799 1543 4575 1405 4968 788 1428 27 2944
West V irginia 325 522 1928 1249 443 793 1222 3511 3811 2739 867 1119 244 1159 534 329 24 1392
Wisconsin 2453 721 5183 3134 751 1869 5287 10804 8280 7100 1464 2609 599 2967 959 1537 47 3634
Wyoming 65 1938 608 343 137 326 422 1191 1022 752 415 328 188 1018 159 104 3 365

0-9 Table A-2. Total Domestic Import for each state by USC service sector, 2001
199
200

Table A. III-3. Total Intrastate Flows for each state by USC service sector, 2001
Total Intrastate Flows ($M .)
States USC30 USC31 USC32 USC33 USC34 USC35 USC36 USC37 USC38 USC39 USC40 USC41 USC42 USC43 USC44 USC45 USC46 USC47
Alabama 5291 13080 7892 4207 1662 11262 2623 8081 15920 8047 908 3233 1450 15512 940 5861 17399 7538
Alaska 1188 2568 703 934 269 1686 691 1034 2769 1572 246 754 148 2198 291 1095 5346 2012
Arizona 4392 21040 8772 4655 1999 14633 3826 11669 22157 12809 1833 5184 794 17131 1969 7374 18875 8370
Arkansas 2451 7492 4820 3123 1111 6427 1688 4119 9033 3425 1000 1795 779 9170 623 3448 8499 4749
California 34240 113187 92054 35305 14978 107249 33588 100130 178847 147085 17140 44211 13837 131932 19812 56412 148559 72959
Colorado 5825 23659 12234 4918 2224 15114 6074 14688 25392 18432 2213 6471 1287 17467 2492 7841 21009 10662
Connecticut 4125 10625 9649 2867 1788 12412 3065 11726 21036 13455 2330 4640 1830 17429 1828 5612 14637 7787
Delaware 823 2623 992 748 358 2418 382 1624 3738 1944 285 503 176 3323 299 1260 3103 1467
District of Columbia 608 1166 771 537 583 1352 1654 2518 5203 8100 255 2170 305 3329 318 1334 29316 757
Florida 15266 61392 31009 15934 6212 51189 13541 44063 77395 44613 6153 20015 3960 65757 7153 25576 55520 29119
Georgia 9589 29535 22220 10513 4259 25170 9070 22548 39317 26226 4321 10263 3140 27561 2998 13183 35944 16759
Hawaii 1117 3221 1734 1086 374 3366 821 2235 5334 3203 370 1338 278 4194 499 1953 8301 1988
Idaho 816 4925 2429 1357 400 3561 733 2323 5083 2426 466 977 319 4166 439 1791 5203 2095
Illinois 14489 42939 37724 17957 6637 38377 9807 42589 65136 45513 7333 16690 5376 53685 6378 20137 45637 26703
Indiana 6635 20413 14489 8022 2894 17621 3590 12933 26596 11300 2920 4706 2176 23806 2623 8799 19265 12716
Iowa 3708 8842 7472 3270 1445 8701 1788 7813 11594 5486 647 2449 1061 11601 1217 4524 10806 5600
Kansas 3601 8752 7338 3220 1285 7980 2208 6815 12008 6007 1143 2566 964 10939 884 3798 11520 6446
Kentucky 3732 12373 8387 4062 1339 10741 2579 7330 14376 7069 1528 3105 1401 15228 1276 5753 16044 6772
Louisiana 4286 13659 6917 6013 1665 11463 3002 7755 16453 13083 2025 4056 1475 15560 1666 6216 17459 12828
Maine 1016 4096 1878 1286 464 3906 814 2419 4501 2100 373 741 441 5111 456 1916 4613 2152
Maryland 5900 20824 9462 4799 2890 17711 4799 13039 27296 23856 964 7348 2218 24116 2041 9051 37630 10868
M assachusetts 6993 19796 18493 6296 3397 21621 6692 20775 36955 28231 3771 8905 3210 32493 3015 11443 24400 12984
M ichigan 12283 29594 20271 8845 4376 29314 6677 20565 44338 27726 5528 10332 3933 40618 4651 14705 33780 21930
M innesota 5469 18177 16708 6299 2665 16365 3843 16791 27141 17481 3162 6782 2143 21575 2413 8752 18972 11506
Mississippi 2675 7058 3751 2794 1007 6610 1524 4181 7934 4027 753 1422 745 8367 833 3663 10729 4583
Missouri 5496 19326 15135 7624 2787 16445 5030 15330 26285 17145 3036 5702 2086 23531 2402 8532 20255 12504
Montana 1150 3214 1075 954 258 2238 641 1556 3005 1908 115 587 259 3334 330 1288 4050 1720
Nebraska 2222 6410 4748 2741 835 5379 1433 5211 7017 4405 850 1718 596 6920 751 2662 6653 3667
Nevada 1954 10857 2759 1801 795 6828 1437 4425 9648 4946 674 2105 138 6390 875 3278 7985 3167
New Hampshire 1211 4233 2430 840 610 4404 829 2881 5930 3209 605 1195 530 5607 583 2182 3650 2279
New Jersey 8762 24158 22786 10838 4439 28598 8807 22208 47825 35013 5222 11364 2677 40850 4522 11914 33732 17919
New M exico 1574 5728 2382 1454 473 4614 1210 2700 6349 4334 454 1902 419 5587 554 2330 10245 3727
New York 22222 49269 47720 18073 8402 52731 21014 68582 109290 67412 8957 27700 8823 91354 9882 28108 75004 34500
North Carolina 6824 32743 18902 9015 3599 24515 6621 16729 33565 15607 3786 5749 2902 29993 3019 12064 35760 14789
North Dakota 888 1911 1117 807 268 1790 451 1332 2019 1122 259 487 238 2433 221 975 3146 1467
Ohio 10498 34529 3168 13950 5303 34605 8141 26667 51525 30141 6067 12478 4200 45892 5312 16151 39581 25072
Oklahoma 4161 9167 6100 3952 1427 9190 2953 6752 13735 7624 1129 3174 1016 13085 983 4596 15132 9987
Oregon 3636 10779 9096 3869 1441 10307 2426 7838 15942 8455 1585 3613 1178 13607 1355 5271 12790 6492
Pennsylvania 14322 38568 30104 15347 6103 39075 9164 34601 60965 40652 6058 14513 4967 53170 5106 19319 37744 25134
Rhode Island 702 2004 1277 551 281 2409 667 2947 4613 1585 203 467 405 4467 371 1521 4196 1773
South Carolina 4011 14614 6090 3830 1585 11451 2234 6767 15074 6059 970 2020 1296 12497 1400 5648 17052 6738
South D akota 789 2680 1838 824 248 2205 519 1619 2937 1058 227 441 265 2942 304 1168 3174 1399
Tennessee 5453 17181 15164 8182 2553 17322 4439 14269 25586 11770 2071 5636 2032 22351 2479 8832 17765 11271
Texas 29573 85768 58470 29786 9539 65311 19810 62619 105661 78198 4291 29883 7773 74207 7870 31345 83390 61441
Utah 1785 7747 4903 2476 952 6000 1606 4859 8895 6087 1021 2107 658 7390 845 3013 9861 3873
Vermont 568 2171 794 567 285 1840 416 1146 2448 1429 183 473 220 2500 229 965 2339 1045
Virginia 7769 29985 12589 8673 3269 22471 10021 16274 35291 38778 3746 9688 2989 30539 2702 12256 47789 15012
Washington 6892 21850 15628 7044 2556 19476 4382 15620 28674 17463 2106 5829 1726 23741 2807 10011 27977 11909
West Virginia 2057 4926 2414 1723 410 4448 1154 2232 5236 2647 326 967 473 6304 504 2399 7039 3148
W isconsin 5308 17478 13807 6931 2645 15987 3159 13743 22218 12085 3128 5225 1950 21476 2134 8458 19368 10910
Wyoming 710 2473 729 687 148 1484 336 816 2399 1167 74 361 45 1341 228 796 2728 2091

0-10 Table A-3. Total Intrastate Flows for each state by USC service sector, 2001
Note: Values are revised by adding foreign imports to the calculated domestic products supported from the 2001 IMPLAN data.
200
201

Appendix C: Chapter IV Appendix
Table A. IV-1. Comparison of Value Added Between IMPLAN and BEA Data Sets
USC Total Value Added of U.S. (TVA_US) Total Value Added of Sum of States (TVA_SS)
Sector IMPLAN_US BEA_US Diff_US IMPLAN_SS BEA_SS Diff_SS
USC 1 32,149 39,482 -22.8% 32,149 42,815 -33.2%
USC 2 53,336 53,831 -0.9% 53,336 53,559 -0.4%
USC 3 6,573 8,622 -31.2% 6,573 8,914 -35.6%
USC 4 31,985 46,471 -45.3% 31,985 47,955 -49.9%
USC 5 69,117 84,445 -22.2% 69,117 80,324 -16.2%
USC 6 22,754 13,358 41.3% 22,754 13,312 41.5%
USC 7 30,137 9,404 68.8% 30,137 5,324 82.3%
USC 8 11,045 12,054 -9.1% 11,045 11,498 -4.1%
USC 9 4,135 4,499 -8.8% 4,135 4,763 -15.2%
USC 10 89,620 117,465 -31.1% 89,620 117,201 -30.8%
USC 11 25,143 25,180 -0.1% 25,143 24,776 1.5%
USC 12 61,902 62,462 -0.9% 61,902 65,433 -5.7%
USC 13 5,985 5,989 -0.1% 5,985 6,028 -0.7%
USC 14 53,351 53,914 -1.1% 53,351 52,013 2.5%
USC 15 69,988 73,021 -4.3% 69,988 72,812 -4.0%
USC 16 35,716 39,441 -10.4% 35,716 38,064 -6.6%
USC 17 45,101 44,412 1.5% 45,101 44,355 1.7%
USC 18 103,614 96,840 6.5% 103,614 96,764 6.6%
USC 19 57,847 43,529 24.8% 57,847 46,887 18.9%
USC 20 44,457 46,798 -5.3% 44,457 47,025 -5.8%
USC 21 32,230 33,906 -5.2% 32,230 34,080 -5.7%
USC 22 81,546 82,094 -0.7% 81,546 81,729 -0.2%
USC 23 131,457 135,809 -3.3% 131,457 136,426 -3.8%
USC 24 294,755 240,998 18.2% 294,755 242,349 17.8%
USC 25 104,416 111,924 -7.2% 104,416 111,733 -7.0%
USC 26 35,470 41,161 -16.0% 35,470 40,676 -14.7%
USC 27 53,785 45,854 14.7% 53,785 45,268 15.8%
USC 28 30,619 30,466 0.5% 30,619 30,473 0.5%
USC 29 48,510 50,953 -5.0% 48,510 50,497 -4.1%
USC 30 165,574 202,286 -22.2% 165,574 202,287 -22.2%
USC 31 425,995 469,535 -10.2% 425,995 469,533 -10.2%
USC 32 587,645 607,078 -3.3% 587,645 607,076 -3.3%
USC 33 225,713 219,602 2.7% 225,713 220,303 2.4%
USC 34 114,420 77,346 32.4% 114,420 76,644 33.0%
USC 35 579,625 691,578 -19.3% 579,625 691,578 -19.3%
USC 36 339,334 339,077 0.1% 339,334 338,699 0.2%
USC 37 774,908 782,627 -1.0% 774,908 782,625 -1.0%
USC 38 1,269,276 1,276,570 -0.6% 1,269,276 1,276,572 -0.6%
USC 39 784,151 698,825 10.9% 784,151 698,821 10.9%
USC 40 153,223 177,636 -15.9% 153,223 177,635 -15.9%
USC 41 315,921 289,419 8.4% 315,921 289,416 8.4%
USC 42 55,924 59,775 -6.9% 55,924 60,234 -7.7%
USC 43 653,680 679,552 -4.0% 653,680 679,093 -3.9%
USC 44 98,774 95,664 3.1% 98,774 95,661 3.2%
USC 45 259,094 265,805 -2.6% 259,094 265,804 -2.6%
USC 46 1,211,813 1,188,505 1.9% 1,211,813 1,188,506 1.9%
USC 47 395,364 282,894 28.4% 395,364 284,607 28.0%
TOTAL 10,077,179 10,058,158 0.2% 10,077,179 10,058,145 0.2%
0-11 Table A-1. Comparison of value added between IMPLAN and BEA data sets
Note: Diff_US=(IMPLAN_US-BEA_US)/IMPLAN_US and
Diff_SS=(IMPLAN_SS-BEA_SS)/IMPLAN_SS.
Source: BEA GSP data from http://www.bea.doc.gov/bea/regional/gsp and 2001 IMPLAN data.
202

Table A. IV-2. Comparison of Total Output Between IMPLAN and BEA Data Sets
USC Total Output of U.S. (TO_US) Total Output of Sum of States (TO_SS)
Sector IMPLAN_US BEA_US Diff_US IMPLAN_SS Diff_SS
USC 1
173,097 195,922 -13.2% 173,097 0.0%
USC 2 118,853 117,442 1.2% 118,853 0.0%
USC 3 44,785 50,631 -13.1% 44,785 0.0%
USC 4 84,932 105,455 -24.2% 84,932 0.0%
USC 5 286,070 309,327 -8.1% 286,070 0.0%
USC 6 61,546 42,350 31.2% 61,546 0.0%
USC 7 52,637 17,100 67.5% 52,637 0.0%
USC 8 19,049 18,364 3.6% 19,049 0.0%
USC 9 9,129 8,813 3.5% 9,129 0.0%
USC 10 371,603 376,181 -1.2% 371,603 0.0%
USC 11 76,034 75,787 0.3% 76,034 0.0%
USC 12 134,457 134,295 0.1% 134,457 0.0%
USC 13 16,209 16,157 0.3% 16,209 0.0%
USC 14 142,133 142,483 -0.2% 142,133 0.0%
USC 15 203,666 200,586 1.5% 203,666 0.0%
USC 16 101,676 103,276 -1.6% 101,676 0.0%
USC 17 142,353 141,106 0.9% 142,353 0.0%
USC 18 203,883 213,869 -4.9% 203,883 0.0%
USC 19 172,998 130,168 24.8% 172,998 0.0%
USC 20 97,801 97,201 0.6% 97,801 0.0%
USC 21 121,498 118,894 2.1% 121,498 0.0%
USC 22 184,519 182,176 1.3% 184,519 0.0%
USC 23 331,350 322,670 2.6% 331,350 0.0%
USC 24 601,195 608,161 -1.2% 601,195 0.0%
USC 25 447,184 441,871 1.2% 447,184 0.0%
USC 26 118,010 118,170 -0.1% 118,010 0.0%
USC 27 114,130 112,783 1.2% 114,130 0.0%
USC 28 73,637 73,417 0.3% 73,637 0.0%
USC 29 110,923 105,886 4.5% 110,923 0.0%
USC 30 296,699 343,430 -15.8% 296,699 0.0%
USC 31 1,013,113 899,778 11.2% 1,013,113 0.0%
USC 32 875,258 851,286 2.7% 875,258 0.0%
USC 33 502,771 469,464 6.6% 502,771 0.0%
USC 34 162,269 101,953 37.2% 162,269 0.0%
USC 35 942,803 1,021,032 -8.3% 942,803 0.0%
USC 36 586,269 710,579 -21.2% 586,269 0.0%
USC 37 1,287,273 1,361,671 -5.8% 1,287,273 0.0%
USC 38 1,681,503 1,775,397 -5.6% 1,681,503 0.0%
USC 39 1,008,256 1,105,607 -9.7% 1,008,256 0.0%
USC 40 210,209 290,414 -38.2% 210,209 0.0%
USC 41 443,881 481,024 -8.4% 443,881 0.0%
USC 42 85,680 107,147 -25.1% 85,680 0.0%
USC 43 1,188,873 1,094,726 7.9% 1,188,873 0.0%
USC 44 154,279 154,140 0.1% 154,279 0.0%
USC 45 498,852 500,925 -0.4% 498,852 0.0%
USC 46 1,288,980 2,019,174 -56.6% 1,288,980 0.0%
USC 47 755,883 534,941 29.2% 755,883 0.0%
TOTAL 17,598,207 18,403,229 -4.6% 17,598,207 0.0%
0-12 Table A-2. Comparison of total output between IMPLAN and BEA data sets
Note: Diff_US=(IMPLAN_US-BEA_US)/IMPLAN_US and
Diff_SS=(IMPLAN_SS- IMPLAN _US)/IMPLAN_SS.
Source: BEA GSP data from http://www.bea.doc.gov/bea/regional/gsp and 2001 IMPLAN data.
203

Table A. IV-3. Definitions of USC Two-Digit Sectors
Classification USC Description SCTG NAICS
USC01 Live animals and live fish & Meat fish seafood and their preparations (1+5)
USC02 Cereal grains & Other agricultural products except for Animal Feed (2+3)
USC03 Animal feed and products of animal origin, n.e.c. 4
USC04 Milled grain products and preparations, and bakery products 6
USC05 Other prepared foodstuffs and fats and oils 7
USC06 Alcoholic beverages 8
USC07 Tobacco products 9
USC08
Nonmetallic minerals (Monumental or building stone, Natural sands, Gravel and crushed stone,
n.e.c.) (10~13)
USC09 Metallic ores and concentrates 14
USC10 Coal and petroleum products (Coal and Fuel oils, n.e.c.) (15~19)
USC11 Basic chemicals 20
USC12 Pharmaceutical products 21
USC13 Fertilizers 22
USC14 Chemical products and preparations, n.e.c. 23
USC15 Plastics and rubber 24
USC16 Logs and other wood in the rough & Wood products (25+26)
USC17 Pulp, newsprint, paper, and paperboard & Paper or paperboard articles (27+28)
USC18 Printed products 29
USC19 Textiles, leather, and articles of textiles or leather 30
USC20 Nonmetallic mineral products 31
USC21 Base metal in primary or semi-finished forms and in finished basic shapes 32
USC22 Articles of base metal 33
USC23 Machinery 34
USC24 Electronic and other electrical equipment and components, and office equipment 35
USC25 Motorized and other vehicles (including parts) 36
USC26 Transportation equipment, n.e.c. 37
USC27 Precision instruments and apparatus 38
USC28 Furniture, mattresses and mattress supports, lamps, lighting fittings, and illuminated signs 39
Commodity
Sectors
USC29 Miscellaneous manufactured products, Scrap, Mixed freight, and Commodity unknown (40~99)
USC30 Utility 22
USC31 Construction 23
USC32 Wholesale Trade 42
USC33 Transportation 48
USC34 Postal and Warehousing 49
USC35 Retail Trade (44+45)
USC36 Broadcasting and information services* (515~519)
USC37 Finance and Insurance 52
USC38 Real estate and rental and leasing 53
USC39 Professional, Scientific, and Technical services 54
USC40 Management of companies and enterprises 55
USC41 Administrative support and waste management 56
USC42 Education Services 61
USC43 Health Care and Social Assistances 62
USC44 Arts, Entertainment, and Recreation 71
USC45 Accommodation and Food services 72
USC46 Public administration 92
Non-
Commodity
(Service)
Sectors
USC47 Other services except public administration** 81
0-13 Table A-3. Definitions of USC Two-Digit sectors
Source: Park et al. (2007)
204

Appendix D: Chapter V Appendix

Figure A. V-1 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
FlexNIEMO, Direct Losses of Foreign Exports, USC Sector 10 (Coal and Petroleum Products), August
2005
Legend
AUG._05: As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO, August 2005
.
800,000 0 800,000 400,000 Meters

0-5Figure A. VI-1 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), August 2005
205

Figure A. V-2 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
FlexNIEMO, Direct Losses of Foreign Exports, USC Sector 10 (Coal and Petroleum Products), September
2005
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO, September 2005
.
800,000 0 800,000 400,000 Meters
Legend
SEP ._05: As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-6 Figure A-2 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), September 2005
206

Figure A. V-3 The State-by-State Economic Impacts of the Customs District of Louisiana:
Application of FlexNIEMO, Direct Losses of Foreign Exports, USC Sector 10 (Coal and
Petroleum Products), October 2005
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO, October 2005
.
800,000 0 800,000 400,000 Meters
Legend
OCT._05: As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-7 Figure A-3 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), October 2005
207

Figure A. V-4 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
FlexNIEMO, Direct Losses of Foreign Exports, USC Sector 10 (Coal and Petroleum Products), November
2005
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO, November 2005
.
800,000 0 800,000 400,000 Meters
Legend
NOV._05: As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-8 Figure A-4 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), November 2005
208

Figure A. V-5 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
FlexNIEMO, Direct Losses of Foreign Exports, USC Sector 10 (Coal and Petroleum Products), December
2005
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO, December 2005
.
800,000 0 800,000 400,000 Meters
Legend
DEC._05: As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-9 Figure A. IV-5 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), December 2005
209

Figure A. V-6 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
FlexNIEMO, Direct Losses of Foreign Exports, USC sSctor 10 (Coal and Petroleum Products), January
2006
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO, January 2006
.
800,000 0 800,000 400,000 Meters
Legend
JAN._06: As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-10 Figure A. IV-6 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), January 2006
210

Figure A. V-7 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
FlexNIEMO, Direct Losses of Foreign Exports, USC Sector 10 (Coal and petroleum products), February
2006
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO, February 2006
.
800,000 0 800,000 400,000 Meters
Legend
FEB._06: As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-11 Figure A-7 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), January 2006

211

Figure A. V-8 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
FlexNIEMO, Direct Losses of Foreign Exports, USC Sector 10 (Coal and Petroleum Products), March
2006
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO, March 2006
.
800,000 0 800,000 400,000 Meters
Legend
MAR._06: As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-12 Figure A. IV-8 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), March 2006
212

Figure A. V-9 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
FlexNIEMO, Direct Losses of Foreign Exports, USC Sector 10 (Coal and Petroleum Products), Total
Impacts Between Aug. 05 and Mar. 06.
The state-by-state economic impacts
of Customs District of Louisiana:
Application of FlexNIEMO
.
800,000 0 800,000 400,000 Meters
Legend
FLEXNIEMO_SUM:
As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-13 Figure A. IV-9 The state-by-state economic impacts of Customs District of Louisiana:
Application of FlexNIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), Total Impacts between Aug. 05 and Mar. 06.
213

Figure A. V-10 The State-by-State Economic Impacts of the Customs District of Louisiana: Application of
NIEMO, Direct Losses of Foreign Exports in USC, Sector 10 (Coal and Petroleum Products), Total
Impacts Between Aug. 05 and Mar. 06.
The state-by-state economic impacts
of Customs District of Louisiana:
Application of NIEMO
.
800,000 0 800,000 400,000 Meters
Legend
NIEMO_SUM:
As Percent of Total
<0.001%
<0.01%
<0.1%
<1%
>=1%

0-14 Figure A. IV-10 The state-by-state economic impacts of Customs District of Louisiana:
Application of NIEMO, Direct losses of foreign exports in USC sector 10 (Coal and
petroleum products), Total Impacts between Aug. 05 and Mar. 06.

Abstract (if available)

Abstract Therefore, this dissertation addresses methodologies on spatial and/or temporal economic IO models extended from the classic IO model, as well as from a temporally extended national IO model. Also, this dissertation includes an essay that offers theoretical support to the supply-driven IO model, especially in light of some well known interpretation controversies. Hence, this dissertation includes three essays on extensions of supply-driven IO and spatial expansions of the classic demand-driven IO: 'The Supply-Driven Input-Output Model: A Reinterpretation and Extension', 'A Two-Step Approach Estimating State-by-State Commodity Trade Flows', 'Estimation of Interstate Trade Flows for Service Industries'. A fourth essay, 'An Evaluation of Input-Output Aggregation Error Using a New MRIO Model', provides information on the accuracy of the National Interstate Economic Model (NIEMO) an operational multi-regional input-output model developed as part of this research . The final essay on 'Constructing a Flexible National Interstate Economic Model (FlexNIEMO)' deals with the temporal extensions of all types of IO models. ❧ Ever since the 2001 terrorist attacks on the U.S., several studies have evaluated the socioeconomic impacts on the U.S. economy. Economic impacts are not restricted to just the immediate impact area

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Empirical essays on trade liberalization and export diversification

PDF

The spatial economic impact of live music in Orange County, CA

PDF

Essays on the econometric analysis of cross-sectional dependence

PDF

From structure to agency: Essays on the spatial analysis of residential segregation

PDF

Modeling and analysis of parallel and spatially-evolving wall-bounded shear flows

PDF

Essays on beliefs, networks and spatial modeling

PDF

Network-based simulation of air pollution emissions associated with truck operations

PDF

Modeling and predicting with spatial‐temporal social networks

PDF

Essays on business cycle volatility and global trade

PDF

Spatial and temporal expenditure-pricing equity of rail transit fare policies

PDF

Inter-temporal allocation of human capital and economic performance

PDF

Essays on interest rate determination in open economies

PDF

Two essays on major macroeconomic shocks in the Japanese economy: demographic shocks and financial shocks

PDF

Essays on narrative economics, Climate macrofinance and migration

PDF

The economic and political impacts of U.S. federal carbon emissions trading policy across households, sectors and states

PDF

Novel techniques for analysis and control of traffic flow in urban traffic networks

PDF

Essays on congestion, agglomeration, and urban spatial structure

PDF

Essays on development economics and adolescent behavior

PDF

Reconstruction and estimation of shear flows using physics-based and data-driven models

PDF

A million-plus neuron model of the hippocampal dentate gyrus: role of topography, inhibitory interneurons, and excitatory associational circuitry in determining spatio-temporal dynamics of granul...

Asset Metadata

Creator Park, Jiyoung (author)

Core Title Essays on economic modeling: spatial-temporal extensions and verification

School School of Policy, Planning, and Development

Degree Doctor of Philosophy

Degree Program Planning

Publication Date 08/07/2007

Defense Date 03/06/2007

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag computable general equilibrium,economic modeling,multiregional input-output,OAI-PMH Harvest,spatial econometrics,trade flows

Place Name USA (countries)

Language English

Advisor Gordon, Peter (committee chair), Moore, James Elliott, II (committee member), Richardson, Harry W. (committee member)

Creator Email jiyoungp@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m773

Unique identifier UC161468

Identifier etd-Park-20070807 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-542574 (legacy record id),usctheses-m773 (legacy record id)

Legacy Identifier etd-Park-20070807.pdf

Dmrecord 542574

Document Type Dissertation

Rights Park, Jiyoung

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu