Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Two essays on financial crisis and banking
(USC Thesis Other)
Two essays on financial crisis and banking
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
TWO ESSAYS ON FINANCIAL CRISIS AND BANKING
by
Mihye Lee
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulllment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(ECONOMICS)
August 2012
Copyright 2012 Mihye Lee
to my parents
ii
Acknowledgements
First of all, I am deeply thankful for the support, patience and guidance I have
received from my advisor Robert Dekle. Without his helpful advisement, contin-
uous encouragement and stimulating discussion this work would never have been
completed. I have greatly beneted from his extensive experience in empirical in-
ternational nance.
I am also grateful for the support and guidance I received from a number of pro-
fessors, including Vincenzo Quadrini, Je Nugent and Daniel Carvalho, for their
generosity with their time and valuable comments.
I am also deeply indebted to Professor Tack Yun for his enormous support and
guidance, which helped me to pursue my Ph.D.
During my six years of study at USC, I received an abundance of help and support
from department sta members Young Miller and Morgan Ponder at the Depart-
ment of Economics. My deepest thanks to my friends Kyoung-Eun Kim, Jaejin An
and Ahram Moon, for being sincere and entertaining friends at all times. Special
thanks to Mary Kwon and Emily Page for their helpful comments and suggestions.
This thesis is dedicated to my parents who have always supported and encouraged
me during my graduate studies.
iii
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables vi
Abstract ix
Chapter 1 Introduction 1
Chapter 2 How did Financial Development Aect the Performance of Euro-
pean Firms Before and After the 2008 Credit Crisis? 4
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Description of Data . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Empirical Specications . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.1 Multilevel Regression . . . . . . . . . . . . . . . . . . . . . . 18
2.4.2 Panel Estimation . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.3 Summary of Findings . . . . . . . . . . . . . . . . . . . . . . 39
2.4.4 Extension : Eect of Strength of Legal Rights . . . . . . . . 40
2.4.5 Extension : Financial Development and Growth of Firm . . 43
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Chapter 3 Do Foreign Bank Aliates Cut Their Lending More than the Do-
mestic Banks In a Financial Crisis: A Quasi-Natural Experiment
with Global Individual Bank Data 48
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2 Bank Balance Sheets and Foreign Lending . . . . . . . . . . . . . . 52
3.3 Empirical Identication Methodology . . . . . . . . . . . . . . . . 54
3.4 Description of Data . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.5 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.5.1 Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.5.2 Additional Robustness Checks . . . . . . . . . . . . . . . . . 66
3.5.3 Why did European-bank aliates cut their lending by more? 74
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
iv
References 77
Appendix Chapter A Robustness Check using \Placebo" Specications 82
Appendix Chapter B Variable Denitions 85
v
List of Tables
Table 2.1 Industry Classication . . . . . . . . . . . . . . . . . . . . . . 12
Table 2.2 Description of Firm Size . . . . . . . . . . . . . . . . . . . . . 13
Table 2.3 Firm Size and Dependence on External Financing . . . . . . . 14
Table 2.4 Number of Firms per Age-Cohort . . . . . . . . . . . . . . . . 14
Table 2.5 Description of EBIT Before and After the Crisis . . . . . . . 16
Table 2.6 Results based on Firm Earnings . . . . . . . . . . . . . . . . 21
Table 2.7 Results based on Firm Earnings including the Crisis Dummy 26
Table 2.8 Results based on Firm Earnings . . . . . . . . . . . . . . . . 28
Table 2.9 Results based on Firm Earnings using
p
Age . . . . . . . . . 29
Table 2.10 Results based on Firm Earnings . . . . . . . . . . . . . . . . 31
Table 2.11 Results based on Firm Earnings using Industry DEF . . . . . 32
Table 2.12 Results based on De
ated Earnings . . . . . . . . . . . . . . . 33
Table 2.13 Results based on Firm Earnings . . . . . . . . . . . . . . . . 36
Table 2.14 Result based on Panel and Panel IV Estimation . . . . . . . . 39
Table 2.15 Results based on Firm Earnings . . . . . . . . . . . . . . . . 42
Table 2.16 Results based on Firm Earnings . . . . . . . . . . . . . . . . 44
Table 2.17 Results based on Sales Growth . . . . . . . . . . . . . . . . . 45
Table 3.1 Description of Sample . . . . . . . . . . . . . . . . . . . . . . 58
Table 3.2 Lending by U.S. and European Banks . . . . . . . . . . . . . 59
Table 3.3 Eects of Ownership on Bank Lending . . . . . . . . . . . . . 61
Table 3.4 Eects of Ownership on Bank Lending with Europe Owned
Dummy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
Table 3.5 Eects of Ownership on Bank Lending including Bank-Level
Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Table 3.6 Eects of Ownership on Bank Lending based on 06-09 sample 65
vi
Table 3.7 Eects of Ownership on Bank Lending Interacted with Capital
Market Openness . . . . . . . . . . . . . . . . . . . . . . . . . 67
Table 3.8 Eects of Ownership on Bank Lending Interacted with Capital
Market Openness . . . . . . . . . . . . . . . . . . . . . . . . . 68
Table 3.9 Eects of Ownership on Bank Lending based on lnLoan
t
. 69
Table 3.10 Eects of Ownership on Bank Lending based on lnLoan
t
Interacted with Capital Market Openness . . . . . . . . . . . 70
Table 3.11 Eects of Ownership on Bank Lending During the Recession (1) 71
Table 3.12 Eects of Ownership on Bank Lending During the Recession (2) 72
Table 3.13 Eects on Lending of Debt-GDP Ratio of Countries of the
Foreign Bank Aliates (1) . . . . . . . . . . . . . . . . . . . 73
Table 3.14 Eects on Lending of Debt-GDP Ratio of Countries of the
Foreign Bank Aliates (2) . . . . . . . . . . . . . . . . . . . 75
Table A.1 Random Assignment . . . . . . . . . . . . . . . . . . . . . . . 83
Table A.2 Comparison of \Correct" and \Random" Assigned of Country
Financial Development Indices . . . . . . . . . . . . . . . . . 84
Table B.1 Summary of Capital Market Openness . . . . . . . . . . . . . 85
vii
Abstract
This dissertation asks two empirical questions : How did nancial development
aect the performance of European rms before and after the 2008 Credit Crisis,
and do foreign bank aliates cut their lending more than the domestic banks in a
nancial crisis.
Chapter 2 studies the impact of the recent credit crisis on rm performance.
The recent credit crisis led to a deep recession in the U.S. and the rest of the
world. This chapter seeks to identify the signicant relationship between nancial
development and rm-level performance in advanced European economies based
on the recent credit crisis. To evaluate rm-level performance, it includes cross-
country dierences in nancial development and cross-rm dierences in dependence
on external nancing, and study how nancial development interacts with rm
dependence on external nancing. The results show that nancial development is
positively related to a rm's earnings in tranquil times. Surprisingly, however, that
same nancial development can also exacerbate the impact of a crisis. The results
are robust to estimation using various instruments for the endogenous variables,
and are statistically signicant across dierent specications.
Chapter 3 studies the impact of the recent credit crisis on bank lending. It con-
tributes to the literature on the international transmission of balance sheet shocks
that pummeled the banks of the industrialized countries in 2008 and 2009. It ex-
amines over time bank level data on 250,000 banks located around the world. Our
identication strategy relies on the dierential responses of foreign and domestic
viii
banks to the post-Lehman 2008 crisis. If in a particular market, say in Korea, a
foreign-aliated bank's (Citibank, Korea's) lending falls by more than a domestic
bank's (Kookmin's) lending, then we attribute this additional decline to the tight-
ening of the foreign aliates internal capital market at its headquarters. We control
for the decline in market conditions common to all banks in a particular region by
the decline in lending by the banks other than the foreign aliated bank. We nd
evidence that internal capital markets do indeed aect cross-border lending. In
particular, European bank aliates in Latin America and Asia cut their lending by
more than the domestic banks located in these regions.
ix
Chapter 1 Introduction
This dissertation consists of two essays on the issues surrounding the recent credit
crisis that erupted in late 2007. The recent credit crisis, also known as the nancial
crisis or the global nancial crisis, caused the failure of many nancial institutions
and led many countries into recession. It also triggered the European sovereign-debt
crisis.
The rst essay, \How did Financial Development Aect the Performance of
European Firms Before and After the 2008 Credit Crisis?", is an empirical study
which discusses how rms respond to the credit crisis. The study is based on
European rm-level data. Since a typical credit crisis is accompanied by a decline
in lending, this essay evaluates the performance of rms based on their dependence
on external nancing and their accessibility to nancial markets.
This essay tries to answer the following basic questions : (1) Does nancial
development promote higher rm-level earnings? (2) Is nancial development more
benecial to rms that are dependent on external nancing? (3) Does nancial
development help rms buer a shock (that is, does a rm in a country with higher
nancial development experience a decline in earnings when there is a crisis)?
The main ndings are as follows. First, nancial development promotes higher
rm-level earnings during calm times. Second, nancial development is more ben-
ecial to rms that need to access external nancing. Third, however, nancial
development contributes to a decline in rm-level earnings during a crisis, which
1
means that rms in a country with higher nancial development undergo more se-
vere drops in their earnings when there is a crisis. The second essay, \Do Foreign
Bank Aliates Cut Their Lending More than the Domestic Banks In a Financial
Crisis: A Quasi-Natural Experiment with Global Banking Data", uses bank-level
data to study changes in lending based on the recent credit crisis.
Drawing from bank balance sheets and ultimate owner information, this essay
examines whether bank ownership (for example, foreign aliates) matters in de-
termining bank lending during a crisis. To study the impact of ownership on bank
lending, this essay applies a panel random eects model including and excluding
country xed-eects.
The main ndings are as follows. First, the recent crisis contributes to a decline
in lending in the U.S. and Canada. Second, it has an unusual eect on bank
lending compared to the early 2000 recession, as the early 2000 recession does not
have a signicant eect on bank lending. Third, foreign banks in Latin America,
particularly European ones, cut their lending by more than the domestic banks
located in that region during the crisis.
The dissertation is organized as follows. Chapter 2, \How did Financial De-
velopment Aect the Performance of European Firms Before and After the 2008
Credit Crisis?", has ve sections. Section 2.1 gives an introduction. Section 2.2 is
a literature review of the growth, volatility and nancial development. Section 2.3
describes the data used in the analysis. Section 2.4 provides the empirical method-
ology and shows the results. Section 2.5 provides the conclusion. Chapter 2, \Do
2
Foreign Bank Aliates Cut Their Lending More than the Domestic Banks In a
Financial Crisis: A Quasi-Natural Experiment with Global Banking Data", has six
sections. Section 3.1 gives an introduction. Section 3.2 gives an overview how a
bank responds or reallocate its capital in response to a sudden shock. Section 3.3
shows the empirical identication methodology that is used in the empirical analysis.
Section 3.4 describes the data used in the analysis and summarizes it. Section 3.5
presents empirical results based on dierent ownership and other variables. Section
3.6 provides the conclusion.
3
Chapter 2 How did Financial
Development Aect the Performance of
European Firms Before and After the
2008 Credit Crisis?
2.1 Introduction
At the end of 2007, many rms and institutions collapsed because of market-
wide defaults in mortgage-backed securities. Firms with investments in these sub-
prime mortgages suered heavy losses, and many were liquidated, including well-
established rms like Lehman Brothers. Other nancial institutions experienced a
severe drop in asset prices, which eroded their net worth and resulted in a serious
short fall in their lending, as well as a deep recession in the U.S. and the rest of the
world. We have seen subsequent attempts to rejuvenate economies, yet most coun-
tries continue to struggle with the consequences of the subprime crisis. According
to a recent report from Credit Suisse, European banks are still (in late September,
2011) holding $532 billion of subprime debt.
The goal of this paper is to unveil factors that have aected rm-level perfor-
mance and to explain European cross-country dierences in response to the recent
4
credit crisis. We focus on European rms, given the availability of data, and the on-
going economic diculties in that region. Since a typical credit crisis is accompanied
by a decline in the supply of credit, rms that are nancially constrained or have
a shortage of liquidity are unable to withstand diminished funding opportunities
in the nancial markets. Those credit constraints are expected to be exacerbated
during times of severe distress, such as the crisis of 2008.
A rm's borrowing or liquidity conditions are closely related to its access to
nancial markets; thus we use a country's nancial development, as dened by
Rajan and Zingales (1998), as a measure of its access to external funding. Financial
development can be dened as the factors, policies, and institutions that lead to
eective nancial intermediation and markets, and deep and broad access to capital
and nancial services. Hence it can be a good proxy to represent how a rm can
access credit markets.
My contribution is the idea that, during times of crisis such as that of 2008,
nancial development can be a curse. That is, the degree of nancial development
during the recent nancial crisis played a signicant role in worsening rm-level
earnings. Overall, I nd that nancial development increases systemic fragility. I
analyze the role of nancial development at a disaggregated level, using a com-
prehensive European rm-level dataset with almost 10 million rms. Using this
rm-level dataset, I provide strong empirical evidence that rms located in coun-
tries that are more nancially developed experienced a sharper decline in earnings.
5
Existing research on nancial development focuses on the impact of nancial
development on growth and volatility; however, the literature on the role of nan-
cial development during times of nancial crisis is rather scarce. While there is
a literature addressing the role of nancial development in emerging market crises
(Kaminsky and Reinhart, 1996; Dekle and Kletzer, 2001), there is hardly any litera-
ture that relates nancial development to crises in advanced industrialized countries,
such as those in Europe. This paper seeks to identify the signicant relationship
between nancial development and rm-level performance in advanced European
economies.
My baseline empirical specication during the crisis of 2008 is straightforward. I
interact cross-country dierences in nancial development and cross-rm dierences
in dependence on external nancing, and regress rm-level performance on these
interacted variables. The coecient on the interacted variable represents how a
country's nancial development aects how sensitive the rms in that country are
to changes in external nancing such as a nancial crisis. Therefore, by including a
measure of a country's nancial development and dependence on external nancing
of a rm located in that country as explanatory variables, we can identify how
nancial development impacts a rm's earnings.
My results are highly robust to estimation using various instruments for the en-
dogenous variables. The main endogenous variables in my models are dependence
on external nancing and nancial development. I instrument rm-level depen-
dence on external nance using: 1) dependence on external nancing dened at the
6
industry level, and 2) a measure of working capital that is uncorrelated with rm-
level variables, other than measures of external nancing. I instrument nancial
development using the Strength of Legal Rights Index from Djankov et al. (2008).
My nding is that nancial development magnies the impact of a crisis is robust
to these various instruments. As a further robustness check, I estimate a \placebo"
specication in which the country-level nancial development indices are assigned
randomly to the rm-level observations of a country. In the \placebo" specica-
tion, since the nancial development indices are randomly assigned, they should not
matter (be statistically insignicant). Indeed, I nd that in the \placebo" speci-
cations, the coecients on the variables interacted with the nancial development
indices are indeed insignicant. Thus, in the specications when the nancial de-
velopment indices are assigned correctly, we can ascribe same casual eect to the
variables interacted with the nancial development indices.
The rest of this paper is organized as follows: Section 2 reviews the relevant
literature on nancial development and rm-level external nancing. Section 3
describes the data used for the analysis and depicts some descriptive statistics.
Section 4 presents the main empirical ndings and implications of those results.
Section 5 concludes.
2.2 Literature Review
A large literature has related nancial development to economic growth and volatil-
ity. However, the literature is mostly silent about the relationship between nancial
7
development and rm-level performance during times of crises. Typically this lit-
erature has found that nancial development and volatility are negatively related.
Rajan and Zingales (1998) nd that an industry which is more dependent on ex-
ternal nancing grows faster, with better nancial development. Since there are
industries which require either a high volume of investment during the gestation
period or capital investment in the industry's production phase, nancial devel-
opment promotes growth in those industries. Moreover, this literature nds that
nancial development reduces GDP volatility (Larrain, 2004). I extend this lit-
erature and explore how nancial development contributes to rm-level earnings
separately during both tranquil and crisis times.
My paper is also related to the literature on the eects of nancial liberalization
in emerging market crises. While easier access to foreign capital should help emerg-
ing market countries better weather the nancial storms, the literature shows { as
does this paper { that nancial liberalization actually destabilized the emerging
market economies during crises. Kaminsky and Reinhart (1996) and Demirguc-
Kent and Detragiache (1998) show that nancial liberalization precedes nancial
crises. Furman and Stiglitz (1998) state that \the evidence that nancial liberal-
ization increases the vulnerability of countries to crises is overwhelming". Kroszner
et al. (2007) look at 38 countries and investigate whether the impact of a nan-
cial crisis on sectors dependent on external nancing varies with dierent levels of
nancial development. Their ndings suggest that rms in nancially vulnerable
sectors are hit harder than rms that are in other sectors. Beck et al. (2006) show
8
that rms which depend more on external nancing suer more through the bank
lending channel when their economy is hit by a large shock. The results in this
paper are complementary to this literature. Surprisingly, I show that the results for
emerging market countries carry over to the more developed markets of Europe. I
nd that nancial liberalization, which spurs nancial market development, causes
systemic volatility and negatively impacts emerging market countries during crisis.
It is surprising since we would expect that advanced economies such as Europe will
have the institutions such as legal rules to limit the negative eects of nancial crisis
on their rms.
2.3 Description of Data
I use the Amadeus database from 2006 to 2008 for the rm-level data in this paper.
The Amadeus is a pan-European nancial database containing information on over
10 million public and private companies from 41 countries. The database contains
nancial, ownership and descriptive information for about 5,000,000 rms in 34
countries, including all the EU countries and Eastern Europe. The Amadeus data
is distinctive from the rm-level data used in previous studies. Previous studies use
the Compustat Global database, which covers 33,900 active and inactive publicly
held companies outside the U.S. and Canada. The available data for the Amadeus
dates from 1984, but it lacks the comprehensiveness and consistency of company
observations for each country, which is necessary for any analysis of data before
9
2000 to be signicant. Therefore, for the main analysis it considers data from the
period 2006 to 2008, to capture the eects of the recent nancial crisis.
In the Amadeus database, company activity codes provide information on their
industry classication. The NAICS core code is used to classify a rm by industry.
Since we are only concerned with non-nancial rms, nance insurance and public
administration (sectors 52 and 92, respectively) are dropped. In addition, rms
that do not have information on assets, earnings, date of incorporation, number of
employees and industry classication, are excluded from the analysis.
To evaluate the role of nancial development on rm-level performance, we
need a measure of nancial development and rm-level earnings. As a measure of
nancial development for each country, I use the ratio of private credit to GDP
from the World Bank Development Indicator (WDI). Since some countries do not
have information on this, those are also excluded from the analysis. Therefore,
after merging this nancial development data with the Amadeus, the number of
rms decreases. The nal dataset used contains 22,181,156 observations: 8,657,573
observations in 2006, 9,409,631 in 2007 and 4,113,952 in 2008. As a measure of
rm performance, I consider Earnings Before Interest and Tax (EBIT) which is the
dierence between revenue and expenses before taking into account interest and
taxes; EBIT is also called operating prot.
Besides nancial development and rm-level earnings, we also need to decide
how a rm's dependence on external nancing is measured. Rajan and Zingales
(1998) dene dependence on external nance as capital expenditures minus cash
10
ow from operations divided by capital expenditures, as shown in the following
equation (1).
DEF =
capitalexpenditurescashflowfromoperations
capitalexpenditures
: (2.1)
A negative DEF means that the operating revenues exceed the capital expen-
ditures, and a rm does not need to depend on external nancing. However, when
the DEF is positive, a rm needs to raise funds from outside - either by borrowing
or by issuing debt. Although both Amadeus and Compustat provide data on cash
ow from operations, Amadeus only tracks total changes in assets and not total
capital expenditures, \capital expenditures" in the equation with \changes in total
assets:"
DEF =
changesintotalasset cashflowfromoperations
changesintotalassets
: (2.2)
In the actual regressions I include the rst dierence of DEF, IDEF, as a regres-
sor. If a rm's changes in assets are funded by its operating revenues, its IDEF is
denoted as 1, otherwise its IDEF is denoted as 0.
In addition to capturing the eects of other rm-level characteristics in deter-
mining rm-level earnings, rm size and age are also considered. There are dierent
ways to dene a rm's size. For example, rm size can be dened by asset value. In
the following analysis, the size dummy is dened based on the number of employees.
11
If the rm has more than 250 employees, it is considered a large rm with a variable
of 1; otherwise it is considered a small rm with a variable of 0.
To investigate the recent credit crisis impacts on rm-level outcomes, the crisis
dummy is dened as 1 if the year is 2008; otherwise it is dened as 0. Firm age is
calculated from the date of incorporation.
Finally, industries are classied into four sectors using the NAICS core code as
in Table 2.1. I use agricultural and mining as the base, and I introduce dummy
variables for the remaining sectors.
Table 2.1: Industry Classication
NAICS code Sector
11 and 21 Agricultural and mining
22, 23, 53 Related to housing
31-33, 42, 44-49 Manufacturing
51-56, 61-62, 71-72 and 81 Services
Before proceeding, I describe the two sets of data and provide some summary
statistics. The rst set is based on those companies that continuously reported their
information from 2006 to 2008, and is called the working sample { there are 96,709
observations in each year for the working sample; the second set considers the whole
data sample during this time period { there are 675,394 observations from 2006 to
2008.
12
Table 2.2 and Table 2.3 show a summary of the size and dependence on external
nancing for the working sample and the sample. About 20% of them are considered
large rms and the rest are small. For the sample, about 81% of rms in the data
are small and the remaining 19% are considered large. Even if the number of rms
is dierent between the two sets of data, the distribution of rm size is similar,
considering that about 20% of rms are small in both datasets.
Table 2.2: Description of Firm Size
Description of Working Sample Description of the Sample
2006 2007 2008 Total 2006 2007 2008 Total
Small 77,555 76,821 75,789 230,165 239,993 224,586 81,170 545,749
Large 19,154 19,888 20,920 59,962 53,924 52,546 23,175 129,645
Total 96,709 96,709 96,709 290,127 293,917 277,132 104,345 675,394
Table 2.3 also summarizes dependence on external nancing by rm size in each
year. The working sample tells us that in 2006, 60% of small rms and 48% of large
rms depended on external nancing. In contrast, 31% of small rms and 30% large
rms required access to external nancing in 2008. Similarly, the sample also shows
that small rms tend to be more dependent than large rms on external nancing.
Table 2.4 shows the number of rms according to its age bin based on the
working sample. As Rajan and Zingales (1998) already note, a rm's dependence
on external nancing is correlated with its age. Since young rms tend to be more
dependent on external nancing for initial investment, it is important to make sure
that there are enough observations in each age cohort in the data. Even though
rms under 5 years old and over 15 years old account for 72% of the data, there
13
Table 2.3: Firm Size and Dependence on External Financing
2006 2007 2008
Dependent Not Dependent Dependent Not Dependent Dependent Not Dependent
Description of Working Sample
Small 46,353 31,202 34,838 41,983 23,238 52,551
Large 9,180 9,974 7,660 12,228 6,183 14,737
Total 55,533 41,176 42,498 54,211 29,421 67,288
Description of the Sample
Small 152,991 87,002 53,598 170,988 36,936 44,234
Large 27,880 26,044 8,299 44,247 11,041 12,134
Total 180,871 113,046 61,897 215,235 47,977 56,368
exists a sucient number of observations to test whether a rm's age explains the
variation in earnings.
Table 2.4: Number of Firms per Age-Cohort
Country Code
Age-Cohorts
Number of rms
<5 [5,9] [10,14] 15<
Austria 0.26 0.14 0.12 0.48 1,924
Belgium 0.07 0.13 0.13 0.67 6,783
Bulgaria 0.21 0.21 0.30 0.27 56
Switzerland 1.00 0.00 0.00 0.00 1
Czech Republic 0.12 0.22 0.48 0.18 1,107
Germany 0.69 0.05 0.05 0.21 8,558
Denmark 0.20 0.23 0.13 0.44 4,495
Estonia 0.13 0.26 0.39 0.21 416
Spain 0.06 0.12 0.16 0.67 454
Finland 0.10 0.15 0.14 0.60 2,098
France 0.44 0.14 0.06 0.36 14,080
United Kingdom 0.14 0.21 0.14 0.51 22,617
Greece 1.00 0.00 0.00 0.00 2,599
Croatia 1.00 0.00 0.00 0.00 5
Ireland 0.12 0.21 0.17 0.50 510
Iceland 0.20 0.20 0.20 0.40 10
Italy 0.13 0.18 0.11 0.59 6,811
14
Table 2.4 { Continued
Country Code
Age-Cohorts
Number of rms
<5 [5,9] [10,14] 15<
Lithuania 0.10 0.23 0.53 0.13 279
Luxembourg 0.00 0.18 0.36 0.45 11
Latvia 0.11 0.24 0.46 0.18 383
Republic of Macedonia 0.10 0.24 0.43 0.24 21
Malta 0.00 0.00 0.50 0.50 2
Norway 0.19 0.23 0.15 0.43 3,664
Poland 1.00 0.00 0.00 0.00 2,070
Portugal 0.06 0.13 0.16 0.66 508
Russian Federation 0.17 0.15 0.13 0.55 1,683
Sweden 1.00 0.00 0.00 0.00 7,650
Slovenia 0.05 0.09 0.20 0.66 512
Slovakia 0.09 0.22 0.49 0.20 90
Ukraine 0.32 0.30 0.37 0.01 7,312
Total 0.35 0.15 0.12 0.37 96,709
Finally, Table 2.5 provides a comparison of the rms' earnings before and after
the crisis, based on the working sample. Two descriptive statistics are provided:
the mean and the median of the rms' earnings per country, only for those rms
that reported their earnings in all three years. The gures are bolded for those
countries in which rms experienced decreased earnings after the crisis. Out of
thirty countries, eighteen experienced a decline in the mean. Twelve countries
underwent a decline in rm-level earnings based on median. This indicates that
even if a credit crunch hits the U.S. and the rest of the world, its impact on each
country is unequal.
15
Table 2.5: Description of EBIT Before and After the Crisis
Country
Mean Median
Before After Before After
Austria 130.33 160.45 8.84 49.92
Belgium 312.36 399.19 0.79 0.81
(Bulgaria) 4.69 0.71 0.80 0.71
(Switzerland) 0.0007 0.0006 0.0007 0.0006
Czech Republic 3.67 4.31 0.00 0.00
(Germany) 119.21 118.47 1.59 1.87
(Denmark) 11.96 14.45 1.31 1.16
(Estonia) 3.78 2.50 1.27 0.92
Spain 28.31 36.98 1.28 1.76
(Finland) 18.00 13.86 1.26 1.50
(France) 19.42 18.99 0.78 0.96
(United Kingdom) 305.40 183.20 1.09 1.13
(Greece) 5.67 4.84 1.33 1.51
(Croatia) 0.50 -5.62 0.22 1.32
(Ireland) 14,139.42 2,493.70 1.59 1.28
(Iceland) 12.41 12.25 9.31 4.89
Italy 5.29 5.60 0.86 1.00
Lithuania 3.62 3.76 1.03 1.16
(Luxembourg) -6.64 -12.11 0.93 2.49
(Latvia) 3.28 2.38 0.92 0.64
(Republic of Macedonia) 7.17 7.18 0.91 0.27
(Norway) 21.89 21.83 0.0012 0.0011
(Poland) 5.28 4.03 0.0011 0.0013
(Portugal) 20.25 19.34 1.00 1.24
Serbia 0.56 0.64 0.09 0.10
(Sweden) 10.00 8.00 0.0011 0.0013
Slovenia 4.20 5.48 1.05 1.26
Slovakia 1.39 2.08 0.74 0.75
(Ukraine) 18.49 -10.96 0.14 0.09
16
Table 2.5 { Continued
Country
Mean Median
Before After Before After
(ALL) 177.37 98.79 0.51 0.54
2.4 Empirical Specications
This section explores the impact of nancial development on rm-level earnings
over the period 2006-2008. I present the results using multilevel regression model:
a multilevel regression model is equivalent to a hierarchical linear model, a mixed-
eects model and a variance component model (Hox, 2010). According to Gelman
and Hill (2007):
The two key parts of a multilevel model are varying coef-
cients, and a model for those varying coecients (which can
itself include group-level predictors). Classical regression can
sometimes accommodate varying coecients by using indicator
variables. The feature that distinguishes multilevel models from
classical regression is in the modeling of the variation between
groups.
Hence, the multilevel regression model enables us to model the variation between
countries based on dierent legal systems or economic environments. By integrating
the level of a country's nancial development with rm-level explanatory variables
- specically, a rm's dependence on external nancing - we can examine how
nancial development aects a rm's dependence on external nancing in times
17
of crisis and tranquility. The main analysis uses a rm's dependence on external
nancing; industry-level dependence on external nancing is introduced to conduct
a robustness check. The main analysis also examines the ndings from the multilevel
regression using a panel approach, which assumes rm random and xed eects.
This section begins by providing empirical evidence based on the multilevel
regression model using (i) rm-level dependence on external nancing (ii) industry-
level dependence on external nancing and (iii) working capital.
2.4.1 Multilevel Regression
Determinants of Firm-level Earnings based on Firm and Country Level
The main goal of the following analysis is to nd the relationship between nancial
development and rm-level earnings. After examining this relationship, I will ex-
plore the question of whether nancial development has the same eects during a
crisis period. Based on dierent specications, I examine whether nancial develop-
ment and rm dependence on external nancing exhibit any signicant relationship
with rm-level earnings. Thus, if nancial development promotes rm-level earn-
ings, the coecient on nancial development will appear as a positive sign. If the
IDEF coecient appears as a positive sign, it means that rms which do not de-
pend on external nancing tend to have higher earnings than those which depend
on external nancing. However, if the coecient of the interaction term of nancial
development and IDEF appears as a negative sign and is statistically signicant, it
implies that nancial development has a dierent impact on rms according to their
18
dependence on external nancing, and that nancial development is more benecial
to those that depend on external nancing.
The empirical specication to test the relationship between nancial develop-
ment and rm-level earnings can be written as follows:
logEBIT
ict
=
0
+
1
FD
ct
+
2
IDEF
ict
+
3
IDEFFD
ct
+e
ict
(2.3)
wherei denotes a rm,c a country andt year. When a country level random eects
is assumed, equation (2.3) can be rewritten as follows:
logEBIT
ict
=
c
+
1
FD
ct
+
2
IDEF
ict
+
3
IDEFFD
ct
+u
ct
+e
ict
(2.4)
According to equation (2.4), the intercept varies with each country since rms
within the same country share a common legal system and other social and economic
characteristics which may in
uence their outcomes. In addition to rm dependence
on external nancing, rm size and age must also be considered in order to assess
the robustness of the ndings on nancial development and rm-level earnings.
The inclusion of rm size and age will enable us to determine the exact relationship
between a rm's size and its earnings. The existing literature points out that large
rms tend to be more protable. With respect to rm age, Rajan and Zingales
(1998) note that besides the size, the age of a rm also matters in deciding its
capital requirement. Young rms tend to be more dependent on external nancing.
Therefore, we need to consider the eect of those factors in determining rm-level
19
earnings. Once examined, I try to nd factors that explain the changes in rm-
level earnings during the crisis period.
The results in Table 2.6 indicate that overall nancial development has a positive
impact on rm-level earnings. Across dierent specications, even if the magnitudes
dier depending on specications, the table results show that the coecient of
nancial development consistently appears as a positive sign and is statistically
signicant. This provides empirical evidence that if all other things are equal, a
rm with better nancial development is likely to have higher earnings than a rm
with lower nancial development.
Regarding dependence on external nancing, according to the results, a rm
that does not depend on external nancing should be better o than a rm that
depends on external nancing. This result holds across dierent specications, and
is statistically signicant. The interaction term between nancial development and
dependence on external nancing tells us that the slope for nancial development is
signicantly dierent in the two groups { those which depend on external nancing
and those which do not. In other words, a greater coecient on nancial devel-
opment suggests that in order to sustain operations, a rm can be more sensitive
to the availability of borrowing in the economy. For example, based on the rst
column in the table, if a rm does not depend on external nancing, its slope for
nancial development is 0.01175, but if it does, then the slope will be 0.01201.
When there are two rms which share the same characteristics and both of which
need to borrow for their production or investment, if one belongs to a country with
20
Table 2.6: Results based on Firm Earnings
Dependent Variable: logEBIT
(1) (2) (3) (4) (5) (6) (7) (8)
Intercept
11.9692*** 12.2606*** 12.0865*** 12.2934*** 11.847*** 11.718*** 11.983*** 11.8098***
(0.021720) (0.457700) (0.037990) (0.459200) (0.023090) (0.456400) (0.038510) (0.456900)
[0.022830] [0.612700] [0.038440] [0.621900] [0.024790] [0.584800] [0.039580] [0.587900]
FD
0.01201*** 0.007533** 0.01203*** 0.007439*** 0.01198*** 0.008289*** 0.01198*** 0.008205***
(0.000157) (0.000563) (0.000158) (0.000563) (0.000170) (0.000536) (0.000171) (0.000536)
[0.000160] [0.003006] [0.000163] [0.002944] [0.000179] [0.002085] [0.000181] [0.002033]
IDEF
0.2773*** 0.5908*** 0.2772*** 0.6015*** 0.2749*** 0.4595*** 0.2747*** 0.4666***
(0.011640) (0.027540) (0.011650) (0.027530) (0.011690) (0.026120) (0.011690) (0.026120)
[0.012400] [0.144400] [0.012410] [0.146400] [0.012450] [0.155700] [0.012450] [0.156700]
FD*IDEF
-0.00026*** -0.00174*** -0.00026*** -0.00172*** -0.00028*** -0.00154*** -0.00028*** -0.00151***
(0.000093) (0.000212) (0.000093) (0.000212) (0.000093) (0.000201) (0.000093) (0.000201)
[0.000097] [0.001401] [0.000097] [0.001396] [0.000097] [0.001516] [0.000097] [0.001515]
Size
0.5776*** 2.3369*** 0.5763*** 2.3263***
(0.030040) (0.030120) (0.030060) (0.030130)
[0.036780] [0.675200] [0.036840] [0.666800]
Size*FD
0.000057 -0.00373*** 0.000078 -0.00366***
(0.000231) (0.000225) (0.000231) (0.000226)
[0.000266] [0.005084] [0.000267] [0.005027]
Manufacturing
-0.06056 0.05919* -0.1604** -0.0367
(0.071310) (0.035120) (0.070200) (0.033220)
[0.064870] [0.171700] [0.064150] [0.185600]
Housing
-0.125*** -0.08567*** -0.1399*** -0.1325***
(0.035640) (0.017150) (0.035070) (0.016240)
[0.034920] [0.067360] [0.034440] [0.076070]
Services
-0.1603*** 0.09093*** -0.1779*** -0.01409
(0.038050) (0.018370) (0.037440) (0.017400)
[0.037930] [0.153300] [0.037400] [0.141000]
Country Random NO YES NO YES NO YES NO YES
1)
* signicant at 10%; ** signicant at 5%; *** signicant at 1%.
2)
The size denotes the size dummy if it is a large rm the size is 1 otherwise it is 0.
3)
FD denotes nancial development, dened as private credit to GDP.
4)
Standard errors assuming homoscedastic errors in parentheses and heteroscedastic.
4)
Robust standard errors in square brackets.
21
a relatively well-developed nancial market, its earnings will be greater than the
one in a country that has a relatively less-developed nancial market. The results
provide empirical evidence for nancial development's positive impact on rm-level
earnings, and shows benets to those rms which need access to external nancing.
From the sixth to eighth column, the analysis includes the rm size and the in-
teraction between size and nancial development. The interaction term of nancial
development and size is negative and signicant only when a country random eects
model is assumed. A negative interaction term implies that there is a distinction
between small and large rms in the degrees to which they benet from nancial
development. A negative sign shows that better nancial development tends to be
more benecial to small rms than to large ones. This nding possibly re
ects the
fact that small rms tend to depend on external nancing more than large rms
do.
For rm size, all the signs are not only consistently positive across dierent
specications but they are also statistically signicant. These specications capture
the relationship between rm size and earnings in the existing literature, which is
that larger rms tend to have higher earnings. Even if the size dummy is included
in the analysis, the coecient of the variable does not
uctuate dramatically. For
each model, the intraclass correlation is 0.50, 0.50, 0.53 and 0.50, respectively. Thus,
about 0.50-0.53% of the variance of the log(EBIT) is at the country level, which is
very high and statistically signicant.
22
The results in Table 2.6 indicate that nancial development can aect rm-level
earnings, and its impact on each rm is unequal, depending on a rm's dependence
on external nancing. In other words, nancial development itself plays a positive
role but depending on a rm's size and age, as the interaction terms indicate, its
impact on a rm's earnings will vary. As discussed, most of the results conrm the
ndings in the existing literature on rm dynamics. Based on the results in the
table, I now want to answer the following question: How does the current crisis
have an impact on rm-level outcomes with dierent levels of nancial development
and dierent characteristics?
When a credit crisis hits the economy, does nancial development still explain
a rm's earnings as it does before the credit crisis? To tackle this question more
deeply, I include the crisis dummy variable and its interaction term with other
variables in the equation below. This will enable us to see whether nancial devel-
opment still plays a positive role during the crisis, and what factors are responsible
for explaining a diminution in a rm's earnings. In the following analysis, if the
coecient of the interaction of nancial development and the coecient of the crisis
dummy appear positive and statistically signicant, we can conclude that nancial
development serves as a safeguard; during the crisis, nancial development still
promotes higher earnings rather than lowering earnings. By contrast, if the coe-
cient of the interaction of nancial development appears negative, it indicates that
nancial development fails to act as a buer against shocks during the recent crisis.
23
Similar to the equation (2.4), a new empirical specication can be written with
the crisis dummy variable and its interaction with other variables, while assuming
country-random eects:
logEBIT
ict
=
c
+
1
FD
ct
+
2
IDEF
ict
+
3
IDEFFD
ct
+
4
size
ict
(2.5)
+
5
crisisdummy +
6
FDcrisisdummy +u
ct
+e
ict
Similarly, when a crisis random eects model is assumed it can be also written:
logEBIT
ict
=
c
+
1
FD
ct
+
2
IDEF
ict
+
3
IDEFFD
ct
+
4
size
ict
(2.6)
+
5
crisisdummy +
6
FDcrisisdummy +u
crisis;ct
+e
ict
In equations (2.5) and (2.6), we see a distinctive feature: they now try to capture
the eect of the crisis, and the factors that are responsible for explaining the changes
in earnings over the period. By allowing the intercept to vary by country, equation
(2.5) assumes that there are country level factors in determining rm-level earnings.
Beyond country level eects, because the recent credit crisis has an uneven impact
across countries, I also assume that the coecient of the crisis dummy variable
may be dierent in equation (2.6). If the crisis contributes to changes in earnings
positively, then the coecient of the crisis dummy variable will appear positive.
If nancial development helps rms to withstand a shock during a nancial crisis,
then the interaction between nancial development and the crisis dummy should
also be a positive sign.
24
Table 3.12 shows the results based on dependence on external nancing, nan-
cial development, rm size, and industry dummies, including the crisis dummy and
its interaction with other variables. The results are somewhat puzzling consider-
ing that the all of the values of the crisis dummy appear statistically signicant
and positive except the rst specication. This raises the question of whether all
specications and variables are dened properly. To address this issue, I conduct
an analysis later in the paper using the industry median for rm DEF and log
transformation as a measure of a rm size. Consistent with my previous ndings,
nancial development and rm size have positive coecients, and the interaction
terms of nancial development and size and nancial development and dependence
on external nancing, respectively, are each are also positive.
The interaction term between the crisis dummy and nancial development is
negative and statistically signicant. The results demonstrate that the coecient
of nancial development is signicantly dierent before and after the crisis. Its neg-
ative sign also indicates that during the crisis, nancial development has a smaller
slope than before the crisis. For example, if a rm does not depend on external
nancing and undergoes the crisis, its slope for nancial development is 0.0101.
However, under the same settings, if a rm depends on external nancing then its
coecient for nancial development is 0.0105, which is a little larger than before the
crisis. As the interaction term tells us, there is a dierence between the two periods
- the crisis and pre-crisis - in how nancial development aects a rm's earnings.
25
Table 2.7: Results based on Firm Earnings including the Crisis Dummy
Dependent Variable: logEBIT
(1) (2) (3) (4) (5) (6)
Intercept
11.9183*** 12.7194*** 11.5289*** 12.3199*** 11.4789*** 11.7397***
(0.023710) (0.478800) (0.025950) (0.472900) (0.040420) (0.461100)
[0.023940] [0.605400] [0.026320] [0.618100] [0.040450] [0.688000]
FD
0.01253*** 0.003108*** 0.01594*** 0.006709*** 0.01628*** 0.00897***
(0.000184) (0.000804) (0.000206) (0.000985) (0.000207) (0.000932)
[0.000177] [0.002813] [0.000201] [0.002922] [0.000205] [0.003122]
IDEF
0.2732*** 0.61*** 0.3178*** 0.6202*** 0.3141*** 0.5198***
(0.011680) (0.027650) (0.011690) (0.027690) (0.011730) (0.026220)
[0.012440] [0.145300] [0.012410] [0.147000] [0.012460] [0.132500]
FD*IDEF
-0.00026*** -0.00175*** -0.0007*** -0.00189*** -0.00074*** -0.00192***
(0.000093) (0.000212) (0.000093) (0.000213) (0.000094) (0.000202)
[0.000097] [0.001387] [0.000096] [0.001391] [0.000097] [0.001283]
Size
0.5983*** 1.882***
(0.014100) (0.012530)
[0.016650] [0.377300]
Crisis
-0.02511*** 0.1258*** 0.3136*** 0.2813*** 0.3102*** 0.2667***
(0.004756) (0.016310) (0.010430) (0.029540) (0.010470) (0.027950)
[0.004818] [0.026670] [0.011710] [0.050960] [0.011750] [0.059260]
Crisis*FD
-0.00297*** -0.00159*** -0.00305*** -0.00195***
(0.000082) (0.000252) (0.000082) (0.000239)
[0.000088] [0.000470] [0.000088] [0.000555]
Manufacturing
-0.05207 -0.03614
(0.070010) (0.033240)
[0.063630] [0.184500]
Housing
-0.09648*** -0.1409***
(0.034950) (0.016240)
[0.034180] [0.082420]
Services
-0.1941*** -0.02081
(0.037280) (0.017400)
[0.037160] [0.148600]
Country Random NO YES NO YES NO YES
The crisis denotes the crisis dummy.
26
To examine whether nancial development helps rms to weather a shock, or
if better nancial development prevents rms from a dramatic decrease in their
earnings, one needs to pay attention to the interaction term of nancial development
and the crisis dummy. The sign for the interaction term of nancial development
and the crisis dummy is negative across all dierent specications except the rst.
We can surmise that a decrease in a rm's earnings is its coecient times nancial
development during the crisis. In other words, more nancial development will bring
lower earnings during the crisis. For example, if a country's nancial development is
denoted as 0, its impact on the crisis is 0.3136, based on the third column. However,
if it is 200, then its impact is -0.2804. Therefore, even if the crisis dummy coecient
is positive, the credit crisis can actually have a negative impact. Thus, the impact
of the crisis can be magnied by nancial development.
The results so far show that a rm's age is inversely related to its earnings. As
a robustness check, we use the alternative measure of a rm's age that is dened
as
p
Age and examine whether the results from the previous specications remain
the same. Table 2.9 shows the results. It conrms that even if we use the dierent
measure of a rm's age, it is still negatively related to a rm's earnings.
Robustness Checks using Instrumental Variables
The ndings from the previous section are not well documented in the existing
literature. Claessens, Tong and Wei (2010) recently found evidence which bolsters
ndings in this paper. Instead of focusing on dependence on external nancing
and nancial development in explaining on rm performance, they pay attention
27
Table 2.8: Results based on Firm Earnings
Dependent Variable: logEBIT
(1) (2) (3) (4) (5) (6)
Intercept
13.7132*** 13.5067*** 13.5176*** 13.2663*** 13.6437*** 13.35***
(0.013120) (0.371700) (0.013090) (0.378300) (0.020270) (0.376900)
[0.401100] [0.382300] [0.425900] [0.402900] [0.491200] [0.422900]
IDEF
-0.5972*** 0.5002*** -0.6759*** 0.421*** -0.6219*** 0.418***
(0.020990) (0.024510) (0.020480) (0.023620) (0.020680) (0.023640)
[0.328600] [0.148500] [0.326800] [0.137500] [0.327900] [0.137200]
FD*IDEF
0.008397*** -0.00044** 0.008455*** -0.00049*** 0.008136*** -0.00048**
(0.000141) (0.000178) (0.000137) (0.000171) (0.000138) (0.000171)
[0.002078] [0.001003] [0.002365] [0.000963] [0.002268] [0.000967]
Size
1.1288*** 1.4899*** 1.1515*** 1.4936***
(0.016990) (0.016500) (0.017010) (0.016510)
[0.285400] [0.233500] [0.265900] [0.227800]
Age
0.01646*** 0.009725*** 0.01525*** 0.005803*** 0.01555*** 0.005706***
(0.000380) (0.000365) (0.000370) (0.000352) (0.000372) (0.000355)
[0.004685] [0.002644] [0.005809] [0.002017] [0.005582] [0.002473]
Crisis
-0.6327 0.2058*** -0.828 0.2099*** -0.803 0.2092***
(0.709200) (0.029520) (0.718900) (0.028770) (0.709000) (0.028770)
[0.605900] [0.029860] [0.582100] [0.035440] [0.575800] [0.034760]
Crisis*Age
-0.00851*** -0.00143** -0.01123*** -0.00148** -0.01061*** -0.00147**
(0.000682) (0.000628) (0.000668) (0.000604) (0.000667) (0.000603)
[0.003722] [0.000300] [0.005198] [0.000299] [0.004747] [0.000288]
Crisis*FD
0.006237 -0.00032* 0.007195 -0.00041*** 0.007013 -0.0004**
(0.005297) (0.000176) (0.005368) (0.000169) (0.005294) (0.000169)
[0.003669] [0.000113] [0.003576] [0.000126] [0.003537] [0.000128]
Crisis*Size
0.3424*** -0.01185 0.3107*** -0.01207
(0.030900) (0.027640) (0.030880) (0.027640)
[0.272400] [0.054020] [0.237400] [0.053060]
Manufacturing
-0.7243*** -0.09522***
(0.035980) (0.034250)
[0.234400] [0.202800]
Housing
-0.2021*** -0.08309***
(0.018650) (0.017430)
[0.103700] [0.072110]
Services
-0.0338*** -0.1156***
(0.020150) (0.018790)
[0.168400] [0.146800]
Country Random NO YES NO YES NO YES
Crisis Random YES NO YES NO YES NO
28
Table 2.9: Results based on Firm Earnings using
p
Age
Dependent Variable: logEBIT
Dep. Variable (1) (2) (3) (4)
Intercept
11.5289*** 12.3199*** -1.4133*** -1.0329***
(0.0260) (0.4729) (0.0399) (0.1194)
FD
0.0159*** 0.006709*** 0.0022*** -0.0021***
(0.0002) (0.0010) (0.0001) (0.0006)
IDEF
0.3178*** 0.6202*** 0.1236*** 0.3102***
(0.0117) (0.0277) (0.0117) (0.0170)
FD*IDEF
-0.0007*** -0.0019*** -0.0001 0.000034
(0.0001) (0.0002) (0.0001) (0.0001)
Size
0.8965*** 0.907***
(0.0024) (0.0018)
p
Age
-0.0097 -0.0144***
(0.0028) (0.0019)
Crisis
0.3136*** 0.2813*** 0.0046 0.0284
(0.0104) (0.0295) (0.0108) (0.0182)
Crisis*FD
-0.003*** -0.0016*** -0.0004*** 0.0002
(0.0001) (0.0003) (0.0001) (0.0001)
Random Intercept NO YES NO YES
to the role of working capital. They use the Cash Conversion Cycle (CCC) as a
measure of working capital. The CCC measures how long a rm will be deprived
of cash if it increases its investment in resources in order to expand customer sales.
Therefore, the CCC is used as a measure of a rm's liquidity position. The inclusion
of the CCC will help us to explain of how a rm's liquidity condition plays a role in
determining rm-level earnings along with rm dependence on external nancing.
Following their method, I also dene the CCC at 2-digit NAICS code to capture a
particular liquidity condition. It can be calculated as follows:
CCC = 365 (
inventoriesaccountspayables
costof goodssold
+
accountsreceivables
totalsales
) (2.7)
Since those items are not available on the Amadeus, the Compustat Fundamental
Annual from 2000-2010 is used instead, and its median value of the CCC in each
sector is used.
29
The results in Table 2.10 indicate that when a country random eects model
is considered, the interaction of nancial development and the crisis dummy be-
comes statistically signicant and has a negative sign. The CCC turns out to have
a negative sign across dierent specications and it is statistically signicant. This
means that when a rm faces low liquidity it is more likely to have comparatively
lower earnings than others. The ndings so far indicate that dependence on exter-
nal nancing, dened at the rm or industry level, and liquidity conditions a rm
faces, turn out to be signicant in determining a rm's earnings. Financial devel-
opment, which is closely related to dependence on external nancing and liquidity
conditions, has a positive impact on rm-level earnings; however, the magnitude
of its eect varies depending on a rm's other characteristics. For example, nan-
cial development is more benecial to a rm that depends on external nancing.
Most surprisingly, after taking into account the dependence on external nancing
and the liquidity condition, the results clearly point out that nancial development
aggravates the impact of the nancial crisis. Throughout all specications, the cri-
sis itself does not play a role but when combined with nancial development, it
propagates shock throughout the economy. As discussed above, with more nancial
development, a rm may experience more a severe decline in its earnings.
I noted earlier that a credit crisis is accompanied by a tightening in lending;
therefore, changes in IDEF may re
ect this tightening rather than the commonly
assumed dependence on external nancing at rm-level, a possibility which raises
concerns about endogeneity. One way to resolve this issue is to replace IDEF with
30
Table 2.10: Results based on Firm Earnings
Dependent Variable: logEBIT
(1) (2) (3) (4) (5) (6)
Intercept
13.6855*** 13.4761*** 13.4799*** 13.2408*** 11.6152*** 12.5564***
(0.013440) (0.373400) (0.013420) (0.381600) (0.026180) (0.356800)
[0.391200] [0.379400] [0.418900] [0.399600] [0.583000] [0.444100]
IDEF
-0.8448*** 0.6197*** -0.9679*** 0.5375*** 0.5179*** 0.5461***
(0.024820) (0.029280) (0.024230) (0.028280) (0.029750) (0.028320)
[0.313700] [0.149100] [0.303300] [0.137800] [0.120500] [0.137300]
FD*IDEF
0.01012*** -0.00124*** 0.01056*** -0.00121*** -0.00132*** -0.00143***
(0.000167) (0.000209) (0.000162) (0.000201) (0.000214) (0.000204)
[0.001973] [0.001000] [0.002203] [0.000947] [0.000830] [0.000977]
FD
0.01563*** 0.006059***
(0.000190) (0.001005)
[0.003431] [0.001881]
Size
1.1603*** 1.4528*** 1.2099*** 1.4637***
(0.017220) (0.016770) (0.016810) (0.016870)
[0.267500] [0.226300] [0.192400] [0.229600]
Age
0.01619*** 0.009525*** 0.0151*** 0.005857*** 0.01226*** 0.005703***
(0.000389) (0.000375) (0.000378) (0.000362) (0.000371) (0.000363)
[0.004407] [0.002118] [0.005612] [0.001640] [0.004193] [0.001603]
CCC
-0.00617*** -0.00227*** -0.00721*** -0.00296*** -0.00402*** -0.00289***
(0.000274) (0.000261) (0.000267) (0.000252) (0.000263) (0.000252)
[0.001580] [0.001128] [0.001750] [0.001040] [0.001087] [0.001043]
Crisis
-0.5961 0.225*** -0.8004 0.2272*** 0.2274 0.2511***
(0.767400) (0.033630) (0.768400) (0.032860) (0.753100) (0.033100)
[0.607200] [0.031270] [0.582000] [0.036620] [0.436200] [0.057420]
Crisis*Age
-0.00836*** -0.00149** -0.01095*** -0.00154** -0.00812*** -0.00116*
(0.000700) (0.000645) (0.000684) (0.000621) (0.000668) (0.000624)
[0.003830] [0.000356] [0.005236] [0.000316] [0.003361] [0.000504]
Crisis*FD
0.005938 -0.00034* 0.007088 -0.00042* -0.00154 -0.00135***
(0.005661) (0.000209) (0.005667) (0.000201) (0.005554) (0.000254)
[0.003708] [0.000121] [0.003609] [0.000131] [0.002317] [0.000403]
Crisis*Size
0.2751*** -0.00718 0.2386*** -0.03521
(0.031310) (0.028100) (0.030550) (0.028480)
[0.258400] [0.053410] [0.170000] [0.041000]
Crisis*CCC
0.004007*** 0.000408 0.004614*** 0.00068 0.001758*** 0.000596
(0.000485) (0.000454) (0.000472) (0.000438) (0.000462) (0.000438)
[0.001380] [0.000313] [0.001435] [0.000228] [0.000540] [0.000239]
Country Random NO YES NO YES NO YES
Crisis Random YES NO YES NO YES NO
31
Table 2.11: Results based on Firm Earnings using Industry DEF
(1) (2) (3) (4) (5) (6)
Intercept
11.5459*** 11.8692*** -1.6615*** -1.7245*** -1.6628*** -1.7268***
(0.02705) (0.46650) (0.04152) (0.11940) (0.04160) (0.11940)
FD
0.01711*** 0.01326*** 0.002398*** 0.003748*** 0.002407*** 0.003761***
(0.00021) (0.00100) (0.00011) (0.00057) (0.00011) (0.00057)
DEF
0.03664 0.02523* -0.00641 -0.01075
(0.02239) (0.01441) (0.01288) (0.00915)
FD*DEF
0.000231 0.000201* -0.00009** -0.00005** -0.00005 0.000016
(0.00015) (0.00011) (0.00004) (0.00002) (0.00009) (0.00006)
log(Size)
0.916*** 0.9221*** 0.916*** 0.9221***
(0.00256) (0.00189) (0.00256) (0.00189)
Age
-0.0021*** -0.00247*** -0.0021*** -0.00248***
(0.00028) (0.00019) (0.00028) (0.00019)
Crisis
0.2571*** 0.205*** -0.01962* -0.0017 -0.01955* -0.00157
(0.01113) (0.03105) (0.01139) (0.01912) (0.01139) (0.01912)
Crisis*FD
-0.00288*** -0.00213*** -0.0003*** -0.00048*** -0.0003*** -0.00048**
(0.00009) (0.00027) (0.00008) (0.00015) (0.00008) (0.00015)
Country Random NO YES NO YES NO YES
the industry median value of dependence on external nancing according to Rajan
and Zingales (1998). In addition, I also use log(asset) as a measure of rm size
instead of rm size dummy variable for a further robustness check.
The Compustat North America is used to compute the external nancing needs
of U.S. companies from 1990 to 2005, dening industry at NAICS 3-digits. After
including this industry level dependence on external nancing into the analysis, the
results follow in Table 2.11.
Even after controlling for industry needs of external nancing, the results con-
rm that nancial development promotes higher rm-level earnings while simulta-
neously worsening the impact of the credit crisis. As more variables are added to
32
Table 2.12: Results based on De
ated Earnings
(1) (2) (3) (4) (5) (6)
Intercept
11.2169*** 11.4385*** -1.734*** -1.7998*** -1.7369*** -1.8034***
(0.03288) (0.59990) (0.04423) (0.15310) (0.04428) (0.15310)
FD
0.0191*** 0.01572*** 0.001991*** 0.003132*** 0.002014*** 0.003156***
(0.00025) (0.00120) (0.00013) (0.00067) (0.00013) (0.00067)
DEF
0.0333 0.001198 -0.01877 -0.02086*
(0.02501) (0.01652) (0.01467) (0.01051)
FD*DEF
0.000229 0.000344** -0.00009** -0.00006** 0.000023 0.000072
(0.00016) (0.00012) (0.00004) (0.00002) (0.00009) (0.00007)
log(Size)
0.9232*** 0.9267*** 0.9231*** 0.9267***
(0.00266) (0.00195) (0.00266) (0.00195)
Age
-0.00202*** -0.00235*** -0.00202*** -0.00235***
(0.00029) (0.00019) (0.00029) (0.00019)
Crisis
0.3692*** 0.3239*** -0.04343** -0.01662 -0.04326** -0.01636
(0.01261) (0.03747) (0.01328) (0.02334) (0.01328) (0.02334)
Crisis*FD
-0.00399*** -0.00333*** -0.00014 -0.00033* -0.00014 -0.00034*
(0.00010) (0.00033) (0.00009) (0.00019) (0.00009) (0.00019)
Country Random NO YES NO YES NO YES
the analysis, the coecient of the crisis dummy is negative { the one notable dif-
ference from the previous results { and the interaction of the nancial development
is negative as well.
Now, I de
ate earnings and examine if this will change the ndings from the
previous results. The results in Table 2.12 rearm the nding from the previous
specications, which is that across all specications, nancial development has a
positive impact on earnings. In addition, as before, nancial development amplies
the shock from the recent crisis. After de
ating earnings, the coecient of the
crisis dummy variable has a negative sign more often than before. This captures
the negative eect of the recent credit crisis.
33
After accounting for a rm's liquidity condition and dependence on external
nancing at industry level, and de
ating earnings using the CPI, the results tell us
that nancial development has a negative impact on rm earnings during the crisis
period.
Robustness Check using Alternative Samples
The observations included in the previous analysis are limited to those rms that
reported their information from 2006 to 2008, consecutively. Therefore, one might
question if the previous results are driven by the sample selection. To resolve this
issue, I allow the whole data sample in the analysis. Except for those that do
not have earnings information, all rms in the data are now included. Specically,
290,127 observations were used for the previous data; now 673,831 observations are
used. The results in Table 2.13 are consistent with previous ndings. Compared
to Table 2.11, as the crisis random eects is considered, some of the coecients
become insignicant and inconsistent with the previous results. The coecients of
size and age are consistently positive and statistically signicant, which conrms
the results from Table 2.6 and 2.10. Again, the crisis dummy appears negative but
is insignicant. When the country random eects model is considered, the crisis
dummy coecient becomes positive and signicant, implying that the crisis itself
does not contribute to a decline in a rm's earnings. However, the interaction of
nancial development and the crisis dummy is negative and statistically signicant
under a random eects, which means that there is a signicant dierence before and
after the crisis. Therefore, the results hold across dierent samples and show that
34
when country random eects is assumed, more reliable results ensue. The nding,
which is that the crisis impact is propagated through nancial development, is
robust.
2.4.2 Panel Estimation
In the previous section, the data is analyzed based on a multilevel approach assum-
ing country and crisis random eects. I nd that the impact of the crisis can be
magnied by nancial development. In this section, I evaluate the ndings from the
previous method - multilevel regression - using panel data analysis. This approach
diers from the previous one, as we now consider rm-level random and xed ef-
fects. The main purpose of this section is to determine whether the ndings from
the multilevel regression still hold when we assume individual rm random or xed
eects. In addition, it is noted that there is an endogeneity issue in the previous
specications; to overcome this limitation, instrumental variable (IV) is introduced.
In the previous section, country and crisis random eects are assumed but here an
individual specic random eects is applied. Firms that have values for their size,
age, earnings for three years are included, making the dataset a strongly balanced
panel of 159,846 observations. As in the previous section, the same models are esti-
mated using the random eects model, and a panel IV is introduced to resolve the
possible endogeneity issue. As an instrumental variable for nancial development,
the strength of legal rights index (SLRI) measures the degree to which collateral
and bankruptcy laws protect the rights of borrowers and lenders, and thus facilitate
35
Table 2.13: Results based on Firm Earnings
Dependent Variable: logEBIT
(1) (2) (3) (4) (5) (6)
Intercept
13.3913*** 12.6677*** 13.2365*** 12.4849*** 13.4493*** 12.6588***
(0.007671) (0.461900) (0.007626) (0.472000) (0.011770) (0.471200)
[0.304900] [0.470800] [0.318200] [0.487200] [0.308600] [0.494900]
IDEF
-0.9759*** 0.5104*** -1.1317*** 0.4452*** -1.0429*** 0.4499***
(0.015050) (0.017610) (0.014780) (0.017080) (0.014940) (0.017070)
[0.528500] [0.236300] [0.526500] [0.207500] [0.515900] [0.201500]
FD*IDEF
0.01117*** -0.00058*** 0.01179*** -0.00077*** 0.01135*** -0.00069***
(0.000105) (0.000131) (0.000103) (0.000127) (0.000104) (0.000127)
[0.003878] [0.001477] [0.003838] [0.001368] [0.003784] [0.001374]
Size
1.191*** 1.3675*** 1.1929*** 1.3707***
(0.010710) (0.009965) (0.010710) (0.009978)
[0.189600] [0.210900] [0.188900] [0.212100]
Age
0.01912*** 0.01274*** 0.01748*** 0.008975*** 0.01839*** 0.009331***
(0.000246) (0.000232) (0.000241) (0.000226) (0.000242) (0.000228)
[0.003665] [0.002689] [0.004122] [0.002528] [0.004079] [0.002782]
Crisis
-0.2987 0.4245*** -0.5367 0.3959*** -0.4897 0.3949***
(0.727800) (0.030040) (0.737100) (0.029410) (0.730600) (0.029380)
[0.461200] [0.108700] [0.447700] [0.107200] [0.441800] [0.107700]
Crisis*Age
-0.01094*** -0.00428*** -0.01321*** -0.0043*** -0.01289*** -0.00435***
(0.000623) (0.000554) (0.000613) (0.000538) (0.000612) (0.000537)
[0.003568] [0.001782] [0.003959] [0.001567] [0.003845] [0.001617]
Crisis*FD
0.005942 -0.00108*** 0.006948 -0.00115*** 0.006613 -0.00115***
(0.005354) (0.000186) (0.005422) (0.000180) (0.005374) (0.000180)
[0.003020] [0.000569] [0.002897] [0.000511] [0.002840] [0.000515]
Crisis*Size
0.2794*** 0.1049*** 0.2716*** 0.1069***
(0.028330) (0.024350) (0.028270) (0.024330)
[0.270600] [0.199900] [0.261500] [0.201800]
Manufacturing
-0.5569*** -0.06908***
(0.026140) (0.023890)
[0.386500] [0.294200]
Housing
-0.3661*** -0.251***
(0.011560) (0.010480)
[0.123900] [0.103300]
Services
-0.09911*** -0.1767***
(0.012870) (0.011680)
[0.132200] [0.127600]
Country Random NO YES NO YES NO YES
Crisis Random YES NO YES NO YES NO
36
lending. The index ranges from 0 to 10, with higher scores indicating that these
laws are better designed to expand access to credit. For nancial development, the
SLRI is used as an IV for interaction terms. For example, the interaction of the
crisis dummy and SLRI is used as an IV for the interaction of the crisis and nancial
development.
According to Beck et al. (2000) and Khan (2000), legal systems originate from
small groups of legal traditions which can account for cross-country dierences in
creditor rights as well as systems for enforcing debt contracts. Therefore, the SLRI
varies with legal origins and traditions because of the degree to which the SLRI
aects the level of nancial development.
A very general model for this allows the intercept and the slope for coecients
to vary over both individual rm and time. This can be written as follows:
y
it
=
it
+x
0
it
+
it
wherei denotes an individual in a cross section, andt equals time. One variation of
this general model assumes that the unobservable individual eects
i
is a random
variable and is therefore called a random eects model. This random eects model
can be written with additional assumptions:
logEBIT
it
= +
i
+
1
FD
t
+
2
IDEF
it
+
3
IDEFFD
t
+
it
:
37
where
i
[;
2
],
it
[0;
2
] both the random eects and the error term are
assumed to be independently, identically distributed. When only nancial develop-
ment and dependence on external nancing are considered, assuming xed eects,
the above equation is as follows:
logEBIT
it
=
i
+
1
FD
t
+
2
IDEF
it
+
3
IDEFFD
t
+
it
As in the multilevel regression model, the same explanatory variables are used in the
analysis. When only nancial development and dependency on external nancing
are considered, it can be written:
logEBIT
it
=
0
+
1
FD
t
+
2
IDEF
it
+
3
IDEFFD
t
+u
it
(2.8)
where i denotes a rm, and t year. But when a rm random eects is considered:
logEBIT
it
=
i
+
1
FD
t
+
2
IDEF
it
+
3
IDEFFD
t
+
it
: (2.9)
The results in Table 2.14 are consistent with the previous ndings. Financial de-
velopment has a positive eect on rm-level earnings. Firms that do not depend
on external nancing tend to have higher earnings, which conrms the previous
ndings. However, in contrast to the previous results, it appears that rms that do
not depend on external nancing benet more from nancial development. When
an instrumental variable is used, the interaction of nancial development and de-
pendence on external nancing becomes negative but statistically insignicant. It
is hard to identify the impact of the crisis itself from Table 2.14.
38
Table 2.14: Result based on Panel and Panel IV Estimation
Dependent Variable: logEBIT
(1) (2) (3) (4)
Intercept 12.316*** 12.437*** 11.724*** 12.211***
(216.66) (495.95) (148.03) (443.39)
FD 0.013*** 0.012*** 0.018*** 0.014***
(27.51) (72.09) (27.33) (74.53)
IDEF 0.040 0.202***
(1.36) (15.11)
FD*IDEF 0.001*** -0.0000621
(4.87) (-0.63)
Age 0.016*** 0.016*** 0.015*** 0.016***
(27.64) (29.27) (19.75) (28.29)
Size 0.601*** 0.610***
(36.86) (37.99)
Crisis -0.097*** -0.080*** 0.549*** 0.124***
(-9.11) (-10.09) (21.33) (9.59)
Crisis*FD -0.006*** -0.002***
(-25.74) (-28.18)
Crisis*Age -0.001*** -0.001*** 0.0001911 -0.0006313***
(-4.34) (-4.15) (0.84) (-2.70)
Crisis*Size -0.156*** -0.152***
(-13.97) (-13.82)
IV YES NO YES NO
In the rst and second columns, the crisis dummy is negative but in the third
and fourth columns it again becomes positive. However, the interaction of nancial
development and the crisis dummy has a negative sign, even if the panel IV approach
is used, and it is statistically signicant. Therefore, based on this approach, nancial
development exacerbates the impact of the crisis.
2.4.3 Summary of Findings
So far, it used dierent methods to investigate nancial development's relationship
to rm-level earnings, and its role during the recent credit crisis. Both multi-level
regression and panel estimations consistently show that nancial development has
39
a positive eect on rm-level earnings during tranquil times. However, the results
also show that nancial development can magnify the impact of a nancial crisis,
as the negative coecient on the interaction of nancial development and the crisis
dummy variable across all specications indicate. This means that rms in more
nancially developed countries will undergo more severe drops in earnings during
times of crisis.
The results are robust to dierent instruments for dependence on external -
nancing and nancial development. In addition, after de
ating earnings using CPI
and using the alternative sample, the results remain the same.
2.4.4 Extension : Eect of Strength of Legal Rights
In the previous section I demonstrated that nancial development has a positive
eect on rm-level earnings. Further, I showed that nancial development serves
as a propagator of the recent credit crisis. Is nancial development the only factor
that aects rm-level outcomes? To answer the question, this section explores the
alternative factors that are related to rm-level earnings based on Djankov et al.
(2008) and Djankov et al. (2007).
This section includes the debt enforcement measure as an explanatory variable
instead of nancial development in determining rm-level earnings. The debt en-
forcement measure is correlated with debt market development, which is referred
to as nancial development in this paper. The debt enforcement measure is also
related to a given country's legal origins as well as its bankruptcy laws. Therefore,
40
we can include this variable in the analysis without worrying about an endogeneity
issue.
Djankov et al. (2008) constructed a measure of the eciency of debt enforce-
ment. They asked practitioners in each country to describe debt enforcement pro-
cedures, then computed the time and cost of each procedure. Based on those
results, they calculated the measure of eciency of the debt enforcement. Since
debt enforcement and nancial development are correlated { that is, debt enforce-
ment predicts debt market development { we do not include both measures in the
analysis.
As in the previous specications, we investigate whether a rm's size, age, crisis
and interactions of the crisis with other variables exhibit the same relationships
when considered with a measure of eciency, rather than a measure of nancial
development. Compared to Table 2.10, the results in Table 2.15 show that most of
the signs are the same and statistically signicant only when a xed eects model
is assumed. The positive coecient for rm size shows that large rms tend to
be more protable. Similarly, older rms' earnings are higher than younger ones',
conrming the previous ndings. The measure of debt enforcement (EFFSALE)
shows a positive relationship with rm-level earnings, since debt enforcement and
debt market development are positively correlated.
Now we run the same specication as in Table 2.11 and see whether debt en-
forcement leads to a decline in earnings during the recent credit crisis. The results
can be found in Table 2.16. As in Table 2.11, the crisis dummy variable appears
41
Table 2.15: Results based on Firm Earnings
Dependent Variable: logEBIT
(1) (2) (3) (4)
Intercept
12.2022*** 11.4875*** 12.1829*** 11.4749***
(0.035250) (1.428500) (0.035300) (1.428300)
[0.030630] [1.415300] [0.030680] [1.420000]
EFFSALE
0.01541*** 0.02697 0.01544*** 0.02697
(0.000459) (0.020350) (0.000459) (0.020340)
[0.000451] [0.021360] [0.000451] [0.021370]
Size
0.6251*** 1.5514*** 0.6648*** 1.5609***
(0.015780) (0.014940) (0.016400) (0.017930)
[0.018370] [0.259500] [0.018720] [0.245100]
Age
0.02348*** 0.005679** 0.02388*** 0.006172**
(0.000599) (0.000317) (0.000604) (0.000381)
[0.000546] [0.002020] [0.000551] [0.001971]
Crisis
0.06154*** 0.07848*** 0.1142*** 0.1188**
(0.004670) (0.012350) (0.007448) (0.019790)
[0.004871] [0.018820] [0.008052] [0.042680]
Crisis*Age
-0.00134*** -0.00153**
(0.000242) (0.000649)
[0.000255] [0.000578]
Crisis*Size
-0.1028*** -0.02841
(0.011310) (0.029700)
[0.011900] [0.088620]
Country Random NO YES NO YES
42
positive and statistically signicant under a country random eects model. The in-
teraction of the debt enforcement and the crisis dummy is negative when a country
random eect is assumed.
The interaction of the debt enforcement and the crisis dummy variable appears
negative and statistically signicant in the fourth and the sixth columns, which indi-
cates that rms in a country with stronger debt enforcement experience a relatively
large decline in earnings compared to those in a country with less debt enforcement.
2.4.5 Extension : Financial Development and Growth of
Firm
So far, earnings has been the dependent variable. In the following I also examine
whether nancial development is benecial to the growth of a rm, and whether
nancial development exacerbates the growth of a rm during times of crisis, based
on the rm's sales growth.
The results in Table 2.17 show that nancial development has a positive eect
and is more benecial to rms that are dependent on external nancing, based on a
rm's sales growth. However, nancial development contributes to a decline in rm
growth during times of crises. Findings from the rm's earnings and sales growth
suggests that nancial development has positive eect on both the level of earnings
and the growth of the rm during calm times; however, rms in countries with
higher nancial development will experience drops in earnings and growth when
there is a crisis.
43
Table 2.16: Results based on Firm Earnings
Dependent Variable: logEBIT
(1) (2) (3) (4) (5) (6)
Intercept
13.2682*** 13.3081*** 13.036*** 13.052*** 13.1947*** 13.195***
(0.014950) (0.455500) (0.014800) (0.452300) (0.022910) (0.453700)
[0.661900] [0.456100] [0.696700] [0.469000] [0.738500] [0.453700]
IDEF
-0.2857*** 0.6931*** -0.6431*** 0.6218*** -0.5752*** 0.6173***
(0.031970) (0.036670) (0.031360) (0.035520) (0.031740) (0.035520)
[0.968100] [0.116500] [0.882700] [0.098710] [0.873300] [0.096410]
IDEF*EFFSALE
0.008965*** -0.0037*** 0.0127*** -0.00382*** 0.01193*** -0.00378***
(0.000404) (0.000486) (0.000395) (0.000470) (0.000399) (0.000470)
[0.013930] [0.001947] [0.012640] [0.001705] [0.012520] [0.001726]
Size
1.5747*** 1.5043*** 1.5865*** 1.5118***
(0.019960) (0.018030) (0.019990) (0.018040)
[0.422100] [0.245100] [0.411200] [0.240300]
Age
0.02453*** 0.01013*** 0.02193*** 0.006117*** 0.02229*** 0.005852***
(0.000442) (0.000392) (0.000430) (0.000381) (0.000433) (0.000384)
[0.011290] [0.002703] [0.011140] [0.002040] [0.010950] [0.002464]
Crisis
-1.1347 0.2175*** -1.4568 0.2073*** -1.4043 0.2025***
(1.573600) (0.040110) (1.542600) (0.040000) (1.540800) (0.039980)
[1.374200] [0.032050] [1.359700] [0.052910] [1.349200] [0.053020]
Age*Crisis
-0.01619*** -0.00166** -0.01775*** -0.00162** -0.01744*** -0.00162**
(0.000796) (0.000677) (0.000777) (0.000656) (0.000776) (0.000656)
[0.010440] [0.000285] [0.010650] [0.000244] [0.010550] [0.000237]
EFFSALE*Crisis
0.02097 -0.00071 0.0252 -0.00099* 0.0245 -0.00093*
(0.022190) (0.000524) (0.021750) (0.000510) (0.021730) (0.000510)
[0.017630] [0.000358] [0.017430] [0.000516] [0.017270] [0.000516]
Size*Crisis
-0.02231 0.04981* -0.03607 0.05121*
(0.035940) (0.030040) (0.035950) (0.030020)
[0.358300] [0.092430] [0.333500] [0.093020]
Manufacturing
-0.5129*** 0.0646*
(0.041370) (0.036340)
[0.284400] [0.311000]
Housing
-0.2317*** -0.1483***
(0.021170) (0.018340)
[0.132500] [0.072500]
Services
-0.1075*** -0.2113***
(0.022750) (0.019670)
[0.180300] [0.150000]
Random Intercept NO YES NO YES NO YES
Crisis Random YES NO YES NO YES NO
44
Table 2.17: Results based on Sales Growth
Dependent Variable: logEBIT
Dep. Variable (1) (2) (3) (4)
Intercept
-33639*** -32355*** -33917*** -33112.00
(5582.26) (6366.17) (5612.99) (6458.07)
FD
96.06** 122.69*** 95.06** 123.52
(38.86) (46.88) (38.91) (46.86)
IDEF
11675*** 12632*** 11562*** 12499.00
(4165.14) (4185.64) (4171.97) (4190.01)
FD*IDEF
-110.15** -119.16*** -109.23** -118.18
(45.80) (46.02) (45.84) (46.04)
Size
1339.56*** 1046.86*** 1363.98*** 1090.08
(319.36) (336.29) (323.47) (341.97)
p
Age
104.70 256.21 98.58 259.94
(364.65) (377.30) (364.88) (377.33)
Crisis
8852* 9229.55* 9491.22* 10182.00
(4761.14) (4785.58) (4947.91) (4990.97)
Crisis*FD
-80.99 -86.91* -84.66* -92.74
(50.29) (50.49) (50.88) (51.23)
Crisis*Size
-1202.34 -1748.95
(2533.00) (2601.89)
Random Intercept NO YES NO YES
45
The ndings suggest that nancial development indeed matters in explaining
the levels of rm earnings and growth. This again means that nancial develop-
ment promotes higher rm-level earnings; at the same time, it also accelerates rm
growth. This adds empirical evidence to the existing literature at the rm-level,
and shows how nancial development is positively related to the growth of a rm,
and how the magnitude of its impact varies depending on a rm's dependence on
external nancing.
2.5 Conclusion
In this paper, I show that nancial development is a propagator of the recent credit
crisis, and that nancial development fails to play a role as a safeguard during crises.
Therefore, nancial development, while promoting higher earnings at the rm-level
during calm times, can be a curse when an economy is hit by a sudden negative
shock.
Based on European rm-level data, this paper provides empirical evidence of
the eect of nancial development on rm-level earnings, and demonstrates how
a shock can spread among rms. In countries with greater nancial development,
this negative propagation is even more severe.
In recent years, many developed and developing economies have experienced
rapid changes in their nancial markets. Most of these changes have been driven
by increasing complexities in nancial instruments (such as derivatives). Although
there are diering opinions among economists and policy makers, many theories
46
focus on the positive merits of nancial development, which can boost an economy,
instead of understanding the potential risks of nancial development. In this con-
text, the ndings in this paper suggest the possible risks that nancial development
can bring to more advanced economies such as those in Europe and perhaps to
non-European countries reaching advanced status as Brazil, Korea and Taiwan.
47
Chapter 3 Do Foreign Bank Aliates
Cut Their Lending More than the
Domestic Banks In a Financial Crisis: A
Quasi-Natural Experiment with Global
Individual Bank Data
3.1 Introduction
As part of liberalizing their international capital markets, countries open up their
banking sector, by allowing foreign banks greater entry and to perform a greater
variety of banking activities. As with other forms of capital liberalization, the
presence of foreign banks may bring in more international capital, raising the re-
cipient country's level of physical investment and growth. The presence of foreign
banks can also lead to supervisory, regulatory, and institutional improvements in
the country (Mishkin, 2009). International organizations such as the International
Monetary Fund, the World Bank, and the Bank of International Settlements have
traditionally pushed for non-discriminatory treatment of foreign banks in domestic
nancial markets.
48
The nancial crisis of 2008, however, has called into question, the benecial
eects of allowing foreign bank entry. When the tightening up of global liquidity
aected the world economy after the Lehman collapse in September 2008, banks
cut their international exposure, especially to emerging market countries. Banks
faced regulatory and internal provisioning requirements that made them rebalance
their portfolios by selling their foreign assets. From 2007 to 2008, international bank
loans fell by 80 percent, from 500 billion to 100 billion. This decline in international
bank loans helped fuel the very sharp decline in output abroad in 2009 and after,
especially in emerging market countries (Didier et al., 2011). Didier et al. (2011)
attribute the decline in output in emerging markets to a reinforcing pattern in which
a decline in foreign banks loans triggered collapses in domestic asset prices and in
the domestic nancial systems of emerging markets.
In this paper, we contribute to the literature on the international transmission
of balance sheet shocks that pummeled the banks all over the world in 2008 and
2009. We interpret the post-Lehman nancial crisis of 2008 and 2009 not only as a
shock to banks in the U.S., but also to banks worldwide. After the Lehman shock,
the crisis broadened and deepened globally, as evidenced by government policies to
support balance sheets in at least 39 countries (Levy and Schich, 2010). We trace
out the impact of these banks balance sheet shocks on multinational and domestic
banks. Given our identication strategy, we can clearly see which foreign banks
reduced their lending by more than the domestic or the other banks in a particular
region. Specically, we nd that European bank aliates in Latin America and
49
in Asia reduced their bank lending by more than the domestic banks operating in
those regions.
More broadly, we are concerned with whether the internal capital markets in the
multinational banking sector are an important transmission mechanism for nancial
shocks. In response to a liquidity shock, a foreign-owned bank may cut its lending to
a particular international market, over and above the cut in lending by the domestic
banks located in that market. Balanced against this desire to cut lending is the
potential loss of franchise value when a foreign bank is quick to cut its lending
to customers. This loss of locked-in customers and the harm done to bank-rm
relations could be quite large. We nd that for European bank aliates in Latin
America and Asia, their cut in lending was larger than the cut in lending of domestic
banks located in these regions.
Our identication strategy is novel and relies on the dierential responses of
foreign and domestic banks to the nancial crisis. For example, the balance sheets
of banks exposed to U.S. subprime mortgages may have been more aected than
the balance sheets of Korean domestic banks. If in a particular market, say in
Korea, a foreign-aliated bank's (say, Citibank, Korea) lending fell by more than a
domestic bank's (say, Kookmin's) lending then it is likely that the foreign bank is
responding to the tightening of its balance sheets at its headquarters. We control
for the decline in market conditions common to all banks in a particular region by
the decline in lending by the other banks. If foreign banks cut their lending by
50
more than the other banks, we attribute this decline in foreign bank lending to the
adverse balance sheet conditions facing the foreign bank at its headquarters.
We nd evidence that internal capital market do indeed aect cross-border lend-
ing. Some foreign banks adversely aected by the original liquidity shock to their
balance sheet cut their local lending in markets abroad. This cut in lending was par-
ticular pronounced for lending by European banks in the Asian and Latin American
markets. We show that compared to the other banks operating in these markets,
the U.S. banks did not cut back their lending.
In a set of important papers, Cetorelli and Goldberg (2012, 2010) also trace
out the implications of the U.S. originated banking crisis on markets abroad. The
authors are mainly concerned with U.S. banks, and examine U.S. bank level data
(Cetorelli and Goldberg, 2012). To the extent that the authors are concerned with
the behavior of non-U.S. banks, the authors mainly rely on country level lending
data, not bank level lending data.
Ours is one of rst papers to use bank level data that includes mostly non-U.S.
banks to show how nancial shocks originating abroad can dierentially aect the
lending behavior of both foreign-aliated banks and domestic banks. The study of
The greater cut in foreign aliate lending may also re
ect the fact that foreign aliates care
less about their franchise value than the domestic banks. Given our identication strategy, when
a foreign aliate cuts its lending by more than the other banks, we cannot distinguish whether
this behavior is because the foreign aliate has a greater balance sheet shock, or that the foreign
aliate has the same balance sheet shock as the other banks, but simply cares less about its
franchise value than the other banks.
51
non-U.S. foreign-aliated banks such as those of European banks operating abroad
is important, since these banks may face dierent shocks than U.S.-banks. We
examine bank level data for 193 countries, both of the domestic banks located in
these countries and of the foreign aliates. In total, there are 249,037 banks in our
sample (See Table 1).
3.2 Bank Balance Sheets and Foreign Lending
When a bank is confronted with a shock to the liabilities side of its balance sheet,
one response would be to cut lending. A shock to bank liabilities, such as a bank
run that reduces the amount of deposits that require reserves, or a tightening of the
interbank market that makes it dicult for a bank to meet its reserve requirements
may { if the bank cannot replace its decient bank liabilities { result in a reduction in
bank lending activity. Those banks under stress with aliates abroad (a \global"
bank) may cut their cross-border lending, lowering the lending activity of their
aliates abroad.
The basic idea is that sources of funds for banks matter for the supply of bank
loans. The sources of funds for a domestic bank in a foreign market include local
deposits, other local sources such as the local interbank market, and cross-border
interbank borrowing. For a global bank operating in a foreign market, there is in
addition, an internal capital market, which includes funding from aliates located
elsewhere and from headquarters located in the home country.
52
In the international transmission of shocks, both cross-border interbank borrow-
ing and cross-border funding from aliates and from the bank holding company or
headquarters matter (Cetorelli and Goldberg, 2010). In a crisis, a foreign-owned
bank hit by a liquidity shock like that of 2008-2009 will reduce its international
lending. With regards to its foreign aliates, the foreign bank may reduce funding
to foreign aliates or actively start transferring funds from its aliates abroad to
sustain its head oce balance sheet.
Domestic banks may also nd that cross-border interbank borrowing has dried
up, and may cut lending, if it cannot substitute other sources for this foreign bor-
rowing. For example, say there is a domestic bank that engages only in lending
within Korea, but borrows from an American bank in the interbank market. Faced
with an initial adverse shock to its liabilities, if the American bank cuts it's lending
to the Korean bank, the U.S. bank's liability shock will be transferred to the Korean
bank, and the lending of the Korean bank may also decline.
The post-Lehman crisis in 2008 was global. Given our data and the fact that
post-2008, the liquidity crisis was global, we cannot identify the decline in say, a
U.S.-bank's branch lending abroad as responding only to the balance sheet deteri-
oration of its headquarters in the U.S. The U.S.-bank branch may be responding
non-U.S. and local conditions. We can control for the eect of non-U.S. and local
conditions by observing the decline in lending by the other banks in that market.
The excess decline in lending by the U.S. bank should then re
ect the tightening of
balance sheet conditions specic to the U.S. bank.
53
The question we are concerned with in this paper is whether in response to
a liquidity shock, a foreign-owned bank will cut its lending to a particular market
abroad, over and above the cut in lending by the other banks located in that market.
We nd evidence that some foreign banks were aected by the original liquidity
shock to their balance sheet, and cut local lending in markets abroad. This cut in
lending was particular pronounced for lending by European bank aliates in the
Asian and Latin American markets.
3.3 Empirical Identication Methodology
Our identication methodology is inspired by the work of Peek and Rosengren
(2000), Khwaja and Mian (2008) and Cetorelli and Goldberg (2012, 2010). Specif-
ically, we apply a \dierence-in-dierence" methodology to examine the impact of
foreign ownership of a bank on domestic lending during times of crisis.
We use bank level data to estimate the following base level specication:
Loans
Assets
= +crisis +foreignowned +
crisisforeignowned (3.1)
+controls +
where
Loans
Assets
is the change in loans divided by the assets of a bank at time t 1,
\crisis" is a dummy variable that takes on a value of 1 in the post-crisis years, in the
years 2008 and 2009, \foreign-owned" is a dummy variables that takes on a value of
54
one when a bank is an aliate of a foreign bank (U.S., European, or non-U.S., non-
European), and \controls" are various bank and country level control variables. As
described in the next Section, our data allows us to distinguish among the ownership
types of the banks located in a given country. The ownership types are \domestic",
and \foreign", and among the foreign-owned banks, we distinguish between those
that are aliated with 1) U.S. banks, 2) European banks, and 3) non-European,
non-U.S. banks.
y
We identify the impact of foreign (U.S., European, non-U.S., non-European)
ownership as follows. Suppose that the occurrence of the 2008 Lehman \crisis" is
exogenous to a region, which is reasonable if the region is located outside of the
U.S. (but not in the U.S. where the subprime mortgage crisis originated).
Then,
if < 0, then on average, banks cut their lending during the \crisis" years,
compared to the years before.
if +
< 0, foreign-owned banks cut their lending during the \crisis" years,
compared to the years before.
if
< 0, foreign-owned banks cut their lending by more than the domestic-owned
banks during the \crisis" years, compared to the years before.
Thus, we can interpret as the impact the crisis had on decreasing total bank
lending in the country, and +
as the impact the crisis had on foreign bank
y
A bank aliate is dened as \U.S. owned" or \European owned" if the owner of the bank
with the highest percentage ownership is a U.S. bank of a European bank.
55
lending in the country.
re
ects the additional cut in lending, solely owing to the
fact that the bank's nationality was foreign.
3.4 Description of Data
In the analysis of this paper, we use the Bankscope database. Bankscope is a
database containing nancial information on over 28,000 banks worldwide. It pro-
vides detailed information on European banks, North American banks, and other
major banks throughout the world; it also has bank ownership information. We
classify the banks in any particular country as U.S., Europe-owned, or Non-U.S. -
Non-European (say, Japanese or Brazilian) owned, and domestically owned. For
example, if the bank is located in Korea and the ultimate owner country of the bank
is Korea, it is classied as a domestically owned bank. If the Korean owned bank
is located in Singapore, then the bank is classied as a Non-U.S, Non-European
(foreign) bank in Singapore. Note that when we are suing the data of a particular
country, that country's bank is always classied as a domestic bank. For example,
a Brazilian bank operating in Argentina is classied as foreign, but the same bank
is classied as domestic in Brazil.
To dene ownership, we utilize the ultimate ownership database of the Bvd
Bankscope data. The Bankscope database denes a bank's ultimate owner country
by the shareholders with the highest direct or total percentage of ownership. For
example, the ultimate owner of Citibank Korea Inc. is the U.S. bank, Citicorp.
56
Based on a given bank's physical location, we divide our data on banks into ve
regions { Europe
z
, Latin America
x
, Asia
{
, Africa
k
, and the U.S. and Canada.
Table 3.1 describes the data used for the empirical analyses. We dene 1999 and
2000 as the pre-recession period and 2001 and 2002 as the post-recession period.
We dene from 2003 to 2007 as the after-recession and pre-crisis periods. 2008 and
2009 we dene as the post-Lehman crisis period. As the number of bank show, this
z
France, Ukraine, Spain, Sweden, Norway, Germany, Finland, Poland, Italy, United Kingdom,
Romania, Belarus, Greece, Bulgaria, Iceland, Hungary, Portugal, Serbia, Austria, Czech Republic,
Ireland
x
Brazil, Colombia, Mexico, Argentina, Venezuela, Peru, Chile, Guatemala, Ecuador, Cuba,
Haiti, Bolivia, Dominican Republic, Honduras, Paraguay, El Salvador, Nicaragua, Costa Rica,
Puerto Rico, Panama, Uruguay, Guadeloupe, Martinique, French Guiana
{
Afghanistan, Bahrain, Bangladesh, Bhutan, Brunei, Cambodia, China, East Timor, India, In-
donesia, Iran, Iraq, Israel, Japan, Jordan, Kazakhstan, Korea Rep. Of, Kuwait, Kyrgyzstan, Laos,
Lebanon, Malaysia, Maldives, Mongolia, Myanmar, Burma, Nepal, Oman, Pakistan, The Philip-
pines, Qatar, Russia, Saudi Arabia, Singapore, Sri Lanka, Syria, Taiwan, Tajikistan, Thailand,
Turkey, Turkmenistan, United Arab Emirates, Uzbekistan, Vietnam, Yemen
k
Algeria, Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Canary Islands, Cape
Verde, Central African Republic, Ceuta, Chad, Comoros, Cote d'Ivoire, Congo, Congo, Demo-
cratic, Djibouti, Egypt, Equatorial Guinea, Eritrea, Ethiopia, Gabon, Gambia, Ghana, Guinea,
Guinea-Bissau, Kenya, Lesotho, Liberia, Libya, Madagascar, Madeira, Malawi, Mali, Mauritania,
Mauritius, Mayotte, Melilla, Morocco, Mozambique, Namibia, Niger, Nigeria, Reunion, Rwanda,
Saint Helena, Sao Tome and Principe, Senegal, Seychelles, Sierra Leone, Somalia, South Africa,
Sudan, Swaziland, Tanzania, Togo, Tunisia, Uganda, Western Sahara, Zambia, Zimbabwe
57
Table 3.1: Description of Sample
Pre-Recession Recession After Recession and Pre-Crisis Crisis
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Number of Countries 186 188 187 185 188 189 191 192 193 192 191
Total Number of Banks 21,030 21,728 21,984 22,804 23,074 23,537 24,795 22,915 22,868 22,511 21,791
{ Number of U.S. Banks 792 845 880 905 928 942 962 940 927 927 908
{ Number of European Banks 1,258 1,367 1,459 1,557 1,627 1,825 1,879 1,882 1,909 1,925 1,871
{ Number of U.S. Banks with Foreign Aliates 144 169 179 188 194 200 201 209 215 219 208
{ Number of European Banks with Foreign Aliates 465 506 533 560 583 642 678 685 714 716 681
{ Foreign
+
Banks with Foreign Aliates 740 812 888 931 972 1,053 1,097 1,109 1,139 1,160 1,137
Total Loans/Assets 0.10 0.07 0.15 0.11 0.27 1.63 3.01 0.16 0.28 0.19
{U.S. 0.12 0.10 0.19 0.07 0.25 0.10 0.17 0.10 0.10 0.04
{U.S. and Canada 0.12 0.10 0.19 0.07 0.25 0.10 0.17 0.11 0.10 0.04
{Europe 0.11 0.03 0.18 0.19 0.20 7.83 12.32 0.20 0.72 0.17
{Latin America 0.07 0.06 -0.07 0.07 1.44 0.60 0.19 0.26 0.09 0.53
{Asia 0.06 -0.03 0.04 0.12 0.12 0.15 0.11 0.11 0.13 1.07
{Africa 0.04 0.10 0.17 0.26 0.11 0.11 0.20 0.25 0.10 0.09
{U.S. Banks 0.15 0.12 0.11 0.11 2.10 0.10 0.92 0.20 0.08 0.13
{European Banks 0.06 0.06 0.15 0.21 1.06 18.83 37.77 0.35 0.14 0.28
{U.S. Banks with Foreign Aliates 0.07 0.06 0.10 0.18 0.15 0.06 0.15 0.20 0.02 -0.03
{European Banks with Foreign Aliates 0.08 0.06 0.07 0.16 2.17 0.16 0.23 0.29 0.09 0.05
{Foreign
+
Banks with Foreign Aliates 0.05 0.06 0.10 0.11 0.15 0.18 0.12 0.22 0.82 2.18
Tier 1 Capital/Assets 0.12 0.11 0.11 0.11 0.14 0.14 1.53 0.13 0.12 0.12 0.16
{ U.S. Banks 0.12 0.12 0.11 0.12 0.12 0.12 0.12 0.12 0.12 0.12 0.13
{ European Banks 0.38 0.09 0.30 0.09 0.10 0.09 0.24 0.23 0.09 0.09 0.10
{ U.S. Banks with Foreign Aliates 0.11 0.12 0.13 0.16 0.18 0.20 0.16 0.15 0.15 0.15 0.19
{ European Banks with Foreign Aliates 0.10 0.10 0.12 0.11 0.12 0.10 0.58 0.53 0.12 0.11 0.12
{ Foreign
+
Banks with Foreign Aliates 0.13 0.13 0.16 0.15 0.15 0.14 0.14 0.14 0.13 0.14 0.15
1)
Banks are dened as U.S. banks if the ultimate owner of a bank is the U.S.
2)
Banks are dened as European banks if the ultimate owner of a bank is a European country.
3)
U.S. banks that are located outside of U.S.
4)
European owned banks that are outside of Europe.
5)
Foreign
+
owned banks which are neither U.S. nor European banks.
6)
Total Loans/Asset is dened as
Loans
Assets
.
7)
The ratio of Tier 1 capital to assets.
database covers many of the banks in the world, and it has over 20,000 observations
per year.
Table 3.1 also summarizes the changes in loans divided by assets according to
the banks' physical location. For example, the third row summarizes those banks
58
which operate or are located in the U.S. Based on the whole sample, we can see that
the loan-to-asset ration decreased from 0.10 to 0.07 during the recent U.S. credit
crisis in 2007-2008. In contrast to the banks in the U.S., banks in Latin American
and Asian countries tended to increase their lending during the recent credit crisis.
Table 3.2: Lending by U.S. and European Banks
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
U.S. Banks/Total Lending
All 0.0530 0.0578 0.0615 0.0630 0.0585 0.0554 0.0685 0.0576 0.0540 0.0519 0.0485
North America 0.1943 0.1999 0.2071 0.2194 0.2194 0.2351 0.2582 0.2403 0.2360 0.2314 0.2467
Europe 0.0097 0.0096 0.0095 0.0103 0.0108 0.0097 0.0137 0.0130 0.0133 0.0150 0.0107
Latin America 0.0819 0.0874 0.0818 0.0877 0.0461 0.0771 0.0618 0.0541 0.0378 0.0442 0.0365
Asia 0.0055 0.0059 0.0050 0.0060 0.0060 0.0063 0.0063 0.0079 0.0091 0.0068 0.0067
Africa 0.0017 0.0017 0.0019 0.0019 0.0013 0.0007 0.0007 0.0010 0.0008 0.0013 0.0008
European Banks/Total Lending
All 0.1464 0.1469 0.1574 0.1731 0.1806 0.2129 0.1952 0.2189 0.2329 0.2218 0.2241
North America 0.0280 0.0280 0.0383 0.0363 0.0365 0.0485 0.0744 0.0636 0.0496 0.0541 0.0490
Europe 0.3843 0.3993 0.4121 0.4210 0.4233 0.4347 0.4168 0.4301 0.4584 0.4522 0.4693
Latin America 0.1936 0.2119 0.2134 0.1953 0.1998 0.2022 0.1838 0.1887 0.1794 0.1883 0.1663
Asia 0.0040 0.0036 0.0044 0.0073 0.0077 0.0081 0.0098 0.0168 0.0221 0.0195 0.0160
Africa 0.2935 0.3186 0.3024 0.3071 0.3541 0.3627 0.3603 0.3431 0.3338 0.2961 0.3096
Table 3.2 shows the ratio of lending by U.S. and European Banks in each region
in our sample. Overall, European banks in our sample account for 15 - 20% of total
lending and U.S. banks account for 5-8% of total lending. The European shares of
lending in Latin America and Africa in our sample are especially high, with 20%
and 30% of total lending in these countries.
3.5 Empirical Results
3.5.1 Main Results
Our baseline specications are based on the dependent variable dened as
Loans
i;j;t
Assets
i;j;t1
.
First, we only consider bank ownership as a determinant of bank lending. Later,
59
we add other bank-level characteristics as control variables. For the foreign bank
aliates, the bank level characteristics refer not to the headquarters' balance sheet
or on a consolidated basis, but only to balance sheet of the bank aliate in the
country.
We examine whether, during the 2008-2009 nancial crisis, the bank aliate's
country of ownership played a signicant role in determining bank lending patterns.
For each specication, we present the results including and excluding country xed-
eects (of where the bank or bank aliate is physically located).
As a preliminary analysis, the following equation is estimated:
Loans
Assets
=c +a
0
crisis +a
1
USBD +a
2
crisisUSBD (3.2)
whereUSBD denotes a dummy variable which takes on a value of one, if the bank
aliate is owned by the U.S. parent. If a bank's parent is located in the U.S., then
it is denoted as 1; otherwise it is 0.
Based on the equation (3.2), we examine how ownership aects lending during
the crisis. The Z test results (a
0
< 0) show that the crisis resulted in a decrease in
lending on average for all banks located in Asia and Africa at the 1% and the 5%
signicance level, respectively.
Next, we ask whether U.S. owned banks cut lending because of the crisis. If the
answer is yes, thena
2
< 0. The Z test results indicate that this is true for U.S.-bank
lending in the U.S. and Canada at the 5% signicance level. However, lending of
60
Table 3.3: Eects of Ownership on Bank Lending
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
1.0913 0.1093*** 1.1126 0.2307 0.8245 0.1632
(37.0949) (0.0220) (25.0537) (1.6752) (13.9288) (0.6002)
U.S. Owned Dummy
-0.5176 0.4591*** -7.6009 -0.1916 -1.7494 -0.1061
(2.4888) (0.0889) (14.8817) (0.8040) (5.5104) (0.2907)
Crisis Dummy
-0.9030 -0.0462 -2.8040 -0.0343 -0.0576*** -0.0609*
(0.7707) (0.0519) (3.4467) (0.3523) (0.0158) (0.0364)
(U.S. Owned Dummy*Crisis)
0.4198 -0.3896* 2.6607 -0.1125 -0.0650 -0.0670
(3.7949) (0.1998) (27.9386) (1.8313) (0.1567) (0.3449)
R-squared 0.0002 0.0003 0.0002 0.0037 0.0010 0.0139
Without Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.96** 0.109*** 0.1094*** 3.0366* 0.3305** 0.8119 0.1876***
(0.4701) (0.0221) (0.0220) (1.7294) (0.1508) (0.6572) (0.0269)
U.S. Owned Dummy
-0.4591 0.4623*** 0.4591*** -2.9181 -0.1453 -0.7209 -0.1339
(2.4595) (0.0895) (0.0889) (14.7101) (0.7993) (5.4576) (0.2862)
Crisis Dummy
-0.8736 -0.0453 -0.0461 -2.8743 -0.0121 -0.0575*** -0.0748**
(0.7659) (0.0523) (0.0519) (3.4427) (0.3514) (0.0158) (0.0361)
(U.S. Owned Dummy*Crisis)
0.4485 -0.3929* -0.3896* 2.7107 -0.1432 -0.0650 -0.0573
(3.7935) (0.2009) (0.1998) (27.9305) (1.8311) (0.1567) (0.3445)
R-squared 0.0000 0.0003 0.0003 0.0000 0.0000 0.0000 0.0007
Number of Observations 206,335 104,100 104,794 45,343 8,476 19,612 5,601
Since we use country dummy to estimate country xed-eects, for a single country (U.S.), we do not report the result
using country xed eects.
U.S. banks in the U.S. economy are not orthogonal to U.S. originated shocks, so
these results are not well-identied.
Table 3.4 divides ownership into : European-owned and non-European owned
banks. The interaction term between the European-owned bank dummy and the
crisis dummy is negative, and it is statistically signicant in Latin America. Eu-
ropean banks appear to cut their lending more than the other banks in the Latin
61
America market. These results are robust to the inclusion of country (of where the
banks are located) xed-eects.
Table 3.4: Eects of Ownership on Bank Lending with Europe Owned Dummy
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.5871 0.1374*** -2.6850 -0.0095 0.8239 0.2909
(37.0886) (0.0214) (23.1388) (2.3258) (13.9275) (0.8628)
Europe Owned
9.3224*** -0.0002 12.762*** 1.7878*** -1.8218 -0.0164
(1.9378) (0.3329) (4.4571) (0.5313) (3.8079) (0.0817)
Crisis Dummy
-0.0085 -0.0691 0.0592 0.1391 -0.0565*** -0.0567
(0.7845) (0.0502) (3.8629) (0.3635) (0.0159) (0.0393)
(Europe Owned Dummy*Crisis Dummy)
-12.0996*** -0.0752 -14.1116* -1.9654* -0.0743 -0.0333
(2.8565) (0.7034) (8.3154) (1.1679) (0.1038) (0.0999)
R-squared 0.0004 0.0000 0.0004 0.005 0.0010 0.0139
Without Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.2389 0.1374*** 0.1375*** 0.4315 0.1786 0.8199 0.1893
(0.4791) (0.0215) (0.0214) (1.9205) (0.1549) (0.6632) (0.0288)
Europe Owned Dummy
9.9124*** -0.0138 0.0006 12.7766*** 1.6767*** -0.5587 -0.0225
(1.7768) (0.3504) (0.3322) (4.2838) (0.5237) (3.6808) (0.0785)
Crisis Dummy
-0.0066 -0.0686 -0.0691 0.0430 0.1557 -0.0564 -0.0726
(0.7799) (0.0506) (0.0502) (3.8580) (0.3626) (0.0159) (0.0390)
(Europe Owned Dummy*Crisis Dummy)
-12.1705*** -0.0574 -0.0751 -14.2537* -1.947* -0.0742 -0.0193
(2.8534) (0.7491) (0.7034) (8.3098) (1.1682) (0.1038) (0.0996)
R-squared 0.0001 0.0000 0.0000 0.0002 0.0012 0.0000 0.0007
Number of Observations 206,335 104,100 104,794 45,343 8,476 19,612 5,601
In the following specications, we classify ownership into three dierent cate-
gories: U.S., Europe, with the other banks as the baseline (including the domestic
banks).
Loans
Assets
= c +a
0
crisis +a
1
USBD +a
2
crisisUSBD (3.3)
+a
3
EBD +a
4
crisisEBD
62
where EBD denotes whether the bank's ultimate owner is in Europe. In this
specication, if the crisis cut lending more for European-owned banks than for
U.S.-owned banks, then a
4
< a
2
< 0. The Z test results (a
4
< a
2
< 0), however,
are not signicant in any sample, suggesting that European-owned banks did not
cut their lending compared to U.S.-owned banks. However, as before European
banks cut their lending in Latin America more than the other banks (excluding the
U.S.-owned banks) located in Latin America.
Table 3.5 shows the results when we add bank control variables 1) the ratio
of Tier 1 to total assets at time (t 1) and 2) the log of assets. These bank-level
characteristics are at the bank or bank-aliate level. Most bank level characteristics
appear to be signicant in a majority of the samples, except in Europe. Specically,
the more capitalized and larger the bank, the greater the lending. Compared to
Table 3.3 and Table 3.5, when we include bank-level variables, the coecient on the
(Crisis Dummy)*(Europe-owned Dummy) becomes signicantly negative in Latin
America and Asia.
The interaction of the Europe-owned bank dummy and the crisis dummy turns
negative and signicant not only in Latin America but also in Asia (the interaction
term appears signicant only in Europe and in Latin America in Table 3.5). These
results again suggest that banks owned by European countries cut their lending
by more than the other banks in Latin America and Asia during the crisis. U.S.
banks, however, do not appear to cut their lending any more than the other banks
(excluding the European-owned banks).
63
Table 3.5: Eects of Ownership on Bank Lending including Bank-Level Character-
istics
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
-5.223 -1.983*** -0.3408 -0.2909 -0.9816 0.075
(6.9392) (0.2318) (1.6466) (0.8208) (26.0164) (0.3499)
Lag(Tier1/Assets)
0.0001 2.8715*** 0.0079 0.00004** 1.3515*** 0.0021
(0.0003) (0.1896) (0.0797) (0.00002) (0.0595) (0.0017)
Log(Assets)
0.3393*** 0.0934*** 0.0337 0.0227 0.0776*** 0.0022
(0.0413) (0.0119) (0.0440) (0.0200) (0.0079) (0.0079)
U.S. Owned Dummy
-0.0058 0.3256*** -1.502 -0.0319 -3.5419 -0.1303
(0.5456) (0.0811) (1.5563) (0.2529) (12.4590) (0.0964)
Europe Owned Dummy
-(0.7273) -(0.3073) -(0.3713) (0.1712) -(5.4337) -(0.0192)
(0.5488) (0.3194) (0.2586) (0.1409) (9.0794) (0.0426)
Crisis Dummy
-0.2123*** -0.0708 -0.2921 -0.1076*** -0.0504*** -0.0645**
(0.0471) (0.0460) (0.2178) (0.0377) (0.0061) (0.0261)
(U.S. Owned Dummy*Crisis Dummy)
-0.5407*** -0.4121** 1.444 -0.055 -0.0538 0.0101
(0.1822) (0.1773) (2.1683) (0.2335) (0.0422) (0.1476)
(Europe Owned Dummy*Crisis Dummy)
-0.0824 -0.1652 0.5039 -0.2889*** -0.0727** -0.0234
(0.2587) (0.7116) (0.4295) (0.0993) (0.0322) (0.0592)
R-squared 0.0012 0.0027 0.0131 0.2159 0.0014 0.0446
Without Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
-5.4963*** -1.9631*** -1.9318*** -0.2856 -0.6091 -0.1057 0.1691
(0.7269) (0.2311) (0.2282) (0.7533) (0.5027) (1.6314) (0.1072)
Lag(Tier1/Assets)
0.0001 2.8742*** 2.859*** 0.0026 0.00004** 1.3514*** 0.0019
(0.0003) (0.1901) (0.1894) (0.0800) (0.00002) (0.0595) (0.0017)
Log(Assets)
0.2932*** 0.0923*** 0.0907*** 0.0285 0.0429* 0.0775*** -0.0007
(0.0365) (0.0118) (0.0117) (0.0357) (0.0236) (0.0079) (0.0053)
U.S. Owned Dummy
0.1737 0.3321*** 0.3306*** -0.3165 -0.1729 -1.7482 -0.126
(0.5396) (0.0814) (0.0810) (1.5279) (0.3111) (12.1327) (0.0947)
Crisis Dummy
-0.1967*** -0.0692 -0.0692 -0.2606 -0.132*** -0.0504*** -0.0422*
(0.0466) (0.0461) (0.0460) (0.2069) (0.0377) (0.0061) (0.0231)
(U.S. Owned Dummy*Crisis Dummy)
-0.5354*** -0.4164** -0.4128** 0.205 -0.0299 -0.0538 -0.0059
(0.1821) (0.1781) (0.1773) (2.1646) (0.2431) (0.0422) (0.1466)
(Europe Owned Dummy*Crisis Dummy)
-0.1693 -0.4918 -0.4784 0.2737 -0.2681*** -0.0727** -0.0547
(0.2512) (0.6687) (0.6397) (0.3638) (0.0945) (0.0322) (0.0442)
R-squared 0.0001 0.0028 0.0027 0.0003 0.0028 0.0001 0.0125
Number of Observations 125,643 102,047 102,399 11,129 755 6,457 861
64
Table 3.6: Eects of Ownership on Bank Lending based on 06-09 sample
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.2721 -1.5861*** -1.5522*** -7.91 -0.4409 -0.5674 0.182
(1.7453) (0.2420) (0.2383) (6.1307) (0.5950) (2.1324) (0.1576)
Lag(Tier1/Assets)
0.00004 2.2424*** 2.2298*** 3.6691*** 0.00004* 2.1802*** 0.3623***
(0.0002) (0.1856) (0.1849) (1.2266) (0.00002) (0.1004) (0.1317)
Log(Assets)
0.0019 0.0749*** 0.0731*** 0.369 0.0359 0.1178*** 0.0015
(0.0853) (0.0123) (0.0121) (0.2303) (0.0279) (0.0213) (0.0076)
U.S. Owned Dummy
0.4718 0.4522*** 0.45*** -0.8474 -0.1162 -2.2808 -0.2293
(2.1957) (0.1019) (0.1013) (33.0830) (0.4190) (15.3741) (0.1600)
Europe Owned Dummy
8.7773*** -0.2786 -0.2479 12.5998* 0.1246 -2.1834 -0.0845
(2.1437) (0.4018) (0.3813) (7.4702) (0.2040) (10.1868) (0.0713)
Crisis Dummy
-0.0813** -0.0395 -0.0398 -0.3229*** -0.136*** -0.0566*** -0.1493***
(0.0342) (0.0370) (0.0368) (0.0981) (0.0435) (0.0083) (0.0344)
(U.S. Owned Dummy*Crisis Dummy)
-0.5521*** -0.5064*** -0.5017*** 0.3436 -0.0906 -0.0716 0.0907
(0.1300) (0.1419) (0.1412) (0.9679) (0.3130) (0.0561) (0.2090)
(Europe Owned Dummy*Crisis Dummy)
-0.1754 -0.1298 -0.149 0.0724 -0.2905** -0.0558 0.0394
(0.1684) (0.5801) (0.5526) (0.1867) (0.1164) (0.0418) (0.0856)
Number of Observations 49,102 37,443 37,608 5,496 615 3,086 590
As a robustness check, Table 3.6 shows the results based only on a sample
from 2006 and 2009, using the same specication in Table 3.5 without country
xed-eects. The ndings from the previous Table remain the same, except that
the interaction of the European owned dummy and the Crisis dummy becomes
insignicant in Asia, though it is still negative.
In Tables 3.7 and 3.8, we also show when we include another ownership variable,
\Foreign-dummy," which takes on the value of one when the bank is owned by a non-
European, non-U.S. foreign country. For example, for Korea, the \Foreign-dummy"
takes on a value of one when the bank is owned by a foreign (non-Korean) country
such as Japan, but not U.S. or European owned. Thus, the \baseline" from which
changes are measured are only the domestic banks. We split Asia into emerging
Asia and Japan. We also interact the ownership variables with the \capital market
65
openness index" from Chinn and Ito (2006). Again, European aliated banks cut
loans more than the domestic banks in Latin America In addition, non-U.S., non-
European foreign aliated banks in Asia cut their lending more than the domestic
banks in Asia. Regardless of how we dene ownership, the capital market openness
index has a positive eect. The decline in crisis-induced bank lending in more open
countries is less.
3.5.2 Additional Robustness Checks
Here we perform two additional robustness checks. First, we will use an alterative
measure of changes in loans, lnLoan
t
lnLoan
t1
; and examine whether the results
remain the same.
The results in Table 3.9 show that all of the results in the previous Tables hold.
In fact, with Loan as the dependent variable, U.S. owned banks in Asia actually
cut their lending more during the crisis than the other banks. Consistent with the
previous ndings, European owned banks in Latin America cut their lending more
than the other banks.
We also re-examine the role of capital market openness. Similar to the previous
results, the results illustrate that capital market openness has a positive eect on
bank lending in Asia and Africa (with country xed-eects).
Our second robustness test is to run our specication using a dierent sample,
the data overlapping the post-2001 recession. The 2001 recession was brief and
damage to the balance sheets of banks were minimal. Our empirical framework tests
66
Table 3.7: Eects of Ownership on Bank Lending Interacted with Capital Market
Openness
With Country Fixed-Eects
ALL Europe Latin America Japan Asia
+
Africa
Constant
-2.8039 0.6856 -0.3362 -0.3914 -2.343
(41.8846) (32.1806) (0.8417) (37.2857) (1.6806)
Lag(Tier1/Assets)
0.00003 0.0065 0.00004** 1.5414*** 0.0011
(0.0004) (0.1154) (0.00002) (0.0815) (0.0020)
Log(Assets)
-0.0179 -0.1252 0.0248 0.0632*** 0.1399***
(0.1447) (0.3023) (0.0207) (0.0099) (0.0302)
U.S. Owned Dummy
-3.2778 -9.2717 -0.0448 -1.5362 -0.3024
(11.1071) (23.1919) (0.2614) (19.3712) (0.8240)
Europe Owned Dummy
5.8607 6.7655 0.1226 -1.8384 -0.0757
(3.5772) (5.2560) (0.1482) (12.3644) (0.2777)
Foreign
+
Owned Dummy
1.594 -15.1048 -0.0832 21.1148** -0.059
(5.4726) (12.8232) (0.1802) (8.4322) (0.2383)
Crisis Dummy
-0.3447** -0.7435** -0.111*** -0.1171*** -0.244***
(0.1494) (0.3225) (0.0398) (0.0107) (0.0432)
(U.S. Owned Dummy*Crisis Dummy)
0.35 0.8875 -0.0863 0.0352 0.1823
(1.5114) (4.8375) (0.3788) (0.0551) (0.2527)
(Europe Owned Dummy*Crisis Dummy)
0.0243 0.3645 -0.4222*** -0.018 0.1014
(0.6554) (1.4032) (0.1468) (0.0386) (0.0846)
(Foreign
+
Owned Dummy*Crisis Dummy)
0.2199 0.5661 0.0153 -0.1091*** 0.095
(0.6123) (1.9338) (0.1943) (0.0258) (0.0756)
(Openness*U.S. Owned Dummy*Crisis Dummy)
-0.0652 -0.1622 0.0341 -0.0695 -0.0243
(0.8313) (2.3606) (0.2874) (0.0558) (0.1097)
(Openness*Europe Owned* Dummy*Crisis Dummy)
0.042 0.0662 0.1935 0.0848** -0.0089
(0.2872) (0.5740) (0.1536) (0.0382) (0.0375)
(Openness*Foreign
+
Dummy*Crisis Dummy)
0.0091 0.0104 0.0217 0.0398** 0.0355
(0.3162) (0.8938) (0.1301) (0.0190) (0.0307)
R-squared 0.0025 0.0027 0.2156 0.0032 0.0050
Without Country Fixed-Eects
ALL Europe Latin America Japan Asia
+
Africa
Constant
0.0159 1.5265 -0.6064 -5.7425*** -1.3601 -1.6833***
(3.3183) (6.6081) (0.5250) (0.3794) (2.9679) (0.5405)
Lag(Tier1/Assets)
0.00003 0.0074 0.00004** 1.2731*** 1.5414*** 0.001
(0.0004) (0.1154) (0.00002) (0.0986) (0.0813) (0.0020)
Log(Assets)
0.0138 -0.0492 0.0434* 0.2427*** 0.0632*** 0.0914***
(0.1428) (0.2987) (0.0246) (0.0158) (0.0098) (0.0269)
U.S. Owned Dummy
-0.2637 -0.3999 -0.1924 -0.0914 -0.2111 -0.0846
(10.7307) (22.4958) (0.3233) (0.5125) (18.6495) (0.7781)
Europe Owned Dummy
7.1779** 9.1699* 0.0237 -0.0691 -0.0526
(3.3300) (5.0156) (0.1860) (11.7440) (0.2572)
Foreign
+
Owned Dummy
5.1757 -0.2621 -0.1676 20.1122*** 0.1273
(4.8816) (11.7480) (0.2229) (7.8054) (0.2104)
Crisis Dummy
-0.3543** -0.7564** -0.1339*** 0.0047 -0.1171 -0.2195***
(0.1492) (0.3224) (0.0404) (0.0076) (0.0107) (0.0430)
(U.S. Owned Dummy*Crisis Dummy)
0.3393 0.8864 -0.0808 0.0352 0.1512
(1.5110) (4.8371) (0.3980) (0.0550) (0.2540)
(Europe Owned Dummy*Crisis Dummy)
0.0059 0.3198 -0.4244*** -0.0181 0.0967
(0.6550) (1.4022) (0.1538) (0.0385) (0.0850)
(Foreign
+
Owned Dummy*Crisis Dummy)
0.204 0.5327 0.0637 -0.1091*** 0.0926
(0.6120) (1.9334) (0.1983) (0.0257) (0.0758)
(Openness*U.S. Owned Dummy*Crisis Dummy)
-0.0484 -0.1036 0.053 -0.0801** -0.0695 -0.0087
(0.8309) (2.3602) (0.3155) (0.0333) (0.0557) (0.1101)
(Openness*Europe Owned Dummy*Crisis Dummy)
0.0516 0.0847 0.2096 -0.0575 0.0848** 0.0022
(0.2870) (0.5736) (0.1715) (0.3284) (0.0381) (0.0374)
(Openness*Foreign
+
Owned Dummy*Crisis Dummy)
0.028 0.081 -0.0066 0.0398** 0.0396
(0.3159) (0.8933) (0.1360) (0.0189) (0.0305)
R-squared 0.0003 0.0003 0.0032 0.0012 0.0022 0.0004
Number of Observations 21,504 11,091 743 2,886 3,145 854
Foreign
+
Owned Dummy denotes banks that are neither owned by U.S. nor European countries.
Asia
+
denotes Asian countries excluding Japan.
67
Table 3.8: Eects of Ownership on Bank Lending Interacted with Capital Market
Openness
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Japan Asia
+
Africa
Constant
-9.6752 -2.0144*** 12.0897 -0.343 -0.4558 0.2448
(25.5808) (0.2334) (29.5079) (0.8409) (37.2723) (0.3558)
Lag(Tier1/Assets)
0.0001 2.8878*** 0.0065 0.00004** 1.5376*** 0.0018
(0.0003) (0.1902) (0.1154) (0.00002) (0.0814) (0.0017)
Log(Assets)
0.6137*** 0.0951*** -0.1268 0.0245 0.0653*** 0.0004
(0.0524) (0.0120) (0.3023) (0.0206) (0.0098) (0.0080)
U.S. Owned Dummy
-0.5427 0.3213*** -9.302 -0.0424 -1.5074 -0.169*
(1.8647) (0.0812) (23.1800) (0.2579) (19.3642) (0.1004)
Europe Owned Dummy
4.3405** -0.3184 6.7691 0.159 -1.8548 -0.0551
(1.8519) (0.3195) (5.2539) (0.1447) (12.3599) (0.0463)
Foreign
+
Owned
1.0544 -0.2245 -15.2298 -0.087 21.1158** -0.0986**
(2.6508) (0.2437) (12.8524) (0.1778) (8.4292) (0.0411)
Crisis Dummy
-0.4313** -0.7778 -0.1221** -0.1244*** -0.1193***
(0.1943) (0.9624) (0.0569) (0.0109) (0.0329)
(U.S. Owned Dummy*Crisis Dummy)
-0.5971*** -0.4134** 0.6579 -0.0538 0.0230 0.0282
(0.1851) (0.1774) (3.3319) (0.2356) (0.0539) (0.1504)
(Europe Owned Dummy*Crisis Dummy)
-0.1912 -0.1678 0.5125 -0.285*** -0.0032 0.0103
(0.2739) (0.7116) (0.6295) (0.1007) (0.0382) (0.0633)
(Foreign
+
Owned Dummy*Crisis Dummy)
-0.0802 -0.1466 0.598 0.0356 -0.0986*** 0.0955*
(0.3490) (0.5323) (1.6542) (0.1285) (0.0253) (0.0544)
(Openness*Crisis Dummy)
0.0471 -0.0285 0.0144 0.0133 0.0231*** 0.0278*
(0.0793) (0.0188) (0.3770) (0.0481) (0.0066) (0.0151)
R-squared 0.0022 0.0028 0.0027 0.2165 0.0032 0.0571
Without Country Fixed-Eect
ALL U.S. U.S. and Canada Europe Latin America Japan Asia
+
Africa
Constant
-11.7407*** -2.0394*** -2.0025*** 1.5661 -0.579 -5.7425*** -1.4048 0.2452
(1.0798) (0.2355) (0.2325) (6.6071) (0.5246) (0.3794) (2.9669) (0.1184)
Lag(Tier1/Assets)
0.0001 2.9035*** 2.886*** 0.0074 0.00004** 1.2731*** 1.5375*** 0.0017
(0.0003) (0.1909) (0.1901) (0.1154) (0.00002) (0.0986) (0.0812) (0.0017)
Log(Assets)
0.6125*** 0.0963*** 0.0944*** -0.0511 0.0421 0.2427*** 0.0653*** -0.0028
(0.0517) (0.0121) (0.0119) (0.2987) (0.0246) (0.0158) (0.0098) (0.0057)
U.S. Owned Dummy
-0.0985 0.3228*** 0.3219*** -0.4322 -0.1834 -0.0914 -0.1804 -0.1637*
(1.8523) (0.0816) (0.0812) (22.4837) (0.3187) (0.5125) (18.6433) (0.0990)
Europe Owned Dummy
5.0069*** -0.362 -0.3265 9.1601* 0.0712 -0.1413 -0.0808 -0.0541
(1.6481) (0.3303) (0.3192) (5.0118) (0.1814) (0.8065) (11.7401) (0.0448)
Foreign
+
Owned Dummy
3.5559 -0.2717 -0.2429 -0.2575 -0.1697 20.1147** -0.0727*
(2.3846) (0.2575) (0.2417) (11.7936) (0.2191) (7.8028) (0.0376)
Crisis Dummy
-0.446** -0.8477 -0.1327** -0.1244*** -0.0874***
(0.1936) (0.9602) (0.0554) (0.0109) (0.0313)
(U.S. Owned Dummy*Crisis Dummy)
-0.5956*** -0.4167** -0.4131** 0.7655 -0.0279 -0.1968** 0.0230 0.0433
(0.1850) (0.1781) (0.1774) (3.3314) (0.2453) (0.0818) (0.0539) (0.1488)
(Europe Owned Dummy*Crisis Dummy)
-0.1828 -0.1501 -0.1707 0.5133 -0.2829*** -0.0032 0.0107
(0.2738) (0.7425) (0.7116) (0.6295) (0.1019) (0.0381) (0.0621)
(Foreign
+
Owned Dummy*Crisis Dummy)
-0.0543 -0.1614 -0.1472 0.6772 0.0565 -0.0986*** 0.0956*
(0.3490) (0.5682) (0.5323) (1.6538) (0.1303) (0.0253) (0.0513)
(Openness*Crisis Dummy)
0.0533 -0.0286 -0.0285 0.0383 -0.0008 0.0019 0.0231*** -0.0023
(0.0790) (0.0189) (0.0188) (0.3761) (0.0448) (0.0031) (0.0066) (0.0103)
R-squared 0.0004 0.0028 0.0027 0.0003 0.0043 0.0012 0.0022 0.0178
Number of Observations 124,126 102,047 102,399 11,091 743 2,886 3,145 835
68
Table 3.9: Eects of Ownership on Bank Lending based on lnLoan
t
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.2368 -0.8326*** 0.0677 -0.4426 -2.0006*** -0.8292
(0.1613) (0.0211) (0.1204) (0.6472) (0.2502) (0.7493)
Lag(Tier1/Assets)
0.00002 3.1777*** 0.0135*** 0.00002 1.764*** 0.0021
(0.00002) (0.0168) (0.0036) (0.00002) (0.0832) (0.0031)
Log(Assets)
-0.0052*** 0.0328*** 0.0113*** 0.0357** 0.097*** 0.0506**
(0.0012) (0.0011) (0.0034) (0.0145) (0.0076) (0.0196)
U.S. Owned Dummy
0.012 0.0061 -0.2397** -0.1433 -0.1887** -0.3235
(0.0112) (0.0078) (0.0990) (0.1873) (0.0954) (0.2747)
Europe Owned Dummy
0.068*** -0.0132 0.0094 0.046 -0.0393 -0.1425
(0.0126) (0.0290) (0.0197) (0.1044) (0.0747) (0.1079)
Crisis Dummy
-0.0855*** -0.0756*** -0.1759*** -0.2013*** -0.0707*** -0.1861***
(0.0023) (0.0021) (0.0102) (0.0395) (0.0108) (0.0476)
(U.S. Owned Dummy*Crisis Dummy)
-0.0388*** -0.0459*** 0.0238 -0.0376 -0.1513* 0.0321
(0.0094) (0.0079) (0.1031) (0.2230) (0.0788) (0.2535)
(Europe Owned Dummy*Crisis Dummy)
-0.1359*** -0.1826*** -0.0313 -0.192* -0.1606*** 0.0251
(0.0122) (0.0325) (0.0200) (0.1037) (0.0597) (0.1095)
R-squared 0.0234 0.1952 0.0377 0.1138 0.0713 0.1191
Without Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.2359*** -0.8172*** -0.8253*** 0.1045* -0.6716** -1.0505*** -0.0396
(0.0207) (0.0209) (0.0211) (0.0626) (0.3348) (0.1364) (0.3021)
Lag(Tier1/Assets)
0.00002 3.1696*** 3.1778*** 0.0139*** 0.00002 1.5948*** 0.0019
(0.00002) (0.0167) (0.0168) (0.0036) (0.00002) (0.0808) (0.0031)
Log(Assets)
-0.0042*** 0.032*** 0.0324*** 0.0054* 0.0484*** 0.0454*** 0.0157
(0.0011) (0.0011) (0.0011) (0.0030) (0.0157) (0.0059) (0.0152)
U.S. Owned Dummy
0.0013 0.0121 0.0065 -0.222** -0.234 -0.1829* -0.1759
(0.0115) (0.0076) (0.0078) (0.0971) (0.2038) (0.1005) (0.2971)
Europe Owned Dummy
0.111*** -0.0411 -0.0192 0.0256 0.0294 -0.0137 -0.094
(0.0121) (0.0296) (0.0290) (0.0193) (0.1151) (0.0766) (0.1133)
Crisis Dummy
-0.0838*** -0.0748*** -0.0755*** -0.1782*** -0.2133*** -0.0484*** -0.126***
(0.0023) (0.0021) (0.0021) (0.0100) (0.0391) (0.0106) (0.0468)
(U.S. Owned Dummy*Crisis Dummy)
-0.0401*** -0.0472*** -0.0458*** 0.0083 0.0018 -0.1433* -0.021
(0.0094) (0.0079) (0.0079) (0.1027) (0.2304) (0.0796) (0.2563)
(Europe Owned Dummy*Crisis Dummy)
-0.1357*** -0.1572*** -0.1833*** -0.0286 -0.2064** -0.1446** 0.0057
(0.0122) (0.0339) (0.0325) (0.0200) (0.1043) (0.0603) (0.1113)
R-squared 0.0054 0.1982 0.1952 0.027 0.0228 0.0181 0.0031
Number of Observations 124,099 100,596 100,945 11,100 747 6,415 860
69
Table 3.10: Eects of Ownership on Bank Lending based on lnLoan
t
Interacted
with Capital Market Openness
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Japan Asia Africa
Constant
1.1366*** -0.8441*** -0.1923 -0.5330 -1.5062*** -0.2638
(0.1934) (0.0212) (0.1504) (0.5672) (0.3416) (0.7564)
Lag(Tier1/Assets)
0.00002 3.1791*** 0.0138*** 0.00002 1.7116*** 0.0022
(0.00002) (0.0168) (0.0037) (0.0000) (0.1144) (0.0031)
Log(Assets)
-0.0055*** 0.0334*** 0.0177*** 0.0358** 0.1177*** 0.047**
(0.0013) (0.0011) (0.0038) (0.0149) (0.0111) (0.0198)
U.S. Owned Dummy
0.0137 0.0044 -0.3216*** -0.1421 -0.218* -0.2569
(0.0117) (0.0078) (0.1081) (0.1899) (0.1244) (0.2818)
Europe Owned Dummy
0.0675*** -0.0179 -0.0050 0.0273 -0.0697 -0.1222
(0.0133) (0.0290) (0.0223) (0.1064) (0.0877) (0.1139)
Foreign
+
Owned Dummy
-0.0086 -0.1078*** -0.2335*** -0.0730 0.0665 0.0312
(0.0191) (0.0235) (0.0580) (0.1313) (0.0597) (0.0974)
Crisis Dummy
-0.2281*** -0.4378*** -0.2974*** -0.1978*** -0.2092***
(0.0091) (0.0314) (0.0604) (0.0205) (0.0613)
(U.S. Owned Dummy*Crisis Dummy)
-0.0423*** -0.0464*** 0.1207 -0.0527 -0.0758 -0.0697
(0.0094) (0.0079) (0.1060) (0.2247) (0.1101) (0.2600)
(Europe Owned Dummy*Crisis Dummy)
-0.1208*** -0.1832*** -0.0165 -0.1716 -0.0667 -0.0114
(0.0126) (0.0325) (0.0207) (0.1049) (0.0758) (0.1174)
(Foreign
+
Owned Dummy*Crisis Dummy)
-0.0092 -0.055** 0.2242*** 0.0203 0.0103 -0.0511
(0.0168) (0.0251) (0.0537) (0.1321) (0.0511) (0.0998)
(Openness*Crisis Dummy)
0.0605*** -0.0307*** 0.1061*** 0.0995** 0.0372*** 0.0736***
(0.0037) (0.0008) (0.0124) (0.0499) (0.0134) (0.0273)
R-squared 0.0237 0.1953 0.0395 0.1194 0.0957 0.124
Without Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Japan Asia Africa
Constant
0.2411*** -0.8325*** -0.8393*** 0.0156 0.6255* -2.7221*** -1.1964*** -0.1466
(0.0218) (0.0210) (0.0212) (0.0701) (0.3485) (0.2812) (0.1997) (0.3203)
Lag(Tier1/Assets)
0.00002 3.171*** 3.1793*** 0.0141*** 0.0000 2.5706*** 1.4687*** 0.0020
(0.00002) (0.0167) (0.0168) (0.0037) (0.0000) (0.1461) (0.1108) (0.0031)
Log(Assets)
-0.0044*** 0.0328*** 0.0331*** 0.01*** 0.0471*** 0.1097*** 0.0575*** 0.0200
(0.0011) (0.0011) (0.0011) (0.0034) (0.0163) (0.0117) (0.0091) (0.0158)
U.S. Owned Dummy
0.0031 0.0100 0.0046 -0.2509** -0.2554 -0.1897 -0.2535* -0.1342
(0.0119) (0.0076) (0.0078) (0.1058) (0.2066) (0.1918) (0.1345) (0.3047)
Europe Owned Dummy
0.1266*** -0.0453 -0.0230 0.0261 0.0062 -0.4768 -0.1023 -0.0682
(0.0126) (0.0295) (0.0290) (0.0216) (0.1176) (0.3007) (0.0927) (0.1193)
Foreign
+
Owned Dummy
0.0756*** -0.137*** -0.1179*** -0.1408** -0.1022 0.0437 0.0931
(0.0180) (0.0245) (0.0233) (0.0548) (0.1428) (0.0621) (0.0970)
Crisis Dummy
-0.1537*** -0.3481*** -0.2634*** -0.1602*** -0.1272**
(0.0080) (0.0283) (0.0548) (0.0203) (0.0610)
(U.S. Owned Dummy*Crisis Dummy)
-0.0417*** -0.0476*** -0.0463*** 0.0779 0.0041 -0.327** -0.0511 -0.0653
(0.0094) (0.0079) (0.0079) (0.1056) (0.2323) (0.1309) (0.1111) (0.2618)
(Europe Owned Dummy*Crisis Dummy)
-0.1322*** -0.1576*** -0.1838*** -0.0206 -0.1898* -0.0492 -0.0150
(0.0126) (0.0339) (0.0325) (0.0207) (0.1056) (0.0770) (0.1193)
(Foreign
+
Owned Dummy*Crisis Dummy)
-0.0275 -0.0469* -0.0552** 0.174*** 0.0536 0.0041 -0.0310
(0.0167) (0.0270) (0.0251) (0.0532) (0.1334) (0.0517) (0.0988)
(Openness*Crisis Dummy)
0.0296*** -0.0304*** -0.0307*** 0.0681*** 0.0441 0.0207*** 0.0132 0.0280
(0.0033) (0.0008) (0.0008) (0.0111) (0.0404) (0.0047) (0.0123) (0.0219)
R-squared 0.0056 0.1984 0.1953 0.0236 0.024 0.0234 0.0262 0.0027
Number of Observations 122,598 100,596 100,945 11,062 735 2,864 3,136 852
70
Table 3.11: Eects of Ownership on Bank Lending During the Recession (1)
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.4324 0.111*** 0.3682 0.2829 0.4272 -0.0889
(83.1749) (0.0260) (50.1496) (1.9382) (0.5725) (1.3815)
U.S. Owned Dummy
-1.1532 0.5168*** -14.6718 -0.2148 -0.0628 -0.0814
(5.4790) (0.1047) (32.7612) (0.9070) (0.2014) (0.3532)
Recession Dummy
-0.1531 -0.0144 -0.8259 -0.3251 -0.093*** -0.0559
(0.9753) (0.0737) (4.5607) (0.4660) (0.0194) (0.0540)
(U.S. Owned Dummy*Recession Dummy)
-0.2584 -0.4754 1.4336 0.0908 0.0548 0.0306
(4.8793) (0.3002) (38.9428) (2.5764) (0.1901) (0.5508)
R-squared 0.0003 0.0003 0.0003 0.0053 0.0091 0.0139
Without Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
1.4072 0.1107*** 0.1112*** 4.8484 0.3734 0.1341*** 0.1981***
(0.9912) (0.0261) (0.0259) (3.7109) (0.1714) (0.0226) (0.0322)
U.S. Owned Dummy
-0.8792 0.5201*** 0.5168*** -4.7170 -0.1824 -0.0541 -0.1156
(5.4097) (0.1053) (0.1047) (32.3629) (0.9021) (0.2018) (0.3461)
Recession Dummy
-0.1680 -0.0136 -0.0145 -0.8816 -0.3169 -0.095*** -0.0518
(0.9748) (0.0741) (0.0737) (4.5588) (0.4659) (0.0194) (0.0538)
(U.S. Owned Dummy*Recession Dummy)
-0.2836 -0.4773 -0.4753 0.8296 0.2693 0.0537 0.0264
(4.8787) (0.3022) (0.3002) (38.9375) (2.5786) (0.1901) (0.5506)
R-squared 0.0000 0.0003 0.0003 0.0000 0.0001 0.0020 0.0004
Number of Observations 165,018 85,286 85,813 35,838 8,476 19,612 5,601
whether internal capital markets are important in the international transmission of
nancial shocks. Since nancial shocks were not important in the 2001 recession,
we should not see any evidence of the international transmission of nancial shocks
in our estimates using the only the pre- and post-2011 recession sample.
Table 3.11 depicts the results. Compared to Table 3.3, the recession eect is
not signicant except in Asia. The interaction of the U.S.- owned dummy and the
Recession dummy are not signicant in any of the regions. Table 3.12 shows the
results that include the post-2001 recession dummy interacted with the European-
owned dummy. None of these interacted variables are signicant; we cannot detect
the eects of that 2001 recession on international lending. Thus, it appears that the
71
Table 3.12: Eects of Ownership on Bank Lending During the Recession (2)
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.4325 0.1109*** -8.3870 0.2718 0.4272 0.1185
(83.1527) (0.0261) (50.3124) (1.9369) (0.5726) (0.8875)
U.S. Owned Dummy
-0.5242 0.5169*** -8.5742 -0.0036 -0.0655 -0.0835
(5.4806) (0.1047) (32.8838) (0.9085) (0.2016) (0.3539)
Europe Owned Dummy
12.9464*** 0.0246 19.4995** 2.0476 -0.0567 -0.0142
(3.9653) (0.3876) (9.5070) (0.6015) (0.1411) (0.0983)
Recession Dummy
-0.0342 -0.0145 -0.1395 -0.1573 -0.0941*** -0.0504
(1.0086) (0.0738) (5.0664) (0.4864) (0.0196) (0.0586)
(U.S. Owned Dummy*Recession Dummy)
-0.3783 -0.4754 0.7695 -0.0678 0.0560 0.0251
(4.8862) (0.3002) (39.0067) (2.5785) (0.1902) (0.5514)
(Europe Owned Dummy*Recession Dummy)
-1.6198 0.0153 -3.1300 -1.8315 0.0601 -0.0365
(3.9546) (1.2136) (11.6353) (1.6841) (0.1452) (0.1509)
R-squared 0.0004 0.00003 0.00004 0.0070 0.0091 0.0140
Without Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
0.2429 0.1106*** 0.111*** 0.5645 0.1976 0.1336*** 0.2005***
(1.0339) (0.0262) (0.0260) (4.1769) (0.1798) (0.0230) (0.0347)
U.S. Owned Dummy
0.2851 0.5202*** 0.5169*** -0.4331 -0.0065 -0.0536 -0.1179
(5.4154) (0.1053) (0.1047) (32.4058) (0.9032) (0.2019) (0.3467)
Europe Owned Dummy
14.251*** 0.0091 0.0270 20.2688** 1.9087*** 0.0178 -0.0174
(3.6172) (0.4081) (0.3866) (9.0859) (0.5925) (0.1380) (0.0940)
Recession Dummy
-0.0380 -0.0138 -0.0145 -0.1709 -0.1409 -0.096** -0.0466
(1.0082) (0.0742) (0.0738) (5.0643) (0.4863) (0.0196) (0.0584)
(U.S. Owned Dummy*Recession Dummy)
-0.4136 -0.4771 -0.4753 0.1188 0.0933 0.0548 0.0212
(4.8856) (0.3022) (0.3002) (39.0014) (2.5809) (0.1902) (0.5511)
(Europe Owned Dummy*Recession Dummy)
-1.6624 0.0504 0.0150 -3.1598 -1.9112 0.0594 -0.0347
(3.9536) (1.2722) (1.2136) (11.6337) (1.6858) (0.1451) (0.1506)
R-squared 0.0002 0.0003 0.0003 0.0002 0.0016 0.0021 0.0005
Number of Observations 165,018 85,286 85,813 35,838 6,913 15,417 4,389
72
Table 3.13: Eects on Lending of Debt-GDP Ratio of Countries of the Foreign Bank
Aliates (1)
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
1.1399 0.1321 0.2525 0.4164 3.9582 0.9083***
(18.3268) (0.3450) (1.4946) (0.4072) (81.3291) (0.3024)
U.S. Owned Dummy
-0.7554 0.1956 -0.0051 0.0450 -3.9441 -0.1164
(2.0055) (0.3662) (0.6135) (0.0800) (14.5305) (0.1370)
Europe Owned Dummy
-0.6790 0.0098 0.2182 -0.0286 -5.0321 0.0223
(1.3968) (0.5607) (0.3968) (0.0534) (10.6967) (0.0617)
Crisis Dummy
-0.1412 -0.1754 -0.0261 -0.1358*** -0.1187 -0.0945
(0.2214) (0.1371) (0.2923) (0.0397) (0.1420) (0.1358)
(PUBOND*Crisis Dummy)
-0.0434 0.0766 -0.2741 -0.19** 0.0723 -0.2085
(0.3875) (0.2640) (0.5103) (0.0860) (0.1612) (0.3417)
R-squared 0.0054 0.0005 0.0106 0.0782 0.0031 0.0957
Without Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
1.5496** 0.1114 0.1321 0.2515 0.3024*** 3.8357 0.2241***
(0.7343) (0.4224) (0.3450) (0.3600) (0.0308) (3.3154) (0.0533)
U.S. Owned Dummy
-1.2405 0.2158 0.1956 -0.1256 0.0673 -3.7777 -0.1557
(1.2717) (0.4408) (0.3662) (0.6097) (0.0724) (13.7708) (0.1306)
Europe Owned Dummy
-1.1076 0.0344 0.0098 0.2380 -0.0219 -3.5626 0.0314
(1.0236) (0.6425) (0.5607) (0.3716) (0.0509) (9.3905) (0.0566)
Crisis Dummy
-0.1205 -0.1824 -0.1754 0.0857 -0.139*** -0.1172 -0.2213*
(0.2201) (0.1481) (0.1371) (0.2802) (0.0388) (0.1420) (0.1175)
(PUBOND*Crisis Dummy)
-0.0854 0.0942 0.0766 -0.5021 -0.1804** 0.0703 0.1372
(0.3843) (0.2870) (0.2640) (0.4817) (0.0834) (0.1612) (0.2882)
R-squared 0.0001 0.0005 0.0005 0.0011 0.0588 0.0001 0.0288
Number of Observations 11,993 2,270 2,373 3,759 857 1,765 490
73
global balance sheet shock present in the post-Lehman 2008 sample, had dierent
eects on foreign aliate bank lending than the 2001 Recession.
3.5.3 Why did European-bank aliates cut their lending
by more?
Why was the cross-border lending of European rms particularly aected by the
crisis? One reason is that European banks are large holders of the debt of their
own governments. If the deep recession following the 2008-Lehman crisis lowered
the abilities of European governments to pay back their debt, then the value (price)
of the government bonds held by the European banks will decline, damaging the
balance of the banks. Unfortunately, we do not have data on government bond
holdings at the individual bank level. Instead, we have data on the outstanding
debt to GDP ratios of the governments of each of the foreign bank aliates. For
example, we can match their government's debt-to-GDP ratio to all of the foreign
bank aliates operating in Korea. The assumption is that the higher the debt-to-
GDP ratio in a given country, the more the banks in that country hold their own
government's debt. In the results below, for all of the foreign bank-aliates, we
interact their home country's debt-GDP ratio with the crisis dummy variable.
The results in Table 3.13 and 3.14 show that during the nancial crisis, foreign
bank aliates from highly indebted countries cut their lending by more. These
results imply that high government indebtedness leads to a cut in bank lending
abroad.
74
Table 3.14: Eects on Lending of Debt-GDP Ratio of Countries of the Foreign Bank
Aliates (2)
With Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
1.1323 0.1095 0.2293 0.4055 3.9554 0.9045***
(18.3247) (0.3545) (1.5239) (0.3899) (81.3393) (0.3026)
U.S. Owned Dummy
-0.7161 0.2196 0.0743 0.0742 -3.9482 -0.1455
(2.0174) (0.3778) (0.7995) (0.0895) (14.5351) (0.2295)
Europe Owned Dummy
-0.583 0.0544 0.2399 0.0011 -4.9996 0.1061
(1.4090) (0.5830) (0.4957) (0.0605) (10.7001) (0.0955)
Crisis Dummy
-0.0686 -0.1182 0.012 -0.1226*** -0.1118 -0.0938
(0.2736) (0.2436) (0.5287) (0.0426) (0.1525) (0.1362)
(U.S. Owned Dummy*Crisis Dummy)
-0.0608 -0.045 -0.1164 -0.0538 0.0053 0.0093
(0.3343) (0.1708) (0.7524) (0.0734) (0.4681) (0.2831)
(Europe Owned Dummy*Crisis Dummy)
-0.1447 -0.0772 -0.0341 -0.0459 -0.0484 -0.1389
(0.2802) (0.2666) (0.4605) (0.0493) (0.3099) (0.1219)
(PUBOND*Crisis Dummy)
-0.0467 0.0454 -0.2809 -0.1787** 0.0687 -0.0033
(0.3877) (0.2835) (0.5152) (0.0868) (0.1638) (0.3991)
R-squared 0.0054 0.0005 0.0107 0.0814 0.0031 0.0987
Without Country Fixed-Eects
ALL U.S. U.S. and Canada Europe Latin America Asia Africa
Constant
1.5011** 0.0764 0.1095 0.2113 0.2911 3.8328 0.1773***
(0.7416) (0.4346) (0.3545) (0.4472) (0.0318) (3.3162) (0.0652)
U.S. Owned Dummy
-1.1978 0.2527 0.2196 -0.0351 0.0987 -3.7812 -0.1574
(1.2907) (0.4548) (0.3778) (0.7969) (0.0843) (13.7769) (0.2228)
Europe Owned Dummy
-1.0094 0.0900 0.0544 0.2806 0.0074 -3.5294 0.1161
(1.0403) (0.6676) (0.5830) (0.4784) (0.0587) (9.3949) (0.0859)
Crisis Dummy
-0.0458 -0.0927 -0.1182 0.1521 -0.126*** -0.1102 -0.1976*
(0.2724) (0.2972) (0.2436) (0.5178) (0.0416) (0.1524) (0.1189)
(U.S. Owned Dummy*Crisis Dummy)
-0.0645 -0.0705 -0.0450 -0.1339 -0.0538 0.0050 -0.0278
(0.3342) (0.2108) (0.1708) (0.7517) (0.0730) (0.4680) (0.2760)
(Europe Owned Dummy*Crisis Dummy)
-0.1478 -0.1010 -0.0772 -0.0652 -0.0450 -0.0490 -0.1482
(0.2802) (0.3099) (0.2666) (0.4597) (0.0491) (0.3099) (0.1143)
(PUBOND*Crisis Dummy)
-0.0888 0.0465 0.0454 -0.5120 -0.1689** 0.0666 0.2978
(0.3844) (0.3162) (0.2835) (0.4857) (0.0840) (0.1637) (0.3183)
R-squared 0.0002 0.0005 0.0005 0.0011 0.0613 0.0001 0.0324
Number of Observations 11,993 2,270 2,373 3,759 857 1,765 490
75
3.6 Conclusion
In this paper, we examine the period before and after the post-2008 Lehman collapse
period to see if the nationalitiy of the bank in a particular market matters in how
much the bank cuts its lending. While banks on average cut their lending after the
crisis, European bank aliates located in Latin America and Asia cut their lending
by more than the other banks and the domestic banks located in those regions. If
the balance sheets of the parents of the European bank branches were the most
harmed by the crisis, then our results imply that internal capital markets matter in
the international transmission of nancial shocks. Since in our sample, European
bank aliates accounted for at least 20 percent of the total loans in the Latin
American banking system, the withdrawal of loans from European-bank aliates
should have had a large eect on the transmission of European nancial shocks to
Latin America.
76
References
Arellano, C., Bai, Y., and Zhang, J. (2009). Firm Dynamics and Financial Devel-
opment. NBER Working Papers 15193, National Bureau of Economic Research,
Inc.
Baker, S. H. (1973). Risk, Leverage, and Protability: An Industry Analysis. The
Review of Economics and Statistics, 55(4): 503{07.
Beck, T., Levine, R., and Loayza, N. (2000). Finance and the sources of growth.
Journal of Financial Economics, 58(1-2): 261{300.
Beck, T., Lundberg, M., and Majnoni, G. (2006). Financial intermediary de-
velopment and growth volatility: Do intermediaries dampen or magnify shocks?
Journal of International Money and Finance, 25(7): 1146{1167.
Becker, B., and Sivadasan, J. (2006). The eect of nancial development on the
investment-cash
ow relationship - cross-country evidence from Europe. Working
Paper Series 689, European Central Bank.
Bilodeau, K. (). The Financial Development Report 2008. World Economic Forum.
isbn 9789295044081.
Bottazzi, G., Secchi, A., and Tamagni, F. (2010). Financial Constraints and
Firm Dynamics. Discussion Papers 2010/99, Dipartimento di Scienze Economiche
(DSE), University of Pisa, Pisa, Italy.
Campello, M., Graham, J. R., and Harvey, C. R. (2010). The real eects of nancial
constraints: Evidence from a nancial crisis. Journal of Financial Economics,
97(3): 470{487.
Cetorelli, N., and Goldberg, L. (2012). Banking Globalization, Monetary Trans-
mission, and the Lending Channel. Journal of Finance.
77
Cetorelli, N., and Goldberg, L. S. (2008). Banking globalization, monetary trans-
mission, and the lending channel. Discussion paper.
Cetorelli, N., and Goldberg, L. S. (2010). Global Banks and International Shock
Transmission: Evidence from the Crisis. NBER Working Papers 15974, National
Bureau of Economic Research, Inc.
Chang, R., and Velasco, A. (1999). Liquidity crises in emerging markets: Theory
and policy. Discussion paper.
Chinn, M. D., and Ito, H. (2008). A New Measure of Financial Openness. Journal
of Comparative Policy Analysis: Research and Practice, 10(3): 309{322.
Clementi, G. L., and Hopenhayn, H. A. (2002). A Theory of Financing Constraints
and Firm Dynamics. RCER Working Papers 492, University of Rochester - Center
for Economic Research (RCER).
Cooley, T., and Quadrini, V. (2006). Monetary policy and the nancial decisions
of rms. Economic Theory, 27(1): 243{270.
Cornett, M. M., McNutt, J. J., Strahan, P. E., and Tehranian, H. (2011). Liquidity
risk management and credit supply in the nancial crisis. Journal of Financial
Economics, 101(2): 297 { 312. issn 0304-405X.
Dekle, R., and Kletzer, K. M. (2001). Domestic Bank Regulation and Financial
Crises: Theory and Empirical Evidence from East Asia. NBER Working Papers
8322, National Bureau of Economic Research, Inc.
Demirguc-Kent, A., and Detragiache, E. (1998). Financial liberalization and -
nancial fragility. Policy Research Working Paper Series 1917, The World Bank.
Demirg u c-Kunt, A., and Huizinga, H. (1999). Determinants of Commercial Bank
Interest Margins and Protability: Some International Evidence. The World Bank
Economic Review, 13(2): pp. 379{408. issn 02586770.
Didier, T., Hevia, C., and Schmukler, S. L. (2011). How resilient were emerging
economies to the global crisis ? Policy Research Working Paper Series 5637, The
World Bank.
Djankov, S., Hart, O., McLiesh, C., and Shleifer, A. (2008). Debt Enforcement
around the World. Journal of Political Economy, 116(6): 1105{1149.
Djankov, S., McLiesh, C., and Shleifer, A. (2007). Private credit in 129 countries.
Journal of Financial Economics, 84(2): 299{329.
78
Dodonov, B. (2009). How Do Capital Requirements Aect the Performance of
the Real Sector? Microevidence from Selected EU Countries. Job market paper,
University of Houston.
Duchin, R., Ozbas, O., and Sensoy, B. A. (2010). Costly external nance, cor-
porate investment, and the subprime mortgage credit crisis. Journal of Financial
Economics, 97(3): 418{435.
Furman, J., and Stiglitz, J. (1998). Economic Crises: Evidence and Insights from
East Asia. Brookings Papers on Economic Activity, 29(2): 1{136.
Gelman, A., and Hill, J. (2007). Data analysis using regression and multi-
level/hierarchical models, volume Analytical methods for social research. New
York: Cambridge University Press.
Holly, S., and Raissi, M. (2009). The Macroeconomic Eects of European Financial
Development: A Heterogenous Panel Analysis. Discussion paper.
Hox, J. J. (2010). Multilevel Analysis: Techniques and Applications. New York:
Routledge. isbn 9781848728455.
Hsiao, C. (1986). Analysis of panel data / Cheng Hsiao. Cambridge University
Press, Cambridge ; New York :. isbn 0521251508.
Huang, Y. (2011). Determinants of Financial Development. Palgrave Macmillan.
isbn 9780230273672.
Joseph E. Murphy, J. (1968). Eect of Leverage on Protability, Growth and
Market Valuation of Common Stock. Financial Analysts Journal, 24(4): 121{123.
Kalemli-Ozcan, S., Srensen, B. E., and Volosovych, V. (2010). Deep Financial
Integration and Volatility. TU"SY'AD-Koc, University Economic Research Forum
Working Papers 1006, TUSIAD-Koc University Economic Research Forum.
Kaminsky, G. L., and Reinhart, C. M. (1996). The twin crises: the causes of
banking and balance-of-payments problems. Discussion paper.
Khan, A. (2000). The nance and growth nexus. Business Review, (Jan): 3{14.
Khwaja, A. I., and Mian, A. (2008). Tracing the Impact of Bank Liquidity Shocks:
Evidence from an Emerging Market. The American Economic Review, 98(4):
1413{1442.
79
Kim, C.-S., Mauer, D. C., and Sherman, A. E. (1998). The Determinants of
Corporate Liquidity: Theory and Evidence. Journal of Financial and Quantitative
Analysis, 33(03): 335{359.
King, R. G., and Levine, R. (1993). Finance and Growth: Schumpeter Might Be
Right. The Quarterly Journal of Economics, 108(3): 717{37.
Kroszner, R. S., Laeven, L., and Klingebiel, D. (2007). Banking crises, nancial
dependence, and growth. Journal of Financial Economics, 84(1): 187{228.
Larrain, B. (2004). Financial development, nancial constraints, and the volatility
of industrial output. Discussion paper.
Levchenko, A. A., Rancie`re, R., and Thoenig, M. (2008). Growth and Risk at
the Industry Level: the Real Eects of Financial Liberalization. CEPR Discussion
Papers 6715, C.E.P.R. Discussion Papers.
Levy, A., and Schich, S. (2010). The Design of Government Guarantees for Bank
Bonds.
Mayer, C. (1990). Financial Systems, Corporate Finance, and Economic Devel-
opment. In Asymmetric Information, Corporate Finance, and Investment, NBER
Chapters, pages 307{332. National Bureau of Economic Research, Inc.
Mendoza, E. G., and Quadrini, V. (2009). Financial Globalization, Financial
Crises and Contagion. NBER Working Papers 15432, National Bureau of Economic
Research, Inc.
Mishkin, F. S. (2009). Why We Shouldn't t Turn Our Backs on Financial Global-
ization. IMF Sta Papers, 56(1): 139{170.
Mitton, T. (2002). A cross-rm analysis of the impact of corporate governance on
the East Asian nancial crisis. Journal of Financial Economics, 64(2): 215 { 241.
Morais, B. (2010). Should I Stay or Should I Go: Investor Protection, Firm
Selection and Aggregate Productivity. Job market paper, UCLA.
Murphy, J. E., and Arditti, J. D. (1967). Risk and the Required Return on Equity.
The Journal of Finance, 22(1): 19{36.
Peek, J., and Rosengren, E. S. (2000). Collateral damage: Eects of the Japanese
bank crisis on real activity in the United States. The American Economic Review,
90(1): 30{45.
80
Raddatz, C. (2006). Liquidity needs and vulnerability to nancial underdevelop-
ment. Journal of Financial Economics, 80(3): 677{722.
Rajan, R. G. (2005). Has nancial development made the world riskier? Proceed-
ings, (Aug): 313{369.
Rajan, R. G., and Zingales, L. (1998). Financial Dependence and Growth. Amer-
ican Economic Review, 88(3): 559{86.
Rose, A. K. (2004). Do We Really Know That the WTO Increases Trade? Amer-
ican Economic Review, 94(1): 98{114. doi 10.1257/000282804322970724.
Tan, C. (2010). Time-Varying Uncertainty and Financially Constrained Invest-
ment: Theory and Evidence. Job market paper, University of California at Berke-
ley.
81
Appendix A Robustness Check using
\Placebo" Specications
One way to see if we are getting spurious results from our regression is to randomly
assign the rm-level observations of a particular country to the nancial develop-
ment index of another country. If the variables interacted with the randomly chosen
nancial development index lose signicance in these \placebo" specications, then
we can ascribe a more causal interpretation to nancial development in our original
specication with the correct assignment of countries.
In Table A-1, I compare the results of the \placebo" random country assign-
ment specications, with the \correct" country assignment specications. In the
\placebo" specications, the variables interacted with randomly assigned nancial
development indices are all insignicant. The insignicance of the \placebo" eect
suggests that we can attach a causal interpretation to our original specication :
the nancial development of a country does matter in propagating a crisis.
82
Table A.1: Random Assignment
Random Assignment of Firm-level Observations Financial Development Index
Austria ! France
Belgium ! Finland
Bulgaria ! Greece
Switzerland ! Ireland
Czech Republic ! Italy
Germany ! Poland
Denmark ! Lithuania
Estonia ! Estonia
Spain ! Austria
Finland ! Slovenia
France ! Czech Republic
United Kingdom ! Latvia
Greece ! Denmark
Croatia ! Serbia
Ireland ! Bulgaria
Italy ! Spain
Lithuania ! Sweden
Luxembourg ! Luxembourg
Latvia ! Ukraine
Poland ! United Kingdom
Portugal ! Portugal
Serbia ! Croatia
Sweden ! Switzerland
Slovenia ! Belgium
Ukraine ! Germany
83
Table A.2: Comparison of \Correct" and \Random" Assigned of Country Financial
Development Indices
Dependent Variable: logEBIT
(1) (2) (3) (4)
Intercept
13.621*** 13.5067*** 13.3503*** 13.2663***
(0.337500) (36.33) (0.348900) (35.07)
IDEF
0.4714*** 0.5002*** 0.3586*** 0.421***
(0.025420) (20.41) (0.024440) (17.83)
FD*IDEF
-0.00035 -0.00044** 0.000019 -0.00049***
(0.000318) (-2.45) (0.000305) (-2.89)
Size
1.4915*** 1.4899***
(0.016500) (90.31)
Age
0.009702*** 0.009725*** 0.005789*** 0.005803***
(0.000365) (26.67) (0.000352) (16.48)
Crisis
0.1449*** 0.2058*** 0.1431*** 0.2099***
(0.033410) (6.97) (0.032400) (7.29)
Crisis*FD
0.000185 -0.00032* 0.000115 -0.00041***
(0.000348) (-1.84) (0.000334) (-2.41)
Crisis*Age
-0.00139** -0.00143** -0.00147** -0.00041***
(0.000629) (-2.27) (0.000605) (-2.41)
Crisis*Size
-0.01769 -0.01185
(0.027640) (-0.43)
Random Assignment YES NO YES NO
84
Appendix B Variable Denitions
The dependent variable is dened as
Loans
i;j;t
Assets
i;j;t1
wherei denotes the country in which
a bank is located,j denotes the bank andt denotes time. As explanatory variables,
the ratio of Tier 1 capital to assets in (t 1) which is dened
Tier 1Capital
i;j;t1
Assets
, the
ratio of the problem loans to total loans
TotalProblemsloans
i;j;t1
Loans
i;j;t1
.
Table B.1: Summary of Capital Market Openness
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
All 2.09 2.09 2.06 2.10 2.11 2.11 2.08 2.02 1.99 2.01 1.99
Europe 2.28 2.27 2.30 2.32 2.32 2.33 2.36 2.36 2.36 2.35 2.35
Latin America 0.51 0.55 0.48 0.63 0.87 1.07 1.15 1.09 1.07 1.17 1.04
Asia 1.32 1.48 1.40 1.37 1.30 1.23 1.17 1.13 1.10 1.28 1.12
Africa -0.50 -0.57 -0.52 -0.52 -0.48 -0.49 -0.49 -0.47 -0.50 -0.39 -0.47
In addition, a bank's liquidity, which is the ratio of liquid assets to assets in
time (t 1), and size variable log(assets
t
) are also considered. We dene the crisis
dummy variable as 1 if the scal year is 2008 and 2009; otherwise it is denoted as
0. We also dene the recession dummy variable as 1 if the year is 2001 and 2002;
otherwise it is 0.
For capital market openness, we use the Chinn-Ito index
which takes on higher
values, the more open a country is to cross-border capital transactions.
We use the data from Chinn-Ito website and use the variable \kaopen" as a measure of capital
market openness. More detailed data descriptions for construction of the data can be found at
Chinn and Ito (2008)
85
Abstract (if available)
Abstract
This dissertation asks two empirical questions : How did financial development affect the performance of European firms before and after the 2008 Credit Crisis, and do foreign bank affiliates cut their lending more than the domestic banks in a financial crisis. ❧ Chapter 2 studies the impact of the recent credit crisis on firm performance. The recent credit crisis led to a deep recession in the U.S. and the rest of the world. This chapter seeks to identify the significant relationship between financial development and firm-level performance in advanced European economies based on the recent credit crisis. To evaluate firm-level performance, it includes cross-country differences in financial development and cross-firm differences in dependence on external financing, and study how financial development interacts with firm dependence on external financing. The results show that financial development is positively related to a firm's earnings in tranquil times. Surprisingly, however, that same financial development can also exacerbate the impact of a crisis. The results are robust to estimation using various instruments for the endogenous variables, and are statistically significant across different specifications. ❧ Chapter 3 studies the impact of the recent credit crisis on bank lending. It contributes to the literature on the international transmission of balance sheet shocks that pummeled the banks of the industrialized countries in 2008 and 2009. It examines over time bank level data on 250,000 banks located around the world. Our identification strategy relies on the differential responses of foreign and domestic banks to the post-Lehman 2008 crisis. If in a particular market, say in Korea, a foreign-affiliated bank's (Citibank, Korea's) lending falls by more than a domestic bank's (Kookmin's) lending, then we attribute this additional decline to the tightening of the foreign affiliates internal capital market at its headquarters. We control for the decline in market conditions common to all banks in a particular region by the decline in lending by the banks other than the foreign affiliated bank. We find evidence that internal capital markets do indeed affect cross-border lending. In particular, European bank affiliates in Latin America and Asia cut their lending by more than the domestic banks located in these regions.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Two essays on major macroeconomic shocks in the Japanese economy: demographic shocks and financial shocks
PDF
Essays on macroeconomics and income distribution
PDF
Essays on government intervention in financial crises
PDF
Essays in corporate finance
PDF
Financial crises and trade policy in developing countries
PDF
Money and trade, hinterland and coast, empire and nation-state: an unusual history of Shanxi piaohao, 1820-1930
PDF
Essays on development and health economics
PDF
The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector
PDF
Essays in international economics
PDF
Emerging market financial crises, investors and monetary policy
PDF
Essay on monetary policy, macroprudential policy, and financial integration
PDF
Early-warning systems for crisis risk
PDF
Does fair value accounting exacerbate the pro-cyclicality of bank lending?
PDF
Essays on firm investment, innovation and productivity
PDF
Comparative study on Asian financial capitals' competitiveness: focused on strengths and weaknesss of the city of Seoul
PDF
Essays on education and institutions in developing countries
PDF
Essays in political economy and mechanism design
PDF
Essays in panel data analysis
PDF
Essays in international economics
PDF
Essays on the new economic history of the Middle East
Asset Metadata
Creator
Lee, Mihye
(author)
Core Title
Two essays on financial crisis and banking
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Economics
Publication Date
08/01/2012
Defense Date
05/09/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
bank lending,external financing,financial crisis,financial development,OAI-PMH Harvest
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Dekle, Robert (
committee chair
), Carvalho, Daniel (
committee member
), Nugent, Jeffrey B. (
committee member
)
Creator Email
himihye@gmail.com,mihyelee@usc.edu
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-80501
Unique identifier
UC11290460
Identifier
usctheses-c3-80501 (legacy record id)
Legacy Identifier
etd-LeeMihye-1095.pdf
Dmrecord
80501
Document Type
Dissertation
Rights
Lee, Mihye
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
bank lending
external financing
financial crisis
financial development