CREDIT MARKETS, FINANCIAL CRISES, AND THE MACROECONOMY
By
Junghwan Hyun

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Economics - Doctor of Philosophy
2014

ABSTRACT
CREDIT MARKETS, FINANCIAL CRISES, AND THE MACROECONOMY
By
Junghwan Hyun
This study consists of three chapters, each of which is an individual paper. The first chapter investigates how the dynamic process of reallocation of credit across firms behaves before and after
financial crises. Applying the methodology proposed by Davis and Haltiwanger (1992) for measuring job reallocation, we construct measures of credit reallocation across Korean firms in the
1981-2012 period. The credit boom preceding the 1997 financial crisis featured a modest intensity
of credit reallocation. By contrast, after the crisis and the associated reforms, credit reallocation
significantly intensified and started to comove with the business cycle, while credit growth slowed
down (deleveraging). The higher dynamism of the credit sector in reallocating liquidity cannot be
explained by “flight to quality” episodes but reflects a structural change in the credit reallocation
process that has persisted since the end of the crisis. The intensification of credit reallocation appears to have been associated with enhanced allocative efficiency.
The second chapter explores the evolution of credit reallocation across Korean non-financial
firms for the period 1981-2012. I employ a dynamic latent factor model that decomposes regional
credit reallocation rates into national, region-specific and idiosyncratic components. I find that
the common factor explaining common movement across 16 regional credit flows increased after
the 1997 financial crisis. The common factor comoves with national excess reallocation. It is
positively and strongly correlated with national excess reallocation, while it is negatively correlated with national net credit growth. It exhibits mild counter-cyclicality. I examine what extent
the volatility of credit reallocation was driven by national, region-specific and idiosyncratic components. This study uncovers evidence that the national factor accounts for a sizable fraction of
regional reallocation rates of total credit and loans, while it plays only a minor role in explaining
the fluctuation in regional reallocation rates of bonds.

The last chapter explores the relationship between religion and bank performance. The study
uses data on credit unions in Korea for the period 2000 to 2007 to investigate the effects of religion
on bank performance. The empirical results show that credit unions based on religious institutions not only suffer less from troubled loans but they also enjoy higher profits relative to ordinary
ones. I find that the religious credit unions unique features, such as non random potential clientele,
rich soft information and reputational incentive to repay, are likely to be what enables them to
outperform.

ACKNOWLEDGEMENTS

I would like to acknowledge my gratitude to my adviser, Raoul Minetti, with whom I co-wrote
Chapter 1, for his encouragement and guidance, without which this dissertation would not have
been written. Especially, he guided me through how to develop a research idea and how to write
a good research paper. I give thanks to Chun (Susan) Zhu for her guidance on my research. She
is one of the kindest people who I have ever met. I would also like to thank Hayong Yun, Luis
Araujo, and Christian Ahlin for their valuable advice and feedback on my research. The course
provided by Luis Araujo is the best one that I have ever taken.
I also thank Professor Chonghyun Nam, Professor Byoungheon Jun, Dr. Shindong Jeong,
Byop-Yong Jeon, Hongbag Jung, Kyusu Kim, Hannyung Lee, and Yunseong Lee, all of whom
encouraged me to start the PhD program when I was working at the Bank of Korea.
I am also grateful to my family and friends. I would like to especially thank my grandmother
for her material and emotional support. Without her support, I would never have made it through
the last five years.

iv

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii
CHAPTER 1
1.1
1.2
1.3

1.4

1.5

1.6
1.7

1.8
1.9

CREDIT REALLOCATION, DELEVERAGING, AND FINANCIAL
CRISES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Prior Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Crisis, Reforms, and the Credit Market . . . . . . . . . . . . . . . . . . . . . . .
1.3.1 Corporate Reforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.2 Financial Reforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3.3 The Effects on the Credit Market . . . . . . . . . . . . . . . . . . . . . .
Data and Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.1 The Data Set . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4.2 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Credit Reallocation, Credit Boom, Deleveraging . . . . . . . . . . . . . . . . . .
1.5.1 Intensity of Credit Reallocation . . . . . . . . . . . . . . . . . . . . . . .
1.5.2 Size and Persistence of Credit Changes . . . . . . . . . . . . . . . . . .
1.5.3 Comparison with job flows . . . . . . . . . . . . . . . . . . . . . . . . .
The Role of “Flights to Quality” . . . . . . . . . . . . . . . . . . . . . . . . . .
The Dynamic Pattern of Credit Reallocation . . . . . . . . . . . . . . . . . . . .
1.7.1 Volatility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.2 Cyclical Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.7.2.1 Unconditional Correlation . . . . . . . . . . . . . . . . . . . .
1.7.2.2 Conditional Correlation . . . . . . . . . . . . . . . . . . . . .
1.7.2.3 Robustness Analysis . . . . . . . . . . . . . . . . . . . . . . .
Allocative Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1
1
5
6
7
8
8
9
9
10
12
12
17
20
20
25
25
26
26
29
30
31
33

.
.
.
.
.
.
.
.
.
.
.

35
35
37
38
38
39
44
44
44
45
46

CHAPTER 2
2.1
2.2
2.3

2.4

2.5

DRIVING FORCES BEHIND THE EVOLUTION OF CREDIT REALLOCATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Spatial Heterogeneity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Data and Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
What is the Common Factor? . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.1 Common factor: Total Credit, Loans, and Bonds . . . . . . . . . . . . .
2.4.2 Common factor: Chaebol Affiliation . . . . . . . . . . . . . . . . . . .
2.4.3 Comovement with National Credit Flows and GDP . . . . . . . . . . .
Variance Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

v

.
.
.
.
.
.
.
.
.
.
.

2.6

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

CHAPTER 3
3.1
3.2
3.3
3.4
3.5

RELIGION AND BANK PERFORMANCE: EVIDENCE FROM CREDIT
UNIONS IN KOREA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
Sources of Better Performance of RCUs . . . . . . . . . . . . . . . . . . . . . . . 57
Data and Model Specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

vi

LIST OF TABLES

Table 1.1

Magnitude of Gross Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Table 1.2

Average Flows due to Large Changes . . . . . . . . . . . . . . . . . . . . . . . 21

Table 1.3

Credit Reallocation in Sub-Groups . . . . . . . . . . . . . . . . . . . . . . . . . 24

Table 1.4

Volatility and Unconditional Correlation . . . . . . . . . . . . . . . . . . . . . . 27

Table 1.5

Properties of Idiosyncratic Flows . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Table 1.6

Decomposition of Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Table 1.7

Conditional Correlation of Credit Reallocation . . . . . . . . . . . . . . . . . . 30

Table 1.8

Efficiency of Credit Reallocation . . . . . . . . . . . . . . . . . . . . . . . . . . 33

Table 2.1

Magnitude of Gross Flows (Total Credit) . . . . . . . . . . . . . . . . . . . . . 40

Table 2.2

Magnitude of Gross Flows (Loans) . . . . . . . . . . . . . . . . . . . . . . . . . 41

Table 2.3

Magnitude of Gross Flows (Bonds) . . . . . . . . . . . . . . . . . . . . . . . . 42

Table 2.4

Cyclicality of the Common Factor . . . . . . . . . . . . . . . . . . . . . . . . . 45

Table 2.5

Variance Decomposition for Credit Reallocation by 16 Regions . . . . . . . . . 48

Table 2.6

Variance Decomposition for Loan Reallocation by 16 Regions . . . . . . . . . . 49

Table 2.7

Variance Decomposition for Bond Reallocation by 16 Regions . . . . . . . . . . 50

Table 2.8

Variance Decomposition (Chaebol-affiliated firms) . . . . . . . . . . . . . . . . 51

Table 2.9

Variance Decomposition (Nonl-affiliated firms) . . . . . . . . . . . . . . . . . . 52

Table 3.1

Summary Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

Table 3.2

Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

vii

LIST OF FIGURES

Figure 1.1: GDP and Business Sector Debt and Equity . . . . . . . . . . . . . . . . . . . . 13
Figure 1.2: Credit Change and Credit Reallocation . . . . . . . . . . . . . . . . . . . . . . . 14
Figure 1.3: Loan and Bond Reallocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Figure 1.4: Large Credit Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Figure 1.5: Efficiency of Credit Reallocation . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Figure 2.1: Evolution of Common Factor (1) . . . . . . . . . . . . . . . . . . . . . . . . . 43
Figure 2.2: Evolution of Common Factor (2) . . . . . . . . . . . . . . . . . . . . . . . . . . 47

viii

CHAPTER 1
CREDIT REALLOCATION, DELEVERAGING, AND FINANCIAL CRISES

1.1 Introduction
The Great Recession has reignited the debate over the long-run effects of financial crises and
of the reforms frequently enacted in their aftermath. In recent decades, while some financial crises
(e.g., those of Chile in 1982 and South Korea in 1997) have been followed by several years of
sustained output growth, other crises (e.g., Mexico in 1982) have marked the onset of prolonged
periods of stagnation. The role of the credit market is at the heart of the analysis of the crises.
Several scholars and policy makers put emphasis on the credit booms that often precede the crises
and the sluggish credit growth (deleveraging) that follows them (Gourinchas and Obstfeld, 2012;
Dell’Ariccia, Igan, Laeven and Tong, 2012; IMF, 2004). A popular argument is that periods of
credit bonanza can fuel excessive investment and a poor allocation of financial resources, culminating in financial crashes. In turn, financial crashes can trigger a drastic shift in the lending
policies of investors and financial institutions, resulting into slow credit growth during the subsequent recoveries (The Economist, 2012; Mendoza and Terrones, 2012). The policies enacted in
response to financial crashes would exacerbate the creditless nature of the recoveries.
In contrast with the breadth of knowledge on the behavior of credit aggregates, we know very
little about the dynamic process of reallocation of credit before and after financial crises. Yet, there
is growing evidence that the allocation of physical and financial inputs plays a role as relevant as
their total volume in affecting aggregate economic activity. Numerous studies find that the impact
of aggregate shocks and structural reforms on the macroeconomy occurs through the allocation
of labor, capital, and financial resources as much as through their total volume (Caballero and
Hammour, 2005, 2001; Eisfeldt and Rampini, 2006; Caballero, Hoshi and Kashyap, 2008). In
particular, the dynamism with which an economy is able to reallocate financial resources across

1

firms is deemed as crucial for efficiency and growth (Beck, Levine and Loayza, 2000; Wurgler,
2000; Galindo, Schiantarelli and Weiss, 2007). These observations elicit fundamental questions:
do the credit booms that precede financial crises feature an intense reallocation of credit or a mere
rollover of credit to firms already served by the credit market? Do financial crises and the subsequent reforms enhance the flexibility with which liquidity is reallocated across businesses? Or
does the deleveraging process that follows the crises stifle the dynamism with which the credit
market reallocates liquidity? Answering these questions can yield critical insights into the interaction between the credit market and the macroeconomy. It can also inform us about the optimal
policy response to credit booms and busts. For example, during a creditless recovery, a policy that
promotes the creation of new lines of credit and a policy that prevents the termination of existing
credit relationships can both boost credit growth. However, these two policies will exert opposite
effects on credit reallocation: promoting credit creation will foster credit reallocation, hindering
credit destruction will depress it. What policy should thus be pursued?
This paper takes a first step towards addressing these questions. We study the dynamic process
of reallocation of credit across South Korean non-financial businesses in the 1981−2012 period
and investigate whether the credit reallocation process changed after the 1997 financial crisis and
the subsequent reforms. The Korean economy and our database constitute an ideal testing ground
for our purposes. Credit is a key source of external finance for Korean firms (accounting for
almost 82% of their external funding in 2000).1 Our unique data set comprises unusually rich
microeconomic data on more than 30,000 non-financial firms, representing about 49.2% of the
employment of Korea in 2000, for example.2 Moreover, the data set covers a long time period (33
years) and features the occurrence of a major financial crisis around the sample midpoint (end of
1997). This allows to separate cyclical changes in the credit reallocation process, as induced by
the crisis, from structural long-lasting changes.
Credit growth was rapid throughout the 1990s and further accelerated during the credit boom
1 Source: Flow of Funds, Bank of Korea.
2 The data source is KISLINE, the business information source provided by the leading Korean

credit rating agency, Korea Investors Service (KIS), which is affiliated with Moody’s.

2

that took place from 1993, till the onset of the crisis in 1997. Prior to the crisis, the allocation of
credit was strongly influenced by government policies. Many firms, especially those affiliated to
industrial groups (chaebols), were guaranteed the renewal of existing loans without close scrutiny
by financial institutions (Hong, Lee and Lee, 2007). The 1997 crisis caused a credit crunch and
a sharp decline in GDP (by 5.7% in 1998). In response to the crisis, the government enacted
structural reforms of the corporate and financial sectors that affected both the demand and the
supply side of the credit market. These reforms aimed at reducing firms’ leverage and inducing
lenders to adopt more selective policies in allocating credit. The economy started to recover from
the crisis in the second half of 1998 and GDP growth rebounded to 10.7% in 1999 and 8.8% in
2000. In the years following the crisis, credit to the business sector grew at a pace significantly
lower than in the pre-crisis period, triggering a deleveraging of the business sector. Economists
debate whether such a deleveraging process was healthy or exerted a drag on the recovery.
To measure credit reallocation, we employ the methodology proposed by Davis and Haltiwanger (1992) for the measurement of job reallocation and used by Herrera, Kolar and Minetti
(2011) for the measurement of credit reallocation across U.S. firms. Average real credit growth
equalled 10.8% in the pre-crisis (1981–1996) period, and peaked at 12.9% during the 1993–1996
credit boom. After the crisis, during the deleveraging period (1999 through 2004) credit shrank at a
rate of 1.5%, and overall, between 1999 and 2012, expanded at an annual rate of only 4.3%. A drop
in credit growth can be attained through a reduction in the rate of credit creation and a relatively
stable credit destruction, thus implying less intense reallocation of credit. Alternatively, it can be
attained through a relatively stable credit creation and an increase in credit destruction, thus implying more intense reallocation of credit. We find that Korea followed the latter path. On average,
inter-firm gross credit reallocation (the sum of credit creation and credit destruction) was about
21.4% between 1981 and 2012, the same order of magnitude of that found for the United States
by Herrera et al. (2011). Most importantly, the intensity of credit reallocation rose significantly
after the crisis, from an average of 17.9% in 1981–1996 to an average of 24.7% in 1999–2012. If
we net out from gross credit reallocation the amount of reallocation strictly needed to accommo-

3

date the net credit change, we obtain that, after being depressed at an average of 7.0% during the
1993–1996 credit boom, excess credit reallocation jumped to an average of 19.0% in 1999–2012.
The reader could wonder whether such a staggering increase in credit reallocation after the
crisis can be explained by the “flights to quality” (e.g., the flights of credit from small to large
firms) that often characterize crises (Bernanke, Gertler and Gilchrist, 1999). The results dispute
this hypothesis. Consistent with the flight to quality argument, we indeed uncover evidence that the
reshuffling of credit across classes of firms different in size, industry, and location intensified during the crisis. However, after the crisis, the importance of the reallocation of credit within groups
of firms relatively homogenous for size, industry, and location, increased while the reshuffling of
credit across different classes of firms (as induced, for instance, by aggregate or sectoral shocks)
became less important.
All in all, this first set of findings are consistent with the view that the structural corporate
and financial reforms enacted in response to the crisis made the process of reallocation of credit
across businesses more frictionless and fluid, for example by ameliorating the lending policies of
financial institutions (Lim, 2010). In support of this argument, we find that, when we break down
credit into loans and bonds, the increase in reallocation occurred for both types of credit but for
loans it was more pronounced. This hints at a change in the dynamism with which after the crisis
financial institutions, such as banks, reallocated loans.
We then turn to explore whether, besides its intensity, the dynamic behavior of credit reallocation also changed after the crisis. The volatility of credit reallocation increased. Moreover, in
line with what found for the intensity of reallocation, the contribution to the volatility of credit
reallocation of the idiosyncratic (firm-level) credit changes grew relative to the contribution of
sectoral and aggregate shocks. Yet, the most interesting finding probably pertains to the cyclical
behavior. Credit reallocation exhibited a procyclical behavior throughout the 1981–2012 period.
However, while prior to the crisis this procyclical behavior was especially driven by credit growth,
after the crisis the comovement with the business cycle was especially driven by the excess credit
reallocation (while credit growth became essentially acyclical).

4

In the last part of the paper, we gather preliminary evidence on whether the higher dynamism
of the credit market in reallocating liquidity was associated with an improvement in the efficiency
of the reallocation process. To this end, we construct an index of credit reallocation efficiency
employing firms’ sales to capital and profits to capital ratios. We uncover evidence that the intensification of credit reallocation after the crisis was associated with enhanced efficiency in the credit
reallocation process.
The remainder of the paper unfolds as follows. Section 2 relates the analysis to prior literature.
Section 3 describes the reforms of the corporate and financial sectors that we expect to have affected credit reallocation. Section 4 describes the data and the empirical methodology. Section 5
investigates the intensity of credit reallocation before and after the financial crisis. Section 6 explores the role of flight to quality episodes, while Section 7 focuses on the time series properties of
credit reallocation. Section 8 investigates the efficiency of the credit reallocation process. Section
9 concludes.

1.2 Prior Literature
This paper relates to two strands of empirical literature. The first strand investigates the interaction between the credit market and the business cycle. Claessens, Kose and Terrones (2012) explore
the interplay between business cycles and financial cycles using aggregate data for advanced and
emerging countries. Mendoza and Terrones (2012) study the anatomy of credit booms and busts in
a large set of emerging countries. Bordo and Haubrich (2010) offer a detailed historical account of
the behavior of money and credit aggregates during recessions. In this strand of literature, a number of studies focus on the “flight to quality” episodes that can occur during recessions. Kashyap,
Stein and Wilcox (1993) document an increase in commercial paper relative to bank loans during
downturns. Lang and Nakamura (1995) and Oliner and Rudebush (1995) provide evidence of a
reshuffling of bank credit from small to large firms after monetary contractions. Only recently few
studies have started to analyze the continuous process of reallocation of funds that occurs in the
credit market. Dell’Ariccia and Garibaldi (2005) study the process of reallocation of loans across
5

U.S. banks. Herrera et al. (2011) document stylized facts of the process of reallocation of credit
across U.S. firms using Compustat data. Neither paper studies the role of financial crises in credit
reallocation and how the process of credit reallocation relates to credit booms and to deleveraging
processes.
The second related strand of literature analyzes the allocation of financial resources prior to
and after financial crises. Using Chilean data, Chen and Irarrazabal (2012) show that a reduction
in resource misallocation after the 1982 financial crisis led to a growth in total factor productivity.
Studying the 2002 Argentine crisis, Neumeyer and Sandleris (2010) uncover evidence that financial
crises can instead increase the misallocation of resources. Midrigan and Xu (2013) and Gilchrist,
Sim and Zakrajsek (2013) examine the effects of financial constraints on total factor productivity
losses using Korean establishment-level data and U.S. firm-level data, respectively.
Some microeconometric studies examine how financial institutions allocate loans after crises
and the associated reforms. Borensztein and Lee (2002, 2005) find that in Korea credit was not
directed to profitable sectors in the 1970-1996 period, whereas profitability was important for
maintaining access to credit during the 1997 financial crisis. Dunchin, Ozbas and Sensoy (2010)
uncover evidence that in the United States the credit crisis of 2008 restrained the supply of external finance to profitable projects. Using Indonesian data, Blalock, Gertler and Levine (2008)
show that foreign owned firms less vulnerable to liquidity constraints fared better than domestically owned firms during the East Asian crisis. Korajczyk and Levy (2003) and Levy and Hennessy
(2007) demonstrate that financially constrained firms and unconstrained firms make different capital structure choices in response to business fluctuations.

1.3 Crisis, Reforms, and the Credit Market
The South Korean economy experienced sustained output growth over the 1981 – 2012 period,
with the real GDP increasing at an average annual rate of 6.6%. At the end of 1997 and beginning
of 1998, a major financial crisis hit the economy (the GDP dropped by 5.7% in 1998). It is often
maintained that excessive investment and poor allocation of capital, labor and financial resources
6

made the economy vulnerable to the crisis (Park and Lee, 2003; Joh, 2003; World Bank, 2000). In
response to the crisis, in 1998 and 1999, the government engaged in deep reforms of the corporate
and financial sectors. This section describes salient aspects of the reforms which can have affected
the flexibility and dynamism with which credit is reallocated across firms.

1.3.1

Corporate Reforms

Prior to the crisis, Korean non-financial businesses expanded by relying heavily on bank loans
and bonds. Firms affiliated to business groups (chaebols ) benefited from the government’s corporate policy that encouraged the growth of chaebols in the belief that large-scale firms would better
compete in global markets (Borensztein and Lee, 2005). In 1995, the top 30 chaebols accounted
for 16.2% of the Gross National Product and 41.0% of the value added of the manufacturing sector. In 1997, the median debt-equity ratio of chaebol -affiliated firms was almost 400% (Lee and
Rhee, 2007). Debt overhang allegedly caused inefficient investments.3 Joh (2003) documents that

chaebols suffered from low productivity and return on equity.
After the onset of the crisis, the government enacted a reform of the corporate sector. Chaebol affiliated firms were forced to lower their debt-equity ratio below 200% by 1999: the debt-equity
ratio of the top 30 chaebols dropped to 171.2% in 2000. Debt guarantees among chaebol affiliates
were abolished: the debt guarantees of the top 30 chaebols dropped from 26.9 trillion won in April
1998 to zero in March 2000 (Chang, 2006). Along with the reform of chaebols, an unprecedented
amount of loans poured into supporting new small and medium-sized firms. At the peak of the
venture business boom in 1999, the ratio of venture companies’ value added to the GDP reached
2%.
3 Kim and Maksimovic (1990) document that higher debt decreases the efficiency of input allo-

cation.

7

1.3.2

Financial Reforms

Prior to the crisis, preferential credit was given to large firms to develop key manufacturing
industries.4 Enjoying little independence in monitoring firms, banks often engaged in a mere
renewal of outstanding loans (Haggard, Lim and Lim, 2010).
After the onset of the crisis, new financial supervision criteria, such as capital adequacy regulation and loan classification standards, were introduced to restrain over-investment and enhance
efficiency in the allocation of liquidity. This allegedly altered lending practices. Financial institutions stopped rolling over loans to companies with high debt and increasingly subjected firms
to loan appraisals (Berger, Clarke, Cull, Klapper and Udell, 2005). To obtain funds, large firms
(including chaebol -affiliated ones) increasingly turned to capital markets, which, in turn, became
more sensitive to firms’ profitability and default risk (Borensztein and Lee, 2002).5

1.3.3

The Effects on the Credit Market

It is commonly agreed that for various years the corporate and financial reforms exacerbated
the deleveraging process of the business sector initiated by the crisis (Bank of Korea, 2003). On
the demand side of the credit market, the reforms prompted the corporate sector to maintain a high
level of liquidity; on the supply side, they forced financial institutions to apply less inertial lending
standards. Figure 1.1 plots the real debt growth rate and real equity issues of Korean non-financial
firms, together with the real GDP growth rate, over the 1981–2012 period. The two financial
variables are constructed using the Flow of Funds Accounts compiled by the Bank of Korea. The
figure clearly illustrates the rapid credit growth before the financial crisis and the credit contraction
after the crisis.6 From the end of 2001, credit to the corporate sector started to increase again, but
4 Furthermore, the liberalization of financial markets - accelerated since 1993 - enabled firms to

borrow from non-bank financial institutions and foreign lenders (Chang, 2006).
5 Banker, Chang and Lee (2010) show that, thanks also to the banking sector reforms, the technical efficiency of the banking sector improved after the financial crisis.
6 In aggregate, the slowdown in the growth of credit to the business sector was partially compensated by an acceleration in loans to households (according to the Bank of Korea, the ratio of
corporate loans to total loans shrank from 76% in 1997 to 55% in 2002).

8

at a very slow pace. It was only in 2006 that credit growth accelerated.
While insightful, conventional credit aggregates are silent on the dynamic process of reallocation of credit across firms. Thus, they do not allow to discern whether the financial crisis and the
associated policy reforms had indeed an impact on the dynamism with which credit was reallocated
across firms.

1.4 Data and Methodology
In this section, we describe the data and the methodology used to measure credit reallocation.

1.4.1

The Data Set

To measure inter-firm credit reallocation, we need microeconomic, firm-level data. Our main
data source is KISLINE, the business information source of the leading Korean credit rating
agency, Korea Investors Service (KIS), which is affiliated with Moody’s. KISLINE provides information on financial statements, public disclosures and corporate governance of Korean businesses.
Our data set covers all the publicly traded firms as well as all the privately held firms subject to
annual external auditing. The 1981 Corporate External Audit Law requires all privately held companies whose assets are above a given level and all publicly traded firms to report their annual
external audit (including financial statements) to financial authorities. Between 1980 and 2012, to
reflect the inflation rate, the asset threshold for privately held firms was raised four times and since
2009 it has been 10 billion won. The coverage of KISLINE implies that our data set covers the
whole period in which a sample company exists. For instance, if a firm was subject to external
auditing only in 2006, our data set would include information about it for all the years in which the
firm was operational during the sample period.
The data set spans 33 years, from 1980 to 2012, and includes 33,463 firms (2,245 publicly
traded firms, 31,218 privately held ones) and 373,685 firm-year observations. We exclude financial firms because we aim at studying the demand side of the credit market. The firms in the data

9

set account for a large fraction of economic activity in Korea. They accounted for 49.2% and
56.6% of regular employment of the non-financial sector and of the manufacturing sector in 2000;
the bank loans they obtained amounted to 81.61% of the bank loans to all non-financial businesses
in 2008. By comparison, the Compustat firms used by Herrera et al. (2011) to document empirical
regularities of credit reallocation in the United States roughly account for one third of the employment of non-financial U.S. businesses. The average sales (total debt) of the privately held firms
and publicly traded firms in the sample are 297 million won (223 million won) and 4.6 billion won
(2.9 billion won), respectively.
The long sample period and the extensive coverage enable us to analyze the effects of the 1997
financial crisis on credit reallocation as well as various cross-sectional properties of credit reallocation. Additionally, our data make it possible to analyze separately the effects on the reallocation
of loans and bonds.

1.4.2

Measurement

Following Herrera et al. (2011), we define total debt as all forms of financial debt except accounts payable to suppliers. We exclude trade credit because it has properties very different from
other kinds of debt. It is for transaction purposes rather than for financial purposes; moreover, it is
based on relationships with suppliers rather than with financial institutions. Finally, trade credit is
very expensive and firms resort to it only when they do not have access to other forms of finance.
These features imply that trade credit has low substitutability with other forms of debt (Rajan and
Zingales 1995; Nilsen 2002). In addition to total credit, we investigate long-term credit, loans, and
bonds. Long-term credit is important because it frequently finances long-term investment plans.
Loans and bonds, in turn, may exhibit different dynamics.
This paper follows Herrera et al. (2011) in addressing a few methodological issues in the measurement of credit reallocation. A first issue regards firm entry and exit. The information provided
by KISLINE on firm inception and exit years enables us to distinguish between newborn firms
and firms that enter the data set but were already operational or were spinned off from other firms.

10

Likewise, we can distinguish between dying firms and firms that exit the data set but continue to
exist. Only when a firm exits due to bankruptcy, liquidation, or merger and acquisition, it is treated
as a dying firm. A second issue is the mismatch between fiscal year and calendar year that occurs
in roughly 5% of the firms in the sample. Following the way Compustat addresses this mismatch
problem, if the fiscal year ends after May 31st, the data of the firm are not reallocated as if there
was no mismatch problem. If, instead, the fiscal year ends before May 31st, the data are allocated
to the previous year. Alternatively, we address this issue by apportioning fiscal year data proportionally to calendar years; this leads to virtually identical results. Lastly, we deflate all the original
variables using the implicit GDP deflator in order to study credit reallocation in real terms and
relate its dynamics with that of real aggregate variables.
To measure credit reallocation, this paper replicates the methodology proposed by Davis and
Haltiwanger (1992) for measuring job reallocation and employed by Herrera et al. (2011) for measuring credit reallocation in the United States. Let c f t denote the average debt of firm f between
year t-1 and year t and Cst denote the average debt of set s of firms between year t-1 and year t.
The debt growth rate g f t of firm f is obtained by dividing the change in debt from year t-1 to year

t by c f t . This growth rate takes values in the [−2, +2] interval and has the advantages of symmetry and boundedness (for more on its statistical properties, see Davis and Haltiwanger, 1992, and
Törnqvist, Vartia and Vartia, 1985). If a firm is born its debt growth rate takes the value of +2; if it
dies, its debt growth rate takes the value of −2.
Five aggregate credit flows are constructed using these building blocks. Credit creation (POSst )
is the sum of the debt growth rates of the firms with growing debt weighted by their debt size (the
firm debt average over the subsample’s debt average). Credit destruction (NEGst ) is the sum of the
debt growth rates of the firms with shrinking debt weighted by their debt size. The third measure,
gross credit reallocation (SUMst ), is the sum of credit creation and credit destruction.7 Net credit
growth (NETst ) is constructed as credit creation less credit destruction. The last measure, excess
7 Since credit creation and destruction are generated with annual data, they do not capture

changes of credit during a year. Thus, they represent lower bounds on the true credit creation
and destruction.

11

credit reallocation (EXCst ), is computed as gross credit reallocation less the absolute value of the
net credit change. That is, EXCst measures credit reallocation in excess of the minimum required
to accommodate the net credit change. These five credit flows can be written as follows:
(

)
cft
POSst =
∑ g f t Cst
1 f ∈st g f t >0
( )
cft
NEGst =
∑ |g f t | Cst
1 f ∈s g <0

(1.1)

(1.2)

t ft

SUM st = POSst + NEGst

(1.3)

EXCst = SUM st − |NET st |

(1.4)

where NETst = POSst – NEGst .

1.5

Credit Reallocation, Credit Boom, Deleveraging

This section studies the intensity of credit reallocation before and after the crisis.

1.5.1

Intensity of Credit Reallocation

Panel A of Figure 1.2 plots gross credit reallocation, excess credit reallocation, and the net
credit change together with the real GDP growth rate for the period 1981 to 2012. Panel B of the
figure plots credit creation and destruction. Panel A of Table 1.1 shows the average flows of credit
for the same period and for the pre-crisis (1981-1996) and post-crisis (1999-2012) sub-periods. It
also shows average credit flows for two narrower sub-periods: the credit boom (1993-1996) and
the deleveraging phase (1999-2004). Over the 1981-2012 period, the net credit change equalled
6.9% on average. Between 1981 and 1996, credit grew at an annual rate of 9.6%. Credit growth
was especially rapid from the beginning of the 1990s and accelerated between 1993 and 1997,
originating a credit boom (average credit growth over 10.3%). After the crisis, credit growth
dropped dramatically, averaging -2.8% between 1999 and 2004 and 3.0% between 1999 and 2012.
12

Panel A: GDP, Debt, and Equity

10

40

0

10

GDP growth

−40 −20

0

20

Equity growth

Debt growth

1980:1

1985:1

1990:1

1995:1

2000:1

2005:1

2010:1 2012:1

8
6
4
2
0

0

2

4

6

8

Panel B: Debt / GDP

1990:1

1993:1

1996:1

1999:1

2002:1

2005:1

Figure 1.1: GDP and Business Sector Debt and Equity Panel A shows the real GDP growth rate
of South Korea and the real growth rate of the total debt and equity of Korean firms. The solid line
is the year -on-year quarterly growth rate of the real GDP (scale on the right Y-axis). The dashed
line and the dotted line represent the year-on-year quarterly real growth rate of the total
outstanding debt and total outstanding equity of Korean firms, respectively (scale on the left Yaxis). Debt consists of total loans from financial institutions and bonds issued. Debt and equity
data are from the Flow of Funds Accounts compiled by the Bank of Korea. Panel B shows the
aggregate leverage ratio (total debt/nominal GDP, solid spike) of Korean firms for the period
1990:1 to 2005:1. The shaded areas in the Panel A and B correspond to the financial crisis.

13

Panel A: Credit Reallocation

20

30

SUM

EXC

0

10

GDP growth

−10

NET

1981

1986

1991

1996

2001

2006

2012

2006

2012

30

Panel B: Credit Creation and Destruction

NEG

0

10

20

POS

1981

1986

1991

1996

2001

Figure 1.2: Credit Change and Credit Reallocation Panel A shows gross credit reallocation
(SUM, solid line), excess credit reallocation (EXC, dashed line), the net credit change (NET,
dotted line), and the annual real GDP growth rate (gray area). Panel B shows credit creation
(POS, solid line) and credit destruction (NEG, dashed line).The vertical shaded areas in the two
panels correspond to the financial crisis.

14

A deleveraging process can be achieved through a reduction in the rate of credit creation and
a relatively stable credit destruction, thus implying lower credit reallocation. Alternatively, it can
be attained through a relatively stable credit creation and an increase in credit destruction, thus
entailing higher credit reallocation. Korea followed the latter path. Over the whole sample period,
the average credit creation and destruction were 14.18% and 7.25%, respectively; the average
credit reallocation equalled 21.42%. Credit destruction surged significantly during the crisis and
thereafter remained permanently higher than in the pre-crisis period: the average credit destruction
was 4.16% before the crisis (1981-1996) and more than double (10.86%) after it (1999-2012).
Credit creation dropped during the crisis and reverted back to the pre-crisis level (about 14%) after
the crisis. As a result of these dynamic patterns of credit creation and destruction, gross credit
reallocation increased after the crisis, rising from an average of 17.89% in the pre-crisis (19811996) period (17% during the credit boom of 1993-1996) to an average of 24.70% in the post-crisis
(1999-2012) period (25% during the 1999-2004 deleveraging period).8 Figure 1.1 makes clear that
the tendency of the gross reallocation of credit to increase in the last phase of the credit boom was
only due to the need to accommodate the acceleration in credit growth. The behavior of the excess
credit reallocation is particularly telling in this respect: the average excess credit reallocation was
8.31% between 1981 and 1996, and it actually slowed down to 6.98% during the credit boom.
After the crisis, due to the significant increase in gross credit reallocation and the drop in the net
credit change, it rose to 18.99%. Altogether, the behavior of credit reallocation and the net credit
change reveal that the credit boom was characterized by a depressed excess reallocation of credit.
By contrast, the creditless recovery after the crisis was characterized by an intensification of the
reallocation of credit that has persisted since then.
We performed Chow tests to assess formally the presence of a structural break in the credit
flows in 1998. The results, shown in Table 1.1, suggest that there was a structural break in credit
destruction, net change, and excess reallocation (with a significance level of 5%) between the pre8 The magnitude of credit reallocation over the full sample is of the same order as that found by

Herrera et al. (2011) for U.S. non-financial businesses over the 1952-2007 period. However, the
net credit change is higher than that of the U.S. business sector.

15

crisis and the post-crisis period, while there was no structural break in credit creation. A concern
with Chow tests might be that they require to posit the year of structural break. To assuage this
possible concern, we followed the approach of Stock and Watson (2003). Specifically, we specified an AR(1) process for the conditional mean of credit reallocation. We then used the Quandt
Likelihood Ratio (QLR) test statistic, also known as the sup-Wald statistic, to test whether the
conditional mean of the AR(1) process had a structural break at some unknown date. We obtained
evidence of a structural break of the excess credit reallocation in 1999 (with a significance level of
1%). The 67% confidence interval for the break date is between 1997 and 2001.9 Furthermore, we
compare the mean of credit flows between pre-crisis period (1981-1996) and the post-crisis period
(1999-2012) using rank sum tests. The results, shown in Table 1.1, again suggest that the means of
credit flows except credit creation differ between the two periods (with a significance level of 5%)
An appealing feature of our data is that they allow to disentangle the behavior of loans and
bonds. Interestingly, after the crisis the inter-firm gross reallocation of loans rose more sharply
than the gross reallocation of bonds (see Table 1.1 and Figure 1.3). This stems from the fact that
both loan creation and loan destruction rose while the increase in bond destruction was approximately offset by the decrease in bond creation. To summarize, gross and excess credit reallocation
have significantly intensified after the crisis than before it, and this has especially been due to the
increase in the intensity of the reallocation of loans.

9 The Quandt test results are available from the authors.

16

Variable
Total credit

Period
POS
NEG
SUM
NET
EXC
81-12
14.177
7.246
21.423
6.930
13.216
81-96
13.730
4.157
17.887
9.573
8.314
93-96
13.768
3.491
17.258
10.277
6.981
99-04
11.189
13.856
25.045
-2.668
22.070
99-12
13.840
10.855
24.696
2.985
18.986
Chow test
0.196
2.315
1.899
2.570
3.983
Rank sum
0.748
-4.448
-4.115
2.827
-4.448
Long-term credit 81-12
18.255
11.090
29.346
7.165
20.222
81-96
17.754
8.087
25.841
9.667
15.929
93-96
17.051
6.395
23.446
10.655
12.790
99-04
15.791
18.343
34.134
-2.552
28.292
99-12
17.979
14.709
32.688
3.269
25.501
Chow test
0.050
6.160
1.099
3.429
6.110
Rank sum
0.042
-3.575
-3.035
2.245
-3.326
Loans
81-12
18.946
12.682
31.628
6.263
22.000
81-96
16.481
8.185
24.666
8.296
15.477
93-96
17.089
7.344
24.433
9.745
14.688
99-04
16.547
22.497
39.045
-5.950
31.652
99-12
20.639
18.090
38.730
2.549
29.954
Chow test
-2.536
-3.908
-4.656
1.538
-4.282
Rank sum
-2.536
-3.908
-4.656
1.538
-4.282
Bonds
81-12
22.066
11.718
33.784
10.349
21.040
81-96
25.146
9.501
34.648
15.645
18.912
93-96
19.737
5.224
24.961
14.513
10.448
99-04
16.413
18.108
34.521
-1.695
25.691
99-12
18.238
14.549
32.787
3.688
24.086
Chow test
3.184
2.262
0.648
3.961
1.239
Rank sum
2.993
-2.702
0.831
3.118
-1.912
Jobs
81-12
8.322
6.760
15.082
1.562
12.243
81-96
7.964
5.678
13.642
2.287
10.707
93-96
6.912
5.178
12.089
1.734
9.990
99-04
9.408
9.935
19.343
-0.527
16.322
99-12
8.868
7.787
16.656
1.081
13.919
Chow test
1.331
0.301
1.113
4.640
1.839
Rank sum
-2.37
-2.702
-2.889
0.416
-3.118
Notes: Panel A reports the average flows of total credit, long-term credit, loans, bonds,
and jobs. The period 1981 to 1996 and the period 1999 to 2012 reflect the pre-crisis
period and the post-crisis one, respectively.

Table 1.1: Magnitude of Gross Flows

1.5.2 Size and Persistence of Credit Changes
An extensive literature demonstrates that, because of non-convex adjustment costs, businesses
prefer adjusting labor and capital in a lumpy way (see, e.g., Davis, Faberman and Haltiwanger,
2006). While it is unclear to what extent a similar argument holds for credit, recent studies suggest
the presence of non-convex adjustment costs in credit changes (Eisfeldt and Muir, 2013; Bazdresch, 2013). It is then important to investigate to what degree the intensification of credit reallocation after the crisis was driven by large credit changes. Following Gourio and Kashyap (2007)
and Herrera et al. (2011), we sort firm credit changes into three groups. If a firm’s debt growth

17

40

50

Panel A: Loan Reallocation

30

SUM

20

EXC
NET

−10

0

10

GDP growth

1981

1986

1991

1996

2001

2006

2012

40

50

Panel B: Bond Reallocation

20

30

SUM

EXC

NET

−10

0

10

GDP growth

1981

1986

1991

1996

2001

2006

2012

Figure 1.3: Loan and Bond Reallocation Panel A shows loan reallocation (SUM, solid line),
excess credit reallocation (EXC, dashed line), the net credit change (NET, dotted line), and the
annual real GDP growth rate (gray area). Panel B shows bond reallocation (SUM, solid line),
excess credit reallocation (EXC, dashed line), the net credit change (NET, dotted line), and the
annual real GDP growth rate (gray area). The vertical shaded areas in the two panels correspond
to the financial crisis.

18

rate exceeds 18% or is below −18%, then this is labeled as a large credit increase and decrease, respectively. Next, using the methodology above, we calculate the credit creation due to large credit
increases (POSbigst ) and the credit destruction due to large credit decreases (NEGbigst ). Based
on these two measures, we then compute the gross and excess credit reallocation (SUMbigst and
EXCbigst ) and the net credit growth (NETbigst ) due to large credit changes. Table 1.2 and Figure
1.4 display the credit flows attributable to large credit changes; numbers in parentheses are the
shares of credit flows due to large changes. On average, between 1981 and 2012 the share of credit
reallocation due to large credit changes equalled 76.80%. This share actually rose from 72.09%
before the crisis to 80.94% after the crisis. Thus, a substantial portion of the increase in credit
reallocation is attributable to large credit adjustments.
Credit changes could be large but temporary, thus reflecting short-lived liquidity shortfalls. To
check whether the increase in the intensity of credit reallocation after the crisis was driven by
temporary debt changes, we assess the persistence of the debt changes using the index proposed
by Davis and Haltiwanger (1992)
)
(
[
g f t,t+2 ]
Pf t = min 1, max 0,
g f t,t+1

(1.5)

where g f t ,t + 2 and g f t ,t + 1 are the debt growth rate between year t and year t+2 and the debt growth
rate between year t and year t+1, respectively. The maximum persistence, equivalent to Pf t =1,
occurs when all the debt change of a firm f from t to t+1 lasts until t+2 ; Pf t =0 means instead that
the debt change of a firm f is purely temporary. In the full sample period, the unweighted average
value of Pf t was 0.71. This value did not change after the crisis (equalling on average 0.72 before
the crisis, 0.70 after it). This indicates that the firm-level debt changes underlying the credit flows
were persistent both before and after the crisis and that the intensification of credit reallocation
after the crisis was not due to temporary liquidity shortfalls.

19

1.5.3

Comparison with job flows

It is useful to compare the intensity of credit reallocation with that of job reallocation. We
consider regular employment, which includes permanent and temporary employment, whose labor
contracts are longer than 1 year. Credit reallocation is more intense than job reallocation in Korea,
consistent with what found for the United States by Herrera et al. (2011). Table 1.1 shows that the
job reallocation rate rose from 13.64% in the pre-crisis period to 16.66% in the post-crisis period.
Interestingly, both job creation and job destruction rose after the crisis.

1.6 The Role of “Flights to Quality”
The macroeconomics literature on credit market imperfections argues that, following negative
aggregate shocks, financiers contract credit to information opaque borrowers, such as small firms,
while they accommodate the increasing credit demand of information transparent borrowers, such
as big firms (Bernanke, Gertler and Gilchrist, 1999). This would induce a reshuffling of credit (a
“flight to quality”) from small to big firms. Similarly, following negative aggregate shocks credit
can flow from industries suffering from tight credit conditions to industries less exposed to tight
credit. The reader may then wonder to what extent the intensification of credit reallocation we
have uncovered reflected a flight to quality triggered by the financial crisis and that persisted after
the crisis.
To probe this point, we break down our sample based on four group categories: size classes,
2-digit manufacturing industries, chaebol -affiliation, locations, and whether listed or not. We measure the reallocation of credit within a group using the following index put forth by Davis and
Haltiwanger (1992)

Wt = 1 −

∑ NET jt
∑ SUM jt

(1.6)

where j denotes a group. Wt =1 if credit reallocation across groups does not occur and all the
reallocation is within groups; Wt = 0 if reallocation within groups does not occur and all credit

20

Variable
Total credit

Long-term credit

Loans

Bonds

Jobs

Period
81-12

POSbig
11.507

NEGbig
5.250

SUMbig
16.757

NETbig
6.257

EXCbig
9.676

(80.006)

(66.988)

(76.800)

(76.940)

(67.939)

81-96

10.693

2.304

12.997

8.389

4.608

(77.166)

(55.325)

(72.093)

(88.193)

(55.325)

93-96

10.331

1.957

12.288

8.374

3.914

(73.809)

(57.379)

(70.195)

(81.541)

(57.379)

99-04

9.258

11.233

20.492

-1.975

18.300

(82.358)

(79.545)

(80.765)

(56.968)

(82.564)

99-12

11.674

8.533

20.206

3.141

15.235

(81.945)

(79.325)

(80.942)

(67.035)

(80.759)

81-12

16.160

9.122

25.282

7.038

16.585

(87.571)

(78.894)

(85.090)

(91.838)

(79.082)

81-96

15.496

5.959

21.455

9.536

11.700

(86.459)

(71.228)

(82.116)

(99.966)

(71.244)

93-96

14.395

4.608

19.003

9.786

9.217

(83.934)

(72.411)

(80.742)

(91.977)

(72.411)

99-04

13.791

16.601

30.392

-2.810

24.934

(88.774)

(87.471)

(88.055)

(84.874)

(88.432)

99-12

15.984

12.816

28.800

3.168

22.088

(88.134)

(86.696)

(87.648)

(81.108)

(87.108)

81-12

16.968

10.794

27.763

6.174

18.536

(88.670)

(80.229)

(86.507)

(91.866)

(80.227)

81-96

14.134

6.078

20.211

8.056

11.418

(85.357)

(71.650)

(81.648)

(84.919)

(71.661)

93-96

14.475

5.218

19.693

9.257

10.436

(84.277)

(70.030)

(80.470)

(117.097)

(70.030)

99-04

15.735

20.630

36.365

-4.895

30.042

(91.551)

(91.511)

(91.508)

(88.066)

(91.830)

99-12

19.589

16.082

35.672

3.507

27.357

(91.992)

(89.885)

(91.352)

(99.250)

(90.021)

81-12

19.920

10.109

30.028

9.811

17.714

(88.347)

(82.598)

(87.785)

(100.028)

(81.116)

81-96

23.070

7.872

30.942

15.198

15.635

(90.645)

(78.196)

(88.121)

(93.594)

(78.145)

93-96

17.171

3.296

20.467

13.875

6.592

(87.159)

(64.246)

(81.943)

(97.441)

(64.246)

99-04

11.342

17.832

29.173

-6.490

22.684

(83.337)

(91.537)

(88.471)

(141.954)

(83.337)

99-12

14.744

13.207

27.951

1.536

20.817

(84.813)

(87.543)

(86.678)

(107.728)

(84.214)

81-12

6.189

5.012

11.201

1.177

9.007

(72.505)

(72.341)

(72.425)

(81.131)

(71.854)

81-96

5.911

4.015

9.926

1.896

7.413

(71.277)

(68.946)

(70.507)

(93.156)

(67.402)

93-96

4.243

3.249

7.493

0.994

6.078

(61.320)

(62.640)

(62.020)

(78.305)

(61.005)

99-04

8.263

6.981

15.244

1.282

13.962

(78.483)

(77.768)

(78.122)

(86.406)

(77.768)

99-12

6.855

5.521

12.376

1.334

11.042

(73.099)

(76.040)

(74.245)

(69.336)

(76.040)

Notes: This table reports the average flows due to large changes. Numbers in parentheses
indicate the shares of total flows due to large changes. The period 1981 to 1996 and the
period 1999 to 2012 reflect the pre-crisis period and the post-crisis one, respectively.

Table 1.2: Average Flows due to Large Changes

21

reallocation occurs across groups.
Table 1.3 displays the results when we partition the sample into size classes (sales quintiles).
Credit reallocation decreases monotonically with size. For instance, the credit reallocation rates
for the 1st sale quintile before and after the crisis were 23.65 and 33.38, respectively, markedly
larger than 16.11 and 21.86 for the 5th quintile.10 The average Wt was 0.57 and rose from 0.44
in the pre-crisis (1981-1996) period to 0.73 in the post-crisis (1999-2012) period (see Panel B). In
unreported tables, we also partition manufacturing firms into 24 two-digit SIC industries. Credit
reallocation exhibits considerably across industries. With the exception of a few manufacturing
industries, such as electronic components and motor vehicles industries, in all industries credit
reallocation increased after the crisis, fueled mainly by an increase in credit destruction. The
average Wt was 0.52 between 1981 and 2012, and, again, rose significantly from 0.43 in the
pre-crisis period to 0.63 in the post-crisis period. Next, we partition the sample into 16 regions
based on the Korean administrative districts (7 metropolitan cities and 9 provinces). We identify
a firm’s location using the headquarter address reported by KISLINE. The average Wt rose from
0.40 in the pre-crisis period to 0.67 in the post-crisis period, suggesting that the share of credit
reallocation across regions shrank after the crisis. Finally, a classification relevant for examining
credit reallocation in South Korea is that between chaebol and non-chaebol firms. When we split
firms based on whether they are affiliated or not to one of the top 30 chaebols (defined according to
the classification of the Korean government), we obtain again that the average Wt rose significantly
after the crisis (0.47 to 0.75).
Altogether, these results suggest that the importance of credit reallocation within relatively homogeneous groups of firms increased after the crisis, while that of the reallocation across groups
dropped. Does this imply that no flight to quality occurred during the crisis? Actually, the sharp
increase in the W-index occurred after a substantial drop during the crisis (see Table 1.3 for sales
quintiles). Such a drop of the index suggests that the crisis was indeed characterized by a reshuffling of credit from risky and informationally opaque small firms to safer and informationally
10 However, such a monotonic pattern cannot be observed for loans and bonds separately.

22

Panel A: Gross Credit Reallocation

20

30

SUM

10

SUM (Large changes)

0

SUM (Small changes)

1981

1986

1991

1996

2001

2006

2012

30

Panel B: Excess Credit Reallocation

20

EXC

10

EXC (Large changes)

0

EXC (Small changes)

1981

1986

1991

1996

2001

2006

2012

Figure 1.4: Large Credit Flows Panel A of this figure shows gross credit reallocation (SUM, solid
line), gross credit reallocation due to large changes (dashed line) and to small changes (dotted line)
for the period 1981 to 2012. Panel B of this figure shows excess credit reallocation (EXC, solid
line), excess credit reallocation due to large changes (dashed line) and to small changes (dotted
line) for the period 1981 to 2012. The vertical shaded areas in the two panels correspond to the
financial crisis.
transparent large firms. Nonetheless, the increase of the W-index after the crisis strongly suggests
that flights to quality among relatively homogeneous classes of firms cannot explain the significant,
persistent intensification of credit reallocation after the crisis.

23

Total Credit
Loans
Bonds
Panel A: Credit reallocation in sales quintiles
Quintile Period
POS
NEG
SUM
NET
EXC
POS
NEG
SUM
NET
EXC
POS
NEG
SUM
NET
EXC
1st
81-12
19.981 7.905
27.885 12.076 15.809 23.281 10.838 34.119 12.443 20.397 41.767 29.185 70.952 12.583 40.988
81-96
18.506 5.148
23.654 13.359 10.296 21.735 8.807
30.542 12.928 15.058 39.072 17.241 56.314 21.831 30.734
93-96
15.381 5.607
20.988 9.775
11.214 18.615 7.293
25.908 11.322 14.586 33.995 22.439 56.434 11.556 43.823
99-04
24.172 11.057 35.229 13.115 22.114 27.950 12.937 40.887 15.013 25.874 54.449 47.270 101.719 7.178
49.169
99-12
22.059 11.321 33.380 10.738 22.642 25.488 13.377 38.866 12.111 26.754 47.748 40.108 87.856 7.641
52.430
2nd
81-12
17.435 7.510
24.944 9.925
14.809 21.532 11.161 32.692 10.371 20.541 36.148 22.724 58.872 13.424 36.144
81-96
16.636 5.009
21.645 11.627 9.905
19.989 9.459
29.448 10.529 15.857 34.171 14.466 48.636 19.705 26.352
93-96
16.505 5.009
21.514 11.496 10.017 22.267 8.421
30.687 13.846 16.841 27.082 15.720 42.802 11.363 31.439
99-04
17.408 13.098 30.506 4.309
25.382 20.366 15.838 36.204 4.528
30.344 49.382 36.554 85.935 12.828 56.916
99-12
18.380 10.314 28.694 8.066
20.279 23.059 13.273 36.332 9.786
25.974 41.310 30.528 71.838 10.782 47.974
3rd
81-12
15.125 7.263
22.388 7.863
13.916 20.962 12.027 32.989 8.935
21.975 24.624 17.331 41.955 7.293
26.185
81-96
15.325 4.710
20.035 10.616 9.419
18.832 9.603
28.435 9.229
16.375 30.782 10.856 41.638 19.925 20.810
93-96
14.587 3.571
18.158 11.016 7.142
20.800 6.754
27.554 14.046 13.507 20.901 6.066
26.967 14.835 12.133
99-04
13.491 11.855 25.346 1.637
22.867 22.569 20.288 42.856 2.281
37.156 17.191 26.434 43.626 -9.243 30.523
99-12
14.508 10.057 24.565 4.451
19.073 22.653 14.870 37.523 7.783
28.275 18.859 24.919 43.777 -6.060 33.430
4th
81-12
16.423 6.669
23.092 9.754
13.164 20.497 11.071 31.568 9.426
20.901 27.672 16.386 44.058 11.286 27.834
81-96
16.785 4.228
21.013 12.557 8.456
19.542 8.574
28.116 10.968 15.206 25.531 9.416
34.947 16.115 18.695
93-96
18.982 2.546
21.528 16.436 5.092
23.552 5.490
29.042 18.062 10.979 17.271 6.941
24.212 10.329 13.883
99-04
12.802 11.569 24.371 1.233
22.210 19.174 17.474 36.648 1.700
33.512 18.999 20.990 39.989 -1.991 22.018
99-12
15.690 9.518
25.208 6.172
18.639 21.528 14.226 35.754 7.302
27.836 30.303 24.291 54.595 6.012
37.452
5th
81-12
13.017 6.183
19.200 6.834
10.733 17.819 12.284 30.103 5.534
19.472 21.391 10.721 32.111 10.670 19.204
81-96
12.639 3.472
16.111 9.168
6.943
15.443 7.502
22.944 7.941
13.793 24.282 9.114
33.396 15.168 18.228
93-96
12.241 2.392
14.633 9.850
4.784
14.941 6.981
21.923 7.960
13.206 19.161 4.317
23.478 14.844 8.634
99-04
9.452
12.902 22.354 -3.450 18.033 15.161 22.902 38.063 -7.740 27.809 12.236 18.137 30.373 -5.900 24.472
99-12
12.638 9.221
21.859 3.418
15.022 19.864 17.640 37.505 2.224
26.859 16.206 13.343 29.549 2.864
21.572
Panel B: W Indexes based on sub-groups
Size
Industry
Chaebol Affiliation
Region
Listing
Total
Loans
Bonds Total
Loans
Bonds Total
Loans
Bonds Total
Loans
Bonds Total
Loans
Bonds
81-12
0.572
0.629
0.610
0.519
0.555
0.435
0.596
0.672
0.606
0.524
0.575
0.444
0.616
0.684
0.658
81-96
0.444
0.544
0.551
0.428
0.503
0.380
0.472
0.624
0.514
0.402
0.471
0.363
0.504
0.619
0.573
93-96
0.400
0.514
0.633
0.490
0.559
0.402
0.419
0.604
0.428
0.427
0.536
0.406
0.419
0.534
0.534
1997
0.227
0.307
0.693
0.255
0.245
0.407
0.174
0.191
0.285
0.240
0.285
0.462
0.228
0.243
0.397
1998
0.821
0.807
0.516
0.691
0.648
0.493
0.918
0.717
0.497
0.703
0.679
0.434
0.929
0.770
0.658
99-04
0.806
0.796
0.620
0.645
0.635
0.476
0.826
0.799
0.759
0.710
0.728
0.471
0.751
0.808
0.702
99-12
0.725
0.736
0.677
0.630
0.630
0.497
0.745
0.759
0.742
0.670
0.707
0.535
0.749
0.784
0.772
Notes: The table shows the average credit flows in sales quintiles (Panel A) and the W indexes for four firm classifications (size, industry, chaebol affiliation, region and listing) (Panel
B). In Panel A, the first (fifth) quintile is the quintile with the smallest (largest) firms. The Chow test statistics come from Chow tests for a structural break in 1997.

Table 1.3: Credit Reallocation in Sub-Groups

24

1.7 The Dynamic Pattern of Credit Reallocation
In the previous section, we found that the intensity of credit reallocation rose significantly
after the 1997 financial crisis. We now turn to examine whether the dynamic pattern of credit
reallocation also changed after the crisis.

1.7.1

Volatility

Table 1.4, Panel A, reports three measures of volatility of credit flows: the standard deviations
of the original flows and of the Hodrick-Prescott filtered flows as well as the coefficient of variation (standard deviation/mean*100) of the flows. Although credit reallocation turns out to be less
volatile than in the United States (see Herrera et al., 2011), its volatility is high: over the full sample
period the coefficients of variation equal 31.73% for credit creation, 59.96% for credit destruction,
22.64% for gross credit reallocation, and 49.96% for excess credit reallocation. Similar to what
found for the United States, credit destruction is more volatile than credit creation. The volatilities of credit creation, credit destruction, net credit change, gross and excess credit reallocation
consistently increased after the crisis. We also computed the rolling standard deviations of credit
reallocation using 5 year and 10 year moving windows. The results confirm that the volatility of
credit reallocation rose after the financial crisis.
As noted, the relative importance of credit reallocation within industries and size classes grew
after the crisis. A related question is to what extent the increase in the volatility of credit reallocation was driven by idiosyncratic, firm-level debt changes. To probe this point, we decompose the
debt growth rate of each firm into the sector growth rate and an idiosyncratic component. Next,
we recompute credit flows using only the idiosyncratic component. Finally, we decompose the
variance of credit reallocation into three parts, the variance caused by idiosyncratic effects, the
variance caused by sectoral or aggregate effects, and the covariance term,
var(SUMt ) = var(SUMti ) + var(SUMt − SUMti ) + 2cov(SUMt − SUMti , SUMti ),

(1.7)

where SUMt i denotes credit reallocation driven by idiosyncratic effects in year t. Table 1.5 sum25

marizes the relative contribution of idiosyncratic effects and of sectoral or aggregate effects to
the variance of credit flows. Consistently across classification schemes, we find that the relative
importance of sectoral or aggregate effects in the volatility of credit reallocation tended to shrink
after the crisis while the importance of idiosyncratic effects rose. Using the classification in size
classes, for example, we obtain that after the crisis the variance of the reallocation of loans explained by idiosyncratic effects amounted to almost 91%, versus 60% explained by sectoral or
aggregate effects.

1.7.2

Cyclical Behavior

To examine the cyclical patterns of credit flows, we start by computing unconditional correlation coefficients between the credit flows and the GDP.

1.7.2.1

Unconditional Correlation

We extract cyclical components from the series using the Hodrick-Prescott filter. Table 1.4,
Panel B, gathers the pairwise coefficients of correlation between the cyclical components of the
credit flows and the cyclical components of real GDP; coefficients significant at the 5% level are
in bold. Since 2008 appears to be an outlier, driven by an economic crisis and a program of credit
subsidies, we also present correlation coefficients excluding 2008. Over the 1981-2012 period,
credit creation was procyclical, while credit destruction was countercyclical. Gross credit reallocation exhibited a mildly procyclical pattern, and this procyclical behavior was present both before
and (to a lesser extent) after the crisis.11 Interestingly, however, when we consider separately the
components of gross credit reallocation (the absolute value of the net credit change and the excess credit reallocation − see formula (1.4)), we find that the forces driving the cyclical behavior
of gross credit reallocation changed after the crisis (see again Table 1.4, Panel B). While before
the crisis credit growth exhibited a procyclical behavior and the excess credit reallocation was es11 Herrera et al. (2011) also find a mildly procyclical behavior of credit reallocation for the

United States.

26

Panel A: Volatility
Total Credit
SUM
NET EXC
POS
s.d.
81-12 4.499 4.345 4.850 7.397 6.602
6.126
81-96 2.387 1.208 2.074 3.164 2.416
3.916
99-12 5.406 3.807 4.364 8.271 4.835
6.372
s.d. of H.P filtered flow
81-12 3.587 1.921 2.710 5.076 3.122
4.440
81-96 2.065 1.085 1.799 2.765 2.038
3.376
99-12 4.095 2.237 3.383 5.666 3.445
4.312
s.d./mean*100
81-12 31.733 59.964 22.639 106.733 49.957
32.333
81-96 17.385 29.058 11.594 33.053 29.058
23.758
99-12 39.266 109.079 25.284 80.475 69.266
37.284
Panel B: Unconditional correlation of credit flows with GDP growth rate
1981-2012
t-2
t-1
t
t+1
t+2
t-2
POS

NEG

Loans
SUM NET
s.d.
7.067 7.918 10.594
4.074 3.205 7.321
5.761 4.118 11.429
s.d. of H.P filtered flow
3.955 2.954 7.873
3.467 2.896 6.201
3.818 3.155 7.510
s.d./mean*100
55.722 25.035 169.149
49.781 12.995 88.239
78.445 16.853 117.279
NEG

EXC

POS

9.573
6.120
5.565

8.306
7.454
6.560

5.125
5.078
4.880

5.730
5.993
4.868

43.514
39.541
37.888

37.640
29.642
33.237

Bonds
SUM
NET
s.d.
5.807 7.698 12.089
5.461 8.230 10.151
5.036 6.921 9.428
s.d. of H.P filtered flow
3.247 5.045 7.830
3.798 4.699 8.866
2.655 5.195 5.874
s.d./mean*100
49.557 22.787 116.818
57.475 23.753 64.882
96.399 27.729 64.958
NEG

EXC
9.699
10.699
8.027
6.739
7.367
6.523
46.100
56.575
76.830

1981-1996
1999-2012
t-1
t
t+1
t+2
t-2
t-1
t
t+1
t+2
Panel B-1: Total credit
POS 0.005 0.298 0.173 -0.695* 0.185
-0.010 0.044 0.219 -0.320 -0.168
-0.068 0.413 -0.079 -0.668* 0.362
NEG 0.248 -0.364* -0.202 0.518* -0.029
0.134 -0.250 -0.170 -0.205 0.597*
0.426 -0.449 0.102 0.460 -0.444
SUM 0.183 0.137 0.086 -0.552* 0.225
0.069 -0.100 0.149 -0.491 0.168
0.199 0.203 -0.028 -0.504 0.142
NET -0.090 0.349 0.198 -0.687* 0.142
-0.060 0.131 0.230 -0.159 -0.359
-0.218 0.475 -0.097 -0.664* 0.437
EXC 0.037 -0.211 -0.208 0.536* 0.016
0.156 -0.230 -0.150 -0.206 0.548*
0.051 -0.217 0.115 0.549* -0.363
Panel B-2: Total credit (Excluding the year 2008)
POS -0.038 0.304 0.264 -0.718* 0.089
-0.010 0.044 0.219 -0.320 -0.168
-0.242 0.520 0.144 -0.527 -0.152
NEG 0.285 -0.357* -0.241 0.489* 0.040
0.134 -0.250 -0.170 -0.205 0.597*
0.538 -0.447 0.005 0.301 -0.251
SUM 0.174 0.109 0.148 -0.535* 0.144
0.069 -0.100 0.149 -0.491 0.168
0.176 0.165 0.145 -0.284 -0.341
NET -0.146 0.358* 0.282 -0.696* 0.044
-0.060 0.131 0.230 -0.159 -0.359
-0.450 0.589* 0.095 -0.516 0.032
EXC 0.068 -0.194 -0.260 0.507* 0.104
0.156 -0.230 -0.150 -0.206 0.548*
0.136 -0.182 -0.009 0.373 -0.057
Panel B-3: Loans
POS -0.111 0.293 0.239 -0.663* 0.069
0.018 -0.014 0.064 -0.080 -0.312
-0.321 0.506 0.017 -0.725* 0.348
NEG 0.285 -0.365* -0.257 0.598* 0.060
0.075 -0.394 -0.002 0.019 0.528*
0.586* -0.412 -0.134 0.810* -0.523
SUM 0.214 -0.048 0.014 -0.195 0.183
0.111 -0.488 0.072 -0.071 0.269
0.270 0.193 -0.138 -0.011 -0.157
NET -0.206 0.348 0.264 -0.674* 0.009
-0.032 0.213 0.036 -0.054 -0.465
-0.482 0.500 0.078 -0.829* 0.466
EXC -0.059 -0.182 -0.125 0.457* 0.122
0.124 -0.229 -0.060 -0.031 0.493
-0.150 -0.141 -0.050 0.654* -0.220
Panel B-4: Bonds
POS 0.299 0.128 -0.557* 0.069 0.142
0.137 -0.120 -0.329 0.155 -0.093
0.438 0.302 -0.651* 0.077
0.235
NEG 0.141 -0.231 0.289 -0.039 0.053
0.110 -0.219 0.275 -0.331 0.320
0.293 -0.255 0.452 -0.151 -0.329
SUM 0.430* -0.004 -0.446* 0.053 0.195
0.263 -0.329 -0.198 -0.070 0.140
0.560* 0.153 -0.379 -0.005 0.052
NET 0.161 0.189 -0.527* 0.067 0.082
0.045 0.013 -0.341 0.247 -0.200
0.231 0.366 -0.744* 0.132
0.343
EXC 0.189 -0.115 0.147 0.046 0.036
0.112 -0.226 0.277 -0.324 0.310
0.323 -0.021 0.184 0.150 -0.428
Notes: Panel A reports three volatility measures for credit flows: the standard deviation (1st to 3rd row), the standard deviation of the HP-filtered credit flows (4th to 6th row)
and the coefficient of variation of the flows (standard deviation/mean) (7th to 9th rows). Panel B reports the unconditional correlation coefficients of the credit flows with
the HP-filtered GDP growth rate. Panel B-1 refers to total credit, Panel B-2 to total credit excluding 2008, Panel B-3 to loans and Panel B-4 to bonds. Each panel displays
correlations for the full sample (1981-2012) period and for the pre-crisis (1981-1996) period and for the post-crisis (1999-2012) period. * denotes statistical significance at the
5% level.

Table 1.4: Volatility and Unconditional Correlation

27

Credit

Size
Loans

Bonds

Credit

Manufacturing
Loans
Bonds

Chaebol Affiliation
Credit
Loans
Bonds

Panel A: Gross credit reallocation
Sectoral effects
81-12
0.609
0.760
0.723
0.830
0.689
0.532
0.574
0.414
0.342
81-96
1.055
3.996
0.627
2.039
1.321
0.333
2.955
2.155
0.196
99-12
0.602
1.404
0.361
0.368
0.860
0.560
0.286
1.107
0.440
Idiosyncratic effects
81-12
0.599
0.722
2.200
0.316
0.161
0.577
1.006
0.753
1.582
81-96
1.748
3.843
2.147
1.595
0.842
0.363
4.123
1.765
1.374
99-12
0.912
1.978
1.404
0.480
0.404
0.200
1.516
1.858
1.520
81-12
-0.209 -0.482
-1.923
-0.146 0.150
-0.109
-0.580 -0.167
-0.925
Covariance term
81-96
-1.802 -6.839
-1.774
-2.633 -1.163
0.304
-6.078 -2.920
-0.570
99-12
-0.514 -2.382
-0.765
0.152
-0.264
0.239
-0.802 -1.965
-0.960
Panel B: Excess credit reallocation
Sectoral effects
81-12
1.086
0.929
1.340
0.986
0.772
1.104
0.791
0.527
1.288
81-96
3.723
2.219
1.198
1.551
1.002
0.849
1.771
0.932
1.075
99-12
1.873
2.038
0.514
1.352
1.243
0.658
1.693
1.581
0.734
81-12
0.402
0.427
0.871
0.064
0.090
0.302
0.329
0.388
0.975
Idiosyncratic effects
81-96
3.518
1.000
0.749
0.524
0.148
0.222
1.735
0.176
0.852
99-12
0.604
0.477
0.892
0.090
0.176
0.189
0.533
0.456
1.149
Covariance term
81-12
-0.489 -0.356
-1.211
-0.050 0.138
-0.406
-0.121 0.084
-1.263
81-96
-6.241 -2.219
-0.947
-1.074 -0.150
-0.072
-2.506 -0.108
-0.927
99-12
-1.476 -1.515
-0.406
-0.442 -0.419
0.153
-1.226 -1.037
-0.883
Notes: Panel A of this table shows the variance decomposition of the gross reallocation of total credit, loans and bonds. Panel
B shows the variance decomposition of the excess reallocation of total credit, loans and bonds.

Table 1.5: Properties of Idiosyncratic Flows

sentially acyclical, after the crisis the patterns appear to have flipped, with the net credit change
becoming almost acyclical and the excess reallocation becoming procyclical. Put differently, in the
pre-crisis period the mildly procyclical pattern of gross credit reallocation mostly reflected the procyclical pattern of credit growth; in the post-crisis period, instead, it mostly reflected the dynamics
of the excess credit reallocation.
Formally, we can decompose the correlation of gross credit reallocation with the GDP using
the following formula

corr(SUM, GDP) =

sd(EXC)
sd(|NET |)
corr(EXC, GDP) +
corr(|NET | , GDP).
sd(SUM)
sd(SUM)

(1.8)

The results of this decomposition (displayed in Table 1.6), confirm that the comovement of
gross credit reallocation with the business cycle was especially driven by the net credit change
before the crisis and by the excess credit reallocation after it. Naturally, this does not inform
us about causality. However, the finding is suggestive, as in the years leading up to the crisis
a credit boom occurred while after the crisis an enhanced dynamism in credit reallocation could

28

1981-1996
1999-2012
t-1
t
t+1
t+2
t-2
t-1
t
t+1
Panel A: Total credit
SUM
0.103 0.071 0.040 -0.324 0.128
0.122 0.058 0.297 -0.375 -0.241
0.125 0.158 0.064 -0.391
SD ratio 1 1.361 1.361 1.361 1.361 1.349
1.165 1.165 1.165 1.165 1.165
1.108 1.108 1.108 1.108
EXC
0.005 -0.120 -0.091 0.281 0.031
0.203 -0.193 -0.112 -0.231 0.327
0.015 -0.100 0.192 0.400
SD ratio 2 1.216 1.216 1.216 1.216 1.188
1.526 1.526 1.526 1.526 1.526
1.484 1.484 1.484 1.484
|NET|
0.079 0.193 0.135 -0.581 0.072
-0.075 0.185 0.280 -0.069 -0.408
0.073 0.181 -0.100 -0.562
Panel B: Loans
SUM
0.079 -0.024 0.004 -0.072 0.077
0.170 -0.311 0.223 0.015 -0.083
0.204 0.219 -0.108 -0.052
SD ratio 1 1.209 1.209 1.209 1.209 1.181
1.909 1.909 1.909 1.909 1.909
1.352 1.352 1.352 1.352
EXC
-0.033 -0.103 -0.050 0.276 0.087
0.150 -0.200 -0.054 -0.049 0.433
-0.071 0.027 -0.165 0.494
SD ratio 2 0.954 0.954 0.954 0.954 0.931
1.896 1.896 1.896 1.896 1.896
1.912 1.912 1.912 1.912
|NET|
0.125 0.106 0.067 -0.425 -0.027
-0.062 0.037 0.171 0.057 -0.480
0.157 0.096 0.060 -0.377
Panel C: Bonds
SUM
0.289 -0.023 -0.315 0.024 0.141
0.205 -0.219 -0.209 -0.197 0.124
0.368 0.074 -0.134 0.041
SD ratio 1 1.260 1.260 1.260 1.260 1.292
1.300 1.300 1.300 1.300 1.300
1.160 1.160 1.160 1.160
EXC
0.126 -0.098 0.122 0.054 0.038
0.121 -0.218 0.130 -0.328 0.297
0.254 0.049 0.195 0.099
SD ratio 2 1.226 1.226 1.226 1.226 1.249
1.215 1.215 1.215 1.215 1.215
0.827 0.827 0.827 0.827
|NET|
0.106 0.082 -0.383 -0.036 0.073
0.039 0.053 -0.311 0.189 -0.216
0.089 0.021 -0.435 -0.089
Notes: This table reports the decomposition of the correlation between credit reallocation and the HP-filtered GDP growth rate.
t-2

t-1

1981-2012
t
t+1

t+2

t-2

t+2
0.092
1.177
-0.295
1.538
0.285
-0.138
1.418
-0.180
1.941
0.060
0.008
1.193
-0.389
0.872
0.541

Table 1.6: Decomposition of Correlation

have promoted economic activity.

1.7.2.2

Conditional Correlation

Unconditional correlations do not control for microeconomic variables that may affect credit
reallocation. To address this issue, we now adopt the approach of Covas and Den Haan (2011)
which allows to control for the impact of micro-variables on credit reallocation. We estimate the
following regression
J
Fi,t
= α0,i + ∑ Ii,t ( j){α j,1t + α j,2t 2 + α j,3 Ytc +
Ai,t−1
j=1

α j,4 (

CFi,t−1 CF j,t−1
−
) + α j,5 (Qi,t−1 − Q j,t−1 )} + ui,t
Ai,t−2
A j,t−2

(1.9)

where Fi,t is the credit change of firm i in year t; Ai,t denotes the total assets of the firm; t and t 2
denote a linear and a quadratic time trend, respectively; Ii,t ( j) is an indicator variable that takes
the value of one if firm i belongs to the class j of firms, zero otherwise; and Ytc is the measure of
the cycle, the HP-filtered GDP. Following Covas and Den Haan (2011), we insert lagged values of
cash flows and Tobin’s Q as independent variables. For unlisted firms, since we lack information
on the Tobin’s Q, we use the two-year-ahead sales growth rate as a proxy. The sample spans
29

1987-2012
1987-1996
1999-2012
Credit
Loans
Bonds
Credit
Loans
Bonds
Credit
Loans
1st quintile
0.057*** 0.047*** -0.002***
0.263*** 0.152*** 0.011***
0.023*
0.040***
(0.007)
(0.006)
(0.001)
(0.020)
(0.017)
(0.004)
(0.012)
(0.010)
2nd quintile
0.036*** 0.035*** -0.002***
0.194*** 0.113*** 0.002
0.008
0.033***
(0.005)
(0.005)
(0.001)
(0.019)
(0.015)
(0.00408)
(0.009)
(0.008)
3rd quintile
0.034*** 0.026*** -0.001
0.132*** 0.073*** 0.005
0.025*** 0.036***
(0.005)
(0.004)
(0.001)
(0.018)
(0.015)
(0.004)
(0.008)
(0.007)
4th quintile
0.039*** 0.023*** 0.003***
0.076*** 0.038*** 0.001
0.047*** 0.052***
(0.005)
(0.004)
(0.001)
(0.018)
(0.014)
(0.005)
(0.009)
(0.007)
5th quintile
0.030*** 0.034*** -0.002
0.043**
0.020
-0.003
0.057*** 0.070***
(0.005)
(0.004)
(0.002)
(0.018)
(0.015)
(0.005)
(0.008)
(0.007)
Observations 145,026
145,026
145,026
23,281
23,281
23,281
113,666
113,666
R-squared
0.068
0.036
0.011
0.065
0.029
0.007
0.076
0.045
Notes: This table shows the coefficient estimates of firm level credit changes on the HP-filtered GDP for the five sales quintiles. The numbers in parentheses denote standard errors. *, ** and *** indicate 10%, 5 % and 1% statistical significance.
The 1st quintile and the 5th quintile are the quintiles of the smallest and the largest firms, respectively.

Bonds
-0.001
(0.001)
-0.002*
(0.001)
-0.007***
(0.001)
-0.009***
(0.001)
-0.009***
(0.002)
113,666
0.004

Table 1.7: Conditional Correlation of Credit Reallocation

from 1987 to 2012 and the number of firm-year observations is 145,026 (data for cash flows and
Tobin’s Q are not available before 1987). The HP-filtered GDP is scaled to be zero at its minimum
observed value and one at its maximum observed value. This enables us to interpret its estimated
coefficient as the change in credit when the economy goes from through to peak over the business
cycle. Moreover, we subtract the group mean from each variable to purge the effect of aggregate
conditions on independent variables.
Table 1.7 reports the estimation results for each sales quintile (to conserve space, we only
display the coefficient estimates for the HP-filtered GDP). As shown by Covas and Den Haan
(2011), it is useful to distinguish firms of different size when studying the cyclical behavior of their
debt. The results confirm the reduced procyclicality of the net credit change documented above
with the unconditional correlation coefficients: after the crisis firm-level credit changes became
less sensitive to the cycle.

1.7.2.3

Robustness Analysis

The reader might have some concern that for the post-crisis period the analysis of the dynamic
behavior of credit flows relies on thirteen annual data (from 1999 to 2012). While higher frequency
data for the whole sample period are not available, we have quarterly data for publicly traded firms

30

for essentially all the post-crisis period (from 2000Q1 to 2012Q4). We then replicated our analysis
for the post-crisis period using quarterly data. The findings confirm those obtained using annual
data. In particular, after the crisis excess credit reallocation exhibited a clearly procyclical behavior
while the net credit change was countercylical. The results of the analysis with quarterly data are
available on request.

1.8 Allocative Efficiency
Although a fully fledged analysis of the efficiency of the credit reallocation process is beyond
the scope of this paper, in this section we take a step towards investigating whether the intensification of credit reallocation after the financial crisis was associated with enhanced efficiency of the
reallocation process. To this end, we use various indicators of firm productivity and efficiency. The
first two indicators consist of the firm ratios of operating profits to capital and sales to capital.12
Using the profits and sales to capital ratios, we construct an index to evaluate the efficiency of
the allocation of credit. We adapt to our context the index for the efficiency of investment allocation
proposed by Galindo, Schiantarelli and Weiss (2007). The index is constructed as a ratio. In the
numerator, in state i and year t, the ratio includes the weighted sum of the sales or profits to capital
ratios of the firms (s f it /k f it ), with the weight for each firm given by the contribution of the firm
debt to the total debt of the firms in the state in that year (c f it /Cit ). In the denominator, the ratio
includes the sum of the sales to capital ratios of the same firms, weighted by the contribution of
the firm debt to the total debt of the firms in the previous year (c f it−1 /Cit−1 ). Formally, the index
reads

sft cft
f k Ct
ft
.
Iit = s c
f t f t−1
fk C
f t t−1

(1.10)

12 There are several reasons to use both sales and profits (see also Galindo et al., 2007). Sales are

measured more accurately than operating profits. Moreover, operating profits are highly correlated
with cash flow. Because cash flow is the main source of internal financing, a relationship between
cash flow and a change in debt may bias the index. Last, operating profits are more volatile than
sales.

31

30

1.2

20

1.1

10

1
.9

0

.8

1985

1990

1995

2000

2005

2010

Figure 1.5: Efficiency of Credit Reallocation This figure shows the annual values of three
efficiency indexes of credit reallocation computed using firms’ sales to capital ratios (squares),
profit to capital ratios (bullet points) and SFA efficiency (triangles). The figure shows six
quadratic fitted lines for each index in the pre-crisis period and the post-crisis period (dotted
lines for sales, dashed lines for profits and dash-dotted lines for SFA efficiency). The efficiency
index using operating profits starts in 1987 due to data availability. The right Y-axis provides the
scale for the magnitude of gross credit reallocation (solid bold line) and excess credit reallocation
(solid light line).

32

Total

Chaebol Affiliation
Chaebols Others

1st

2nd

Size Quintile
3rd
4th

5th

Panel A: Total credit
Average

87-96
93-96
99-12

1.022
1.059
1.040

0.942
0.990
1.009

85-96
93-96
99-12
Panel B: Loans

0.987
1.022
1.071

0.948
0.996
1.027

Average

Operating profits
0.983
0.490
1.030
0.478
0.977
0.932
Sales
0.957
1.202
0.975
1.309
1.004
1.295

0.420
0.441
0.783

0.462
0.697
0.983

0.385
0.398
0.934

0.481
0.503
0.478

1.589
1.716
1.152

1.364
1.311
1.314

1.010
1.300
1.300

0.770
0.738
0.585

Operating profits
1.004
0.513
0.432
0.444
0.436
0.518
1.074
0.415
0.493
0.589
0.408
0.565
0.948
0.824
0.770
0.978
0.855
0.665
Sales
Average
85-96
1.012
0.965
0.980
1.127
1.688
1.399
1.012
0.798
93-96
1.040
1.006
1.015
1.240
2.102
1.282
1.273
0.801
99-12
1.078
1.055
0.982
1.402
1.141
1.352
1.218
0.626
Notes: This table displays the values of the efficiency index of credit reallocation constructed using the profits to capital ratios of the
firms, the sales to capital ratios and the SFA efficiency. Panel A refers to total credit, Panel B to loans. Each panel reports the values
of the index for all firms, for chaebol and non-chaebol firms, and for firms of different size. It also reports values of the index for the
pre-crisis (1981-1996) period, for the credit boom (1993-1996) period and for the post-crisis (1999-2012) period.
Average

87-96
93-96
99-12

1.038
1.086
1.029

0.958
0.992
1.160

Table 1.8: Efficiency of Credit Reallocation

A value of the index greater than one signals that credit was allocated more efficiently in year t
than if the credit distribution had remained as in year t − 1.
Tables 1.8 and Figure 1.5 show the values of the index (the figure also plots fitted lines). The results consistently indicate that the efficiency of credit allocation jumped up after the crisis (indeed,
the pattern of the index tracks that of credit reallocation).

1.9

Conclusion

This paper has investigated the effect of a major financial crisis and the associated corporate
and financial reforms on the process of inter-firm credit reallocation. We have found that during the
credit boom that preceded the 1997 Korean financial crisis, the intensity of credit reallocation was
somewhat depressed. By contrast, after the crisis and the reforms enacted in its aftermath, credit
reallocation rose significantly, while credit growth slowed down. The increase in the intensity
of credit reallocation cannot be explained by episodes of “flight to quality”: the share of credit
reallocation occurring within groups of firms roughly homogeneous for size, industry or location,

33

rose substantially relative to the share of reallocation across such groups.
The analysis has further revealed that before the crisis credit growth comoved with the business
cycle more than credit reallocation, while after the crisis credit reallocation became more procyclical than credit growth. Finally, we have uncovered preliminary evidence that the increase in the
intensity of credit reallocation was associated with enhanced efficiency in the credit reallocation
process.
A large body of research has recently investigated the behavior of credit growth before and after
financial crises, focusing on the credit boom-and-busts that occur in coincidence with the crises.
All in all, our results suggest that financial crises and the associated reforms can play a pivotal role
not only in the dynamics of credit growth but also in the dynamism and flexibility of the dynamic
process of credit reallocation. A credit boom characterized by a depressed dynamism in the credit
reallocation process could be very different from a credit boom characterized by a fluid process
of reallocation of liquidity. Similarly, a creditless recovery characterized by increased dynamism
in the reallocation of credit could spur growth, despite the lower volume of liquidity flowing to
the business sector. Accounting for the behavior of credit reallocation can significantly further
our understanding of the causes and consequences of financial crises and of the impact of credit
markets on aggregate economic activity.

34

CHAPTER 2
DRIVING FORCES BEHIND THE EVOLUTION OF CREDIT REALLOCATION

2.1 Introduction
The Great Recession has reignited the debate over the long-run effects of financial crises and
of the reforms enacted in their aftermath (Allen et al., 2010; Gourinchas and Obstfeld, 2012;
Dell’Ariccia, Igan, Laeven and Tong, 2012). Nonetheless, we know very little about the dynamic
process of reallocation of credit before and after financial crises. The allocation of physical and
financial inputs plays a role as relevant as their total volume in affecting aggregate economic activity. Numerous studies find that the impact of aggregate shocks and structural reforms on the
macroeconomy occurs through the allocation of labor, capital, financial resources (Caballero and
Hammour, 2001, 2005; Davis and Haltiwanger, 1999; Eisfeldt and Rampini, 2006; Caballero,
Hoshi and Kashyap, 2008; Liu, 2013). In particular, the dynamism with which an economy is
able to reallocate financial resources across firms is considered as crucial for efficiency and growth
(Beck, Levine, Loayza, 2000; Wurgler, 2000; Galindo, Schiantarelli and Weiss, 2007). To fill the
gap, Chapter 1 studied the dynamic process of reallocation of credit across Korean non-financial
businesses in the 1981-2012 period. Especially, it investigated whether the credit reallocation process changed after the 1997 financial crisis and the subsequent reforms. However, it did not study
either the driving forces behind dynamic process of credit reallocation or the impact of the crisis
on the driving forces.
This chapter explores the evolution of credit reallocation across Korean non-financial firms in
the 1981-2012 period from the geographical location perspective.1 We examine the driving forces
behind dynamic process of credit reallocation as well as whether the forces changed after the 1997
crisis and the subsequent reforms. To do so, we construct regional credit flows based on each
1 This chapter is my own work. But, in the text, I use "we", which is accepted as a kind of a

customary practice of academic publication.
35

16 administrative region. Then, we employ a Bayesian dynamic latent factor model to estimate
common factor which explains common movement across the 16 regional credit reallocation rates.
Furthermore, we decompose credit reallocation fluctuations into three parts (national, regional and
idiosyncratic components) in order to examine the driving forces behind regional credit reallocation rates.
The Korean economy and our database constitute an ideal testing ground for our purposes.
Credit is a key source of external finance for Korean firms. It accounts for 66.88% of their outstanding external funding in 2002. Our unique data set comprises rich microeconomic data on
more than 33,000 non-financial firms and covers a long sample period, of which the 1997 crisis is
in the middle. This enables us to examine the long-run effects of the 1997 crisis and the associated
reforms on the evolution of credit reallocation and the driving forces behind it.
We uncover evidence that the common factor explaining regional reallocation flows across all
the 16 regions increased after the 1997 crisis and is highly correlated with national excess credit
reallocation. As Kose, Otrok, and Whiteman (2003) pointed out, because the factor is unobservable and we have merely extracted an estimate of it based on the observable time series variables
that are credit flows in our study, it is not easy to define what the common factor is. So we use
various combinations of credit flows to explore its properties. We use three types of credit: total
credit, loans and bonds. We explore the common factor of regional credit reallocation rates after
reconstructing credit flows based on chaebol-affiliated firms and non-affiliated firms, respectively.
In all analyses, the common factors have common patterns. First, they increased after the 1997
financial crisis. Next, they comove with national credit flows and are highly correlated with them.
In addition, the factors exhibit some cyclicality.
Some interesting patterns emerge when we investigate the roles played by the national, regionspecific and idiosyncratic components in driving the fluctuation of regional credit reallocation
rates. When it comes to regional excess reallocation rates of total credit and loans, the common
factors explain a large fraction of the volatility in the rates. By contrast, idiosyncratic components
contribute a large fraction of the volatility of regional bond excess reallocation rates, while the

36

common factor plays only a minor role in explaining the volatility in the flows.
This paper is organized as follows. Section 2 describes the spatial heterogeneity of the 16
regions which constitute South Korea. Section 3 describes the data source and introduces the
dynamic latent factor model. Section 4 examines the common factor affecting credit reallocation
across the 16 regions. Section 5 investigates the driving factors behind the credit reallocation from
the geographical location. Section 6 concludes.

2.2 Spatial Heterogeneity
South Korea is divided into 16 administrative regions: 9 provinces (Kyunggi, Kangwon, Kyungbuk, Kyungnam, Jeonbuk, Jeonnam, Chungbuk, Chungnam, and Jeju) and 7 metropolitan cities
(Seoul, Busan, Incheon, Daegu, Kwangju, Daejeon, and Ulsan). Seoul and the 7 metropolitan
cities respectively account for 26.34% and 50.05% of the national GDP in 2000. The 47.74% of
the national GDP in 2000 is attributed to Seoul, Incheon, and Kyunggi.2
Korea is a small country and the mobility between regions is quite easy. However, each region
has its own economic characteristics in terms of industrialization, economic growth, credit market and financial development. For example, Jeonnam, Jeonbuk, Jeju and Kangwon are relatively
rural areas, while Kyungnam, and Kyunggi are industrialized areas. Among metropolitan cities,
Busan and Incheon are more industrialized and more populated than Kwangju and Daejeon. The
development of regional loan markets varies. Regional banks are operating based on its own operation region and Seoul in the 6 regions (Busan, Daegu, Kwangju, Kyungnam, Jeonbuk, and Jeju).
Furthermore, regions respond to aggregate shocks differently. For example, the Korean economy
underwent a recession due to the adverse external shock in 2008 and 2009. Real GRDP growth
rates of Seoul and Ulsan respectively decreased 4.37% and 4.20% in 2007 to 1.50% and -3.21%
in 2009 due to decreased exports. On the other hand, those of Chungbuk and Jeju, in which service industries dominates, remained stable at 8.91% and 6.71% in 2009, respectively, compared
7.58% and 6.38% in 2007. Over the 1985-2012 period, the average standard deviation of GRDPs
2 These three areas are called the capital area.

37

in the regions is 3.47.3 Such different responses of regional growth rates to economic shocks are
consistent with economic theories which imply that spatial heterogeneous properties lead regions
to respond to economics shocks differently (Anas, Arnott, and Small, 1998; Bertola, 1993; Martin
and Sunley, 1998).
When it comes to credit reallocation, the regions do not only have common properties but they
also have their own properties. Tables 2.1 to 2.3 show that regional credit flows of total credit,
loans, and bonds, respectively. Both gross reallocation and excess reallocation increased after the
financial crisis in all regions. The averages of credit reallocation (total credit) range from 19%
to 28% in the regions over the whole sample period. These uniform pattern of increased reallocation after the crisis and of intense credit reallocation rates indicate that every region underwent
structural impact of the crisis on the credit market after the crisis and has substantial heterogeneity
in the regional credit reallocation rates. Hence, region is a possible good source of differences in
credit reallocation, hence exploring the common factor that explains common movement across
regional credit flows would be very useful to get insight into the driving forces behind dynamics
of credit reallocation.

2.3 Data and Methodology
2.3.1

Data

We obtained firm-level credit data from KISLINE, the business information source of the leading Korean credit rating agency, Korea Investors Service, which is affiliated with Moody’s. Our
data set covers all the publicly traded firms as well as all the privately held firms subject to annual external auditing. The data set includes 33,463 firms (2,245 publicly traded firms and 31,218
privately held ones) and 373,685 firm-year observations over the period 1980 to 2012. The firms
in the data set account for a large fraction of economic activity in Korea. The bank loans they
obtained amounted to 81.6% of the bank loans to all non-financial businesses in 2008. The long
3 Data on Gross Regional Domestic Production (GRDP) are available since 1985.

38

sample period and the extensive coverage enable us to analyze the effects of the 1997 crisis on
the evolution of credit reallocation. Using the methodology proposed by Davis and Haltiwanger
(1992) for measuring job reallocation and employed by Herrera, Kolar, and Minetti (2011), we
construct region-level credit flows of total credit, loans, and bonds (see Tables 2.1 to 2.3).

2.3.2

Methodology

This section outlines the Bayesian dynamic latent factor model, proposed by Kose, Otrok, and
Whiteman (2003, 2008) and Crucini, Kose, and Otrok (2011) that is employed in this paper. The
dynamic latent factor model is written as:
yi jt = βinj Ft + βirj fitr + εi jt

(2.1)

εi jt = ρi j1 εi jt−1 + · · · + ρi jp εi jt−p + ui jt

(2.2)

Ft = ρ1F Ft−1 + · · · + ρ pF Ft−p + utF

(2.3)

f

f

r
r
+ · · · + ρip fit−p
+ urit
fitr = ρi1 fit−1

(2.4)

where i and t respectively denote region (i = 1, ...,16) and year (t = 1981,...., 2012) and j denotes
credit flow (j=1: gross reallocation and j=2: excess reallocation). Ft is a common factor shared by
all variables across 16 regions. fitr is a region-specific factor for each region i. βinj and βirj are factor
loads for the common and region-specific factors, respectively. They measure the degree to which
variation in yi jt can be explained by each factor. εi jt is an idiosyncratic component of each variable.
As shown Equations (2) to (4), the two factors and idiosyncratic components follow an AR(p)
process. We follow the Bayesian procedures proposed by Kose, Otrok, and Whiteman (2003, 2008)
and Crucini, Kose, and Otrok (2011) to estimate factors and idiosyncratic components. Posterior
distribution properties for the model parameters and factors are based on 50,000 Markov Chain
Monte Carlo (MCMC) replications after 5,000 burn-in replications. To save space here, refer to
Kose, Otrok, and Whiteman (2003, 2008) and Crucini, Kose, and Otrok (2011) for more technical
details on the estimation procedure.

39

Region
Seoul

Period
POS
NEG
SUM
NET
EXC
1981-2012
13.521
7.036
20.557
6.484
12.842
1981-1996
13.230
4.202
17.432
9.028
8.404
1999-2012
13.014
10.190
23.204
2.825
17.690
Kyunggi
1981-2012
14.350
6.627
20.977
7.724
11.319
1981-1996
14.171
3.677
17.848
10.494
7.175
1999-2012
13.979
9.764
23.742
4.215
16.410
Incheon
1981-2012
14.927
8.102
23.029
6.825
11.844
1981-1996
13.426
3.599
17.024
9.827
7.197
1999-2012
15.466
13.780
29.246
1.686
17.595
Busan
1981-2012
16.658
9.180
25.838
7.478
16.074
1981-1996
16.216
5.967
22.184
10.249
11.934
1999-2012
17.187
12.857
30.044
4.331
20.488
Daegu
1981-2012
15.859
8.254
24.113
7.605
13.334
1981-1996
18.278
3.870
22.148
14.407
7.741
1999-2012
12.994
13.616
26.610
-0.622
19.975
Kyungbuk
1981-2012
13.431
6.225
19.656
7.206
9.347
1981-1996
11.925
3.817
15.742
8.108
6.632
1999-2012
15.201
8.640
23.840
6.561
12.887
Jeonbuk
1981-2012
17.082
8.412
25.495
8.670
15.376
1981-1996
17.749
4.832
22.581
12.918
9.663
1999-2012
15.656
12.313
27.968
3.343
21.314
Kyungnam
1981-2012
13.771
7.412
21.183
6.358
13.122
1981-1996
11.144
4.325
15.468
6.819
8.649
1999-2012
16.929
10.835
27.765
6.094
17.995
Kwangju
1981-2012
17.942
10.108
28.049
7.834
15.109
1981-1996
16.833
3.582
20.415
13.251
6.001
1999-2012
20.032
17.468
37.500
2.563
26.653
Ulsan
1981-2012
13.940
7.001
20.941
6.938
9.034
1981-1996
14.385
5.830
20.214
8.555
8.101
1999-2012
12.802
8.858
21.660
3.944
10.424
Kwangwon
1981-2012
16.394
7.489
23.883
8.905
12.541
1981-1996
17.912
4.645
22.558
13.267
8.380
1999-2012
14.603
10.074
24.678
4.529
17.027
Jeonnam
1981-2012
18.505
8.297
26.803
10.208
12.221
1981-1996
20.339
3.262
23.602
17.077
5.567
1999-2012
16.291
13.876
30.166
2.415
19.715
Daejeon
1981-2012
14.769
7.433
22.202
7.336
11.710
1981-1996
13.700
3.820
17.521
9.880
7.193
1999-2012
16.803
11.309
28.112
5.494
17.135
Chungbuk
1981-2012
15.860
9.655
25.515
6.205
12.766
1981-1996
15.594
3.952
19.546
11.643
6.241
1999-2012
15.286
15.547
30.833
-0.261
19.823
Chungnam
1981-2012
16.105
8.209
24.313
7.896
13.922
1981-1996
16.450
4.328
20.778
12.121
8.415
1999-2012
15.530
12.802
28.331
2.728
21.365
Jeju
1981-2012
16.424
7.006
23.430
9.419
12.228
1981-1996
14.613
6.210
20.823
8.403
10.301
1999-2012
18.683
7.754
26.437
10.929
13.829
Notes: This table reports the average flows of total credit of the 16 regions. The period 1981 to 1996 and
the period 1999 to 2012 reflect the pre-crisis period and the post-crisis one, respectively.

Table 2.1: Magnitude of Gross Flows (Total Credit)

40

Region
Seoul

Period
POS
NEG
SUM
NET
EXC
1981-2012
18.212
12.941
31.152
5.271
20.949
1981-1996
15.493
7.972
23.465
7.521
13.722
1999-2012
20.669
18.551
39.220
2.117
29.889
Kyunggi
1981-2012
20.364
12.223
32.587
8.141
21.127
1981-1996
19.091
8.099
27.191
10.992
16.199
1999-2012
21.179
16.796
37.975
4.382
27.849
Incheon
1981-2012
18.663
12.159
30.822
6.505
18.966
1981-1996
15.694
7.418
23.111
8.276
12.643
1999-2012
21.015
18.164
39.179
2.852
26.602
Busan
1981-2012
20.979
13.454
34.432
7.525
22.736
1981-1996
19.379
11.154
30.533
8.225
18.505
1999-2012
23.269
16.075
39.344
7.194
28.612
Daegu
1981-2012
20.518
12.514
33.032
8.004
19.506
1981-1996
21.622
7.848
29.470
13.774
12.142
1999-2012
19.119
18.415
37.534
0.704
28.423
Kyungbuk
1981-2012
17.569
12.715
30.284
4.854
17.366
1981-1996
13.423
7.802
21.224
5.621
9.494
1999-2012
22.315
17.608
39.924
4.707
27.167
Jeonbuk
1981-2012
22.736
11.854
34.590
10.882
20.566
1981-1996
23.167
8.250
31.417
14.918
13.451
1999-2012
22.333
16.214
38.547
6.119
28.730
Kyungnam
1981-2012
17.784
11.356
29.141
6.428
19.644
1981-1996
13.168
8.085
21.253
5.082
13.960
1999-2012
22.773
14.609
37.383
8.164
26.217
Kwangju
1981-2012
22.037
14.185
36.222
7.852
19.910
1981-1996
18.187
8.133
26.320
10.054
12.712
1999-2012
25.970
22.439
48.409
3.530
29.603
Ulsan
1981-2012
20.701
15.238
35.939
5.463
16.305
1981-1996
19.653
14.163
33.816
5.490
13.124
1999-2012
22.243
17.719
39.963
4.524
20.711
Kwangwon
1981-2012
23.723
14.095
37.818
9.627
22.398
1981-1996
21.789
9.866
31.655
11.923
12.193
1999-2012
25.954
18.214
44.169
7.740
32.990
Jeonnam
1981-2012
23.410
13.399
36.808
10.011
20.604
1981-1996
25.461
9.686
35.147
15.774
12.573
1999-2012
21.142
17.519
38.661
3.623
29.203
Daejeon
1981-2012
20.450
13.248
33.698
7.201
20.479
1981-1996
17.276
10.145
27.421
7.132
14.545
1999-2012
24.802
16.821
41.623
7.981
28.286
Chungbuk
1981-2012
20.016
13.133
33.149
6.882
18.457
1981-1996
17.897
5.758
23.655
12.139
9.581
1999-2012
21.504
19.815
41.320
1.689
27.766
Chungnam
1981-2012
20.374
11.823
32.197
8.551
20.089
1981-1996
19.932
6.675
26.607
13.257
13.036
1999-2012
20.598
17.992
38.590
2.606
29.964
Jeju
1981-2012
23.038
11.984
35.022
11.054
16.933
1981-1996
22.109
10.411
32.519
11.698
9.624
1999-2012
23.435
14.351
37.785
9.084
25.420
Notes: This table reports the average loan flows of the 16 regions. The period 1981 to 1996 and the
period 1999 to 2012 reflect the pre-crisis period and the post-crisis one, respectively.

Table 2.2: Magnitude of Gross Flows (Loans)

41

Region
Seoul

Period
POS
NEG
SUM
NET
EXC
1981-2012
20.994
10.604
12.779
10.391
18.819
1981-1996
23.916
8.843
15.581
15.072
17.178
1999-2012
15.709
13.411
7.176
2.297
21.944
Kyunggi
1981-2012
21.954
12.802
20.295
9.152
20.295
1981-1996
24.377
11.900
18.757
12.477
18.757
1999-2012
17.499
14.422
22.473
3.077
22.473
Incheon
1981-2012
21.698
11.926
18.693
9.771
18.693
1981-1996
25.297
9.528
17.463
15.769
17.463
1999-2012
14.761
15.652
21.332
-0.892
21.332
Busan
1981-2012
27.592
18.401
24.800
9.191
24.800
1981-1996
29.848
14.184
23.592
15.663
23.592
1999-2012
23.650
24.734
27.493
-1.084
27.493
Daegu
1981-2012
26.423
19.823
23.413
6.599
23.413
1981-1996
30.678
11.032
13.979
19.646
13.979
1999-2012
22.748
29.523
34.629
-6.775
34.629
Kyungbuk
1981-2012
26.810
11.850
12.296
14.960
12.296
1981-1996
32.109
8.572
11.105
23.537
11.105
1999-2012
21.581
16.223
13.483
5.358
13.483
Jeonbuk
1981-2012
24.878
18.345
22.064
6.533
22.064
1981-1996
26.090
7.990
13.569
18.100
13.569
1999-2012
25.587
28.837
32.003
-3.250
32.003
Kyungnam
1981-2012
24.287
15.765
24.650
8.522
24.650
1981-1996
23.156
10.823
19.553
12.333
19.553
1999-2012
26.362
20.324
28.624
6.038
28.624
Kwangju
1981-2012
26.270
18.105
11.719
8.165
11.719
1981-1996
33.277
6.318
7.113
26.959
7.113
1999-2012
21.620
27.491
17.867
-5.871
17.867
Ulsan
1981-2012
35.584
16.024
13.512
19.560
13.512
1981-1996
54.224
13.394
13.285
40.830
13.285
1999-2012
15.230
20.572
14.206
-5.342
14.206
Kwangwon
1981-2012
38.583
19.807
26.476
18.777
26.476
1981-1996
46.855
8.444
14.157
38.411
14.157
1999-2012
32.678
31.188
41.586
1.490
41.586
Jeonnam
1981-2012
26.346
18.465
15.682
7.881
15.682
1981-1996
32.825
11.959
13.555
20.866
13.555
1999-2012
20.504
26.950
18.322
-6.447
18.322
Daejeon
1981-2012
31.368
22.922
22.125
8.446
22.125
1981-1996
23.358
12.279
16.884
11.078
16.884
1999-2012
42.613
35.291
27.687
7.322
27.687
Chungbuk
1981-2012
37.244
25.672
33.071
11.572
33.071
1981-1996
38.195
12.164
21.476
26.031
21.476
1999-2012
33.080
41.667
44.826
-8.587
44.826
Chungnam
1981-2012
25.014
13.462
20.800
11.553
20.800
1981-1996
28.481
7.692
13.556
20.788
13.556
1999-2012
21.394
20.725
29.544
0.669
29.544
Jeju
1981-2012
43.638
28.931
18.725
14.707
18.725
1981-1996
43.261
22.755
6.727
20.506
6.727
1999-2012
48.790
34.469
32.088
14.321
32.088
Notes: This table reports the average bond flows of the 16 regions. The period 1981 to 1996 and the
period 1999 to 2012 reflect the pre-crisis period and the post-crisis one, respectively.

Table 2.3: Magnitude of Gross Flows (Bonds)

42

0

10

−.5 −.25

0

20

.25

30

.5

Panel A: Total credit

1981

1986

1991

1996

2001

2006

2011

2001

2006

2011

2001

2006

2011

0

−1

10

−.5

0

30

.5

1

50

Panel B: Loans

1981

1986

1991

1996

0

10

−1.5 −1 −.5

0

30

.5

1

50

Panel C: Bonds

1981

1986

1991

1996

Figure 2.1: Evolution of Common Factor (1) This figure shows movements of common factors
of regional credit flows for total credit (Panel A), loans (Panel B), and bonds (Panel C). The thick
bold line surrounded by dotted lines in each panel indicates the median of the posterior distribution
of the common factor. The two dotted lines in each panel indicate 33- and 66-percent quantile
bands. The thin bold line and dashed line are respectively national excess credit reallocation and
gross credit reallocation (scale on the right Y-axis)

43

2.4 What is the Common Factor?
2.4.1

Common factor: Total Credit, Loans, and Bonds

Figure 2.1 represents the median of the posterior distribution of the common factors explaining
credit reallocation across 16 regions, along with 33- and 66-percent quantile bands. Panels A, B,
and C respectively show the common factors of reallocation of total credit, loans and bonds. The
narrowness of the bands indicates the factors are estimated quite precisely. In turn, we examine the
features of common factors explaining the all regional reallocation of total credit, loans and bonds.
Panel A, Figure 2.1, exhibits two interesting patterns of the common factor. First, it increased
after the crisis and picked up again in 2008. It tends to decrease during credit booms (1993-1996
and 2005-2007). Panel B shows that the common factor explaining regional loan reallocation rates
follows the two patterns. Panel C depicts the common factor of regional bond reallocation rates.
This factor does not follow the patterns.

2.4.2

Common factor: Chaebol Affiliation

We construct regional credit flows based on chaebol-affiliated firms and non-affiliated firms,
respectively. Panel A of Figure 2.2 shows the common factor when we use regional excess reallocation rates of chaebol-affiliated firms and non-affiliated firms. The factor accounts for common
movement across the 16 regions as well as across chaebol-affiliated firms and non-affiliated firms.
It shows a sharp increase in the common factor in 1997. Since financial and corporate reforms
completed in 2001, the common factor has reduced until 2008. Panel B plots the common factor that accounts for regional gross reallocation rates of chaebol-affiliated firms and non-affiliated
firms. It tends to increase during the whole period. Panel C shows the common factor when we
use regional excess reallocation rates and gross reallocation rates of chaebol-affiliated firms only.
Interestingly, it dropped, then increased sharply in the 1997 financial crisis and in the 2008 recession. The common factor explaining regional excess reallocation rates and gross reallocation rates
of non-affiliated firms exhibits different pattern which is similar to Panel A of Figure 2.1.

44

Credit
Total credit

Period
POS
NEG
SUM
NET
EXC
GDP
1981-2012
-0.780*
0.757*
-0.496*
-0.838*
0.771*
-0.090
1981-1996
-0.702*
0.861*
-0.413
-0.794*
0.866*
-0.269
1999-2012
-0.839*
0.680*
-0.566*
-0.875*
0.686*
0.229
Loans
1981-2012
-0.615*
0.732*
0.056
-0.715*
0.869*
-0.144
1981-1996
-0.647*
0.919*
0.254
-0.827*
0.852*
-0.032
1999-2012
-0.510
0.417
-0.193
-0.505
0.892*
0.005
Bonds
1981-2012
-0.896*
0.605*
-0.628*
-0.907*
0.521*
0.495*
1981-1996
-0.937*
0.711*
-0.620*
-0.938*
0.724*
0.233
1999-2012
-0.803*
0.305
-0.596*
-0.803*
0.168
0.843*
Total credit
1981-2012
-0.844*
0.693*
-0.676*
-0.855*
0.792*
-0.284
(Chaebols)
1981-1996
-0.910*
0.628*
-0.696*
-0.939*
0.649*
-0.134
1999-2012
-0.833*
0.561*
-0.735*
-0.805*
0.725*
0.089
Total credit
1981-2012
-0.758*
0.625*
-0.401*
-0.793*
0.757*
-0.264
(Others)
1981-1996
-0.567*
0.801*
0.056
-0.809*
0.786*
-0.392
1999-2012
-0.834*
0.503
-0.505
-0.809*
0.708*
-0.107
Notes: All variables are HP-filtered. The period 1981 to 1996 and the period 1999 to 2012 correspond to the pre-crisis
period and the post-crisis one, respectively. * denotes statistical significance at the 5% level.

Table 2.4: Cyclicality of the Common Factor

2.4.3

Comovement with National Credit Flows and GDP

As Kose, Otrok, and Whiteman (2003) pointed out, because the factor is unobservable and we
have merely extracted an estimate of it based on the observable time series variables that are credit
flows in our study, we investigate the comovement of the factor with GDP growth and national
credit flows in order to know about what the common factor is. Notably, the factor comoves
with the national excess reallocation. The correlation coefficients range from 0.52 to 0.89. The
correlation is the most stronger for loan reallocation. The only exception is bond reallocation in
the post-crisis period. The corresponding correlation coefficient is 0.17 and is not statistically
significant.
Table 2.4 reports another interesting tendency; common factors are negatively correlated with
national net credit growths. The negative correlations are very strong and statistically significant,
ranging from -0.71 to - 0.94, except for bond reallocation in the post-crisis period (-0.51). In
addition, the common factors tend to be negatively correlated with national credit creation, while
they are positively correlated with national credit destruction. However, the factors do not seem
to have cyclicality. Except for bond reallocation, they are mildly counter-cyclical, but statistically
insignificant.
Based on the findings, we can interpret the pattern of the common factor as follows. Despite

45

spatial heterogeneity and different regional responses of credit flows to aggregate shocks, the pattern of national credit flows well explains the common factor accounting for regional credit reallocation rates. As the credit reallocation, whether gross or excess, increased after the 1997 financial
crisis, the common factor also increased after the crisis. In particular, the increase in the common
factor is largely attributable to the policy reforms on the financial sector during the financial crisis.
Lenders started extending and renewing bank loans based on a profit and risk basis after the crisis.
The uniform financial regulation, which applied to all the banks, could raise the magnitude of the
common factor that influences the regional loan reallocation rates in all the regions. In addition,
the reduced influence of regional banks and continued mergers between banks made the factor
increasing after the crisis. 4

2.5 Variance Decomposition
As noted in chapter 1, the volatility of regional credit reallocation rates are high and the reshuffling of credit within a region is relatively more active than the reallocation of credit between
regions. A related question is to what extent the fluctuation of credit reallocation is driven by national, region-specific and idiosyncratic components. In order to measure the relative contributions
of the common and region-specific factors to variations in credit flows, we estimate the share of
the variance of credit flows due to each factor. We decompose the variance of each credit flow
into the fraction that is due to each of the two factors and idiosyncratic component. The variance
of credit flows for orthogonal factors can be written as Equation (5). θinj is the proportion of the
total variability in credit reallocation of the region i attributable to the common factor, as shown in
Equation (6). Similarly, θirj and θicj are defined as the proportion due to the region-specific factor
and the idiosyncratic component.
4 The common factor of regional bond reallocation rates reflects two bond market reforms. The

factor increased sharply in 1986 when credit rating system was introduced and in 2000 to 2001
when the financial authority brought out several incentives to issue bonds in order to develop bond
markets (e.g., reduction in bond issuance fees).

46

1

1981

1986

1991

1996

2001

2006

2011

10
0

−1

0

−1

−.5

−.5

10

0

0

20

.5

20

.5

30

1

Panel B: Gross reallocation
30

Panel A: Excess reallocation

1981

1991

1996

2001

2006

2011

Panel D: Non−affiliated firms

1981

1986

1991

1996

2001

2006

2011

30
20
10

−.25

0

−.5

0

−2

−1

10

0

0

20

.25

1

30

.5

Panel C: Chaebol−affiliated firms

1986

1981

1986

1991

1996

2001

2006

2011

Figure 2.2: Evolution of Common Factor (2) This figure shows the movements of common
factors of credit flows for the 16 regional credit reallocation rates (Panel A: excess reallocation for
chaebol-affiliated firms and non-affiliated firms, Panel B: gross reallocation for chaebolaffiliated firms and non-affiliated firms, Panel C: gross and excess reallocation for chaebolaffiliated firms, and Panel D: gross and excess realloca-tion for non-affiliated firms). The thick
bold line surrounded by dotted lines in each panel indicates the median of the posterior
distribution of the common factor. The two dotted lines in each panel indicate 33- and 66percent quantile bands. The thin bold line and dashed line are respectively national excess
credit reallocation and national gross credit reallocation (scale on the right Y-axis)
47

1981-2012
1981-1996
1999-2012
33%
50%
66%
33%
50%
66%
33%
50%
66%
Seoul
National
0.563 0.585 0.607
0.006 0.016 0.065
0.180 0.211 0.241
Region-specific 0.130 0.184 0.240
0.117 0.270 0.453
0.509 0.581 0.650
Idiosyncratic
0.173 0.226 0.278
0.382 0.582 0.799
0.128 0.194 0.268
Kyunggi
National
0.612 0.633 0.655
0.167 0.229 0.365
0.344 0.382 0.420
Region-specific 0.011 0.032 0.078
0.088 0.160 0.251
0.233 0.297 0.369
Idiosyncratic
0.274 0.315 0.346
0.267 0.471 0.650
0.223 0.300 0.377
Incheon
National
0.512 0.533 0.554
0.006 0.013 0.025
0.055 0.070 0.087
Region-specific 0.028 0.072 0.154
0.084 0.259 0.494
0.058 0.182 0.392
Idiosyncratic
0.308 0.384 0.428
0.478 0.713 0.882
0.533 0.738 0.855
Busan
National
0.387 0.406 0.424
0.154 0.253 0.341
0.057 0.077 0.098
Region-specific 0.260 0.314 0.375
0.245 0.357 0.468
0.687 0.757 0.817
Idiosyncratic
0.218 0.279 0.333
0.212 0.337 0.482
0.098 0.156 0.226
Kyungnam
National
0.733 0.752 0.771
0.289 0.349 0.399
0.426 0.472 0.518
Region-specific 0.005 0.013 0.030
0.015 0.038 0.092
0.097 0.170 0.260
Idiosyncratic
0.195 0.219 0.242
0.496 0.569 0.633
0.227 0.326 0.420
Ulsan
National
0.216 0.232 0.250
0.101 0.171 0.213
0.249 0.277 0.305
Region-specific 0.072 0.171 0.312
0.214 0.359 0.517
0.040 0.106 0.272
Idiosyncratic
0.451 0.593 0.688
0.330 0.483 0.615
0.433 0.593 0.669
Daegu
National
0.474 0.498 0.523
0.015 0.037 0.123
0.019 0.029 0.042
Region-specific 0.104 0.147 0.204
0.036 0.102 0.214
0.272 0.406 0.591
Idiosyncratic
0.290 0.345 0.390
0.496 0.775 0.913
0.373 0.556 0.687
Kyungbuk
National
0.368 0.388 0.409
0.012 0.028 0.049
0.280 0.308 0.340
Region-specific 0.145 0.237 0.326
0.145 0.292 0.477
0.197 0.309 0.428
Idiosyncratic
0.284 0.373 0.461
0.483 0.669 0.813
0.253 0.370 0.480
Jeonbuk
National
0.525 0.546 0.567
0.004 0.011 0.028
0.181 0.214 0.251
Region-specific 0.079 0.132 0.193
0.062 0.164 0.352
0.398 0.470 0.547
Idiosyncratic
0.255 0.316 0.370
0.609 0.794 0.893
0.209 0.292 0.377
Kwangju
National
0.495 0.515 0.534
0.035 0.062 0.104
0.068 0.090 0.116
Region-specific 0.344 0.372 0.402
0.459 0.560 0.671
0.752 0.794 0.832
Idiosyncratic
0.086 0.112 0.138
0.245 0.352 0.447
0.071 0.106 0.145
Jeonnam
National
0.589 0.607 0.625
0.007 0.019 0.047
0.266 0.298 0.331
Region-specific 0.055 0.093 0.139
0.161 0.343 0.565
0.483 0.533 0.584
Idiosyncratic
0.250 0.296 0.334
0.391 0.608 0.779
0.121 0.165 0.209
Daejeon
National
0.733 0.753 0.772
0.350 0.562 0.685
0.420 0.454 0.488
Region-specific 0.014 0.029 0.059
0.056 0.109 0.183
0.091 0.156 0.252
Idiosyncratic
0.177 0.206 0.231
0.158 0.246 0.390
0.277 0.366 0.437
Chungbuk
National
0.713 0.734 0.753
0.319 0.420 0.488
0.243 0.279 0.316
Region-specific 0.021 0.040 0.069
0.038 0.091 0.193
0.128 0.237 0.393
Idiosyncratic
0.184 0.215 0.244
0.343 0.439 0.533
0.317 0.468 0.574
Chungnam
National
0.756 0.776 0.794
0.532 0.725 0.827
0.198 0.228 0.260
Region-specific 0.059 0.079 0.102
0.012 0.030 0.068
0.600 0.639 0.681
Idiosyncratic
0.115 0.140 0.164
0.112 0.189 0.325
0.088 0.123 0.158
Kwangwon National
0.374 0.392 0.409
0.059 0.091 0.125
0.113 0.134 0.158
Region-specific 0.176 0.265 0.358
0.152 0.347 0.553
0.688 0.727 0.767
Idiosyncratic
0.245 0.341 0.431
0.347 0.552 0.740
0.094 0.129 0.164
Jeju
National
0.103 0.115 0.127
0.129 0.162 0.199
0.005 0.010 0.017
Region-specific 0.104 0.192 0.341
0.161 0.299 0.478
0.100 0.259 0.474
Idiosyncratic
0.541 0.690 0.778
0.346 0.525 0.659
0.510 0.726 0.884
Notes: This table reports the variance decomposition for excess reallocation of total credit by 16 regions
(9 provinces and 7 metropolitan cities).

Table 2.5: Variance Decomposition for Credit Reallocation by 16 Regions

48

1981-2012
1981-1996
1999-2012
33%
50%
66%
33%
50%
66%
33%
50%
66%
Seoul
National
0.747 0.764 0.780
0.092 0.119 0.151
0.352 0.401 0.453
Region-specific 0.008 0.019 0.040
0.085 0.216 0.402
0.221 0.313 0.403
Idiosyncratic
0.182 0.206 0.226
0.464 0.642 0.765
0.148 0.240 0.350
Kyunggi
National
0.541 0.562 0.583
0.525 0.568 0.611
0.347 0.389 0.430
Region-specific 0.023 0.073 0.159
0.025 0.074 0.156
0.096 0.167 0.281
Idiosyncratic
0.271 0.351 0.401
0.241 0.320 0.385
0.302 0.413 0.500
Incheon
National
0.597 0.615 0.633
0.390 0.426 0.464
0.036 0.057 0.082
Region-specific 0.095 0.127 0.165
0.120 0.214 0.320
0.421 0.487 0.563
Idiosyncratic
0.218 0.256 0.289
0.242 0.346 0.436
0.363 0.441 0.513
Busan
National
0.464 0.481 0.499
0.137 0.170 0.206
0.227 0.273 0.322
Region-specific 0.133 0.194 0.268
0.222 0.322 0.452
0.384 0.464 0.547
Idiosyncratic
0.250 0.324 0.385
0.381 0.508 0.599
0.156 0.235 0.318
Kyungnam
National
0.631 0.648 0.665
0.482 0.517 0.553
0.058 0.083 0.110
Region-specific 0.017 0.040 0.081
0.045 0.082 0.134
0.110 0.210 0.368
Idiosyncratic
0.265 0.300 0.326
0.336 0.383 0.424
0.525 0.680 0.790
Ulsan
National
0.217 0.232 0.247
0.007 0.013 0.023
0.011 0.021 0.034
Region-specific 0.187 0.297 0.414
0.042 0.140 0.355
0.397 0.510 0.628
Idiosyncratic
0.351 0.469 0.579
0.623 0.837 0.932
0.340 0.460 0.571
Daegu
National
0.664 0.683 0.702
0.543 0.589 0.635
0.058 0.085 0.116
Region-specific 0.106 0.135 0.166
0.131 0.191 0.258
0.267 0.398 0.539
Idiosyncratic
0.146 0.178 0.209
0.155 0.208 0.258
0.353 0.495 0.630
Kyungbuk
National
0.646 0.663 0.680
0.174 0.205 0.238
0.023 0.038 0.057
Region-specific 0.092 0.144 0.198
0.156 0.273 0.425
0.079 0.193 0.369
Idiosyncratic
0.136 0.192 0.244
0.366 0.519 0.629
0.574 0.753 0.869
Jeonbuk
National
0.656 0.673 0.689
0.158 0.193 0.232
0.252 0.288 0.326
Region-specific 0.045 0.067 0.096
0.028 0.078 0.187
0.468 0.529 0.590
Idiosyncratic
0.225 0.254 0.279
0.581 0.680 0.749
0.113 0.165 0.223
Kwangju
National
0.392 0.408 0.425
0.008 0.014 0.022
0.112 0.141 0.174
Region-specific 0.226 0.298 0.375
0.661 0.736 0.803
0.168 0.297 0.435
Idiosyncratic
0.216 0.293 0.365
0.177 0.243 0.318
0.411 0.545 0.669
Jeonnam
National
0.421 0.440 0.459
0.102 0.125 0.149
0.235 0.278 0.324
Region-specific 0.157 0.206 0.266
0.177 0.263 0.387
0.299 0.382 0.465
Idiosyncratic
0.291 0.348 0.400
0.479 0.605 0.692
0.241 0.322 0.403
Daejeon
National
0.495 0.513 0.531
0.352 0.397 0.444
0.187 0.224 0.262
Region-specific 0.113 0.162 0.229
0.091 0.200 0.322
0.417 0.500 0.587
Idiosyncratic
0.256 0.324 0.374
0.231 0.365 0.497
0.169 0.257 0.346
Chungbuk
National
0.797 0.814 0.830
0.435 0.484 0.538
0.202 0.244 0.288
Region-specific 0.006 0.017 0.040
0.021 0.066 0.160
0.021 0.062 0.167
Idiosyncratic
0.132 0.155 0.177
0.292 0.386 0.467
0.543 0.644 0.714
Chungnam
National
0.654 0.671 0.687
0.203 0.239 0.280
0.088 0.119 0.155
Region-specific 0.043 0.070 0.110
0.060 0.187 0.385
0.419 0.477 0.541
Idiosyncratic
0.215 0.253 0.282
0.337 0.531 0.667
0.314 0.382 0.447
Kwangwon National
0.480 0.498 0.515
0.200 0.231 0.264
0.019 0.032 0.048
Region-specific 0.296 0.335 0.373
0.368 0.461 0.559
0.805 0.850 0.892
Idiosyncratic
0.130 0.167 0.204
0.195 0.294 0.391
0.063 0.102 0.149
Jeju
National
0.312 0.330 0.348
0.010 0.020 0.033
0.041 0.062 0.087
Region-specific 0.053 0.067 0.084
0.413 0.588 0.736
0.379 0.468 0.579
Idiosyncratic
0.569 0.599 0.623
0.234 0.383 0.556
0.352 0.462 0.549
Notes: This table reports the variance decomposition for excess reallocation of loans by 16 regions (9
provinces and 7 metropolitan cities).

Table 2.6: Variance Decomposition for Loan Reallocation by 16 Regions

49

1981-2012
1981-1996
1999-2012
33%
50%
66%
33%
50%
66%
33%
50%
66%
Seoul
National
0.122 0.179 0.244
0.042 0.094 0.170
0.010 0.025 0.056
Region-specific 0.347 0.472 0.587
0.717 0.801 0.862
0.311 0.456 0.609
Idiosyncratic
0.261 0.341 0.426
0.049 0.077 0.114
0.301 0.460 0.619
Kyunggi
National
0.051 0.083 0.122
0.005 0.014 0.032
0.026 0.058 0.105
Region-specific 0.179 0.309 0.458
0.082 0.213 0.442
0.641 0.715 0.785
Idiosyncratic
0.452 0.601 0.716
0.482 0.712 0.862
0.120 0.180 0.252
Incheon
National
0.043 0.068 0.097
0.115 0.186 0.267
0.007 0.017 0.036
Region-specific 0.258 0.372 0.489
0.319 0.418 0.518
0.288 0.453 0.637
Idiosyncratic
0.442 0.551 0.653
0.221 0.324 0.441
0.320 0.501 0.662
Busan
National
0.023 0.036 0.051
0.030 0.048 0.070
0.005 0.012 0.025
Region-specific 0.175 0.270 0.401
0.043 0.114 0.299
0.204 0.319 0.491
Idiosyncratic
0.554 0.686 0.778
0.635 0.813 0.890
0.480 0.652 0.768
Kyungnam
National
0.058 0.083 0.110
0.007 0.019 0.039
0.027 0.055 0.093
Region-specific 0.324 0.403 0.494
0.625 0.704 0.775
0.431 0.516 0.618
Idiosyncratic
0.412 0.502 0.581
0.178 0.248 0.328
0.285 0.387 0.482
Ulsan
National
0.008 0.016 0.026
0.029 0.045 0.066
0.004 0.009 0.019
Region-specific 0.081 0.168 0.311
0.094 0.203 0.394
0.216 0.365 0.554
Idiosyncratic
0.666 0.808 0.892
0.544 0.738 0.844
0.425 0.615 0.764
Daegu
National
0.034 0.054 0.077
0.057 0.084 0.115
0.079 0.128 0.185
Region-specific 0.138 0.247 0.383
0.381 0.483 0.590
0.131 0.341 0.558
Idiosyncratic
0.551 0.688 0.797
0.307 0.409 0.518
0.259 0.459 0.668
Kyungbuk
National
0.015 0.028 0.045
0.016 0.034 0.063
0.132 0.192 0.260
Region-specific 0.368 0.499 0.625
0.409 0.567 0.708
0.125 0.214 0.359
Idiosyncratic
0.334 0.462 0.592
0.216 0.355 0.515
0.399 0.540 0.648
Jeonbuk
National
0.163 0.202 0.244
0.035 0.057 0.082
0.024 0.049 0.082
Region-specific 0.057 0.140 0.270
0.045 0.091 0.169
0.061 0.228 0.524
Idiosyncratic
0.501 0.623 0.716
0.757 0.832 0.879
0.400 0.666 0.835
Kwangju
National
0.067 0.089 0.115
0.047 0.081 0.119
0.011 0.025 0.052
Region-specific 0.088 0.174 0.317
0.520 0.605 0.685
0.101 0.217 0.398
Idiosyncratic
0.577 0.718 0.804
0.204 0.287 0.382
0.535 0.720 0.849
Jeonnam
National
0.003 0.007 0.014
0.030 0.061 0.100
0.066 0.102 0.146
Region-specific 0.071 0.147 0.261
0.516 0.611 0.694
0.050 0.116 0.228
Idiosyncratic
0.724 0.839 0.914
0.205 0.298 0.411
0.620 0.731 0.817
Daejeon
National
0.060 0.084 0.111
0.264 0.320 0.370
0.016 0.035 0.064
Region-specific 0.129 0.210 0.313
0.215 0.365 0.509
0.289 0.461 0.631
Idiosyncratic
0.580 0.687 0.773
0.197 0.314 0.436
0.310 0.479 0.646
Chungbuk
National
0.141 0.177 0.215
0.033 0.065 0.114
0.005 0.013 0.027
Region-specific 0.036 0.108 0.235
0.291 0.448 0.586
0.046 0.119 0.281
Idiosyncratic
0.575 0.684 0.759
0.330 0.455 0.564
0.683 0.842 0.921
Chungnam
National
0.147 0.189 0.232
0.014 0.030 0.053
0.008 0.018 0.037
Region-specific 0.169 0.266 0.368
0.126 0.242 0.414
0.283 0.394 0.519
Idiosyncratic
0.441 0.534 0.619
0.535 0.701 0.815
0.436 0.564 0.679
Kwangwon National
0.198 0.246 0.299
0.008 0.020 0.045
0.045 0.086 0.146
Region-specific 0.025 0.046 0.074
0.104 0.260 0.518
0.070 0.162 0.303
Idiosyncratic
0.622 0.686 0.740
0.413 0.647 0.813
0.515 0.701 0.834
Jeju
National
0.113 0.141 0.173
0.065 0.095 0.127
0.007 0.017 0.036
Region-specific 0.103 0.203 0.332
0.205 0.418 0.611
0.188 0.372 0.568
Idiosyncratic
0.506 0.636 0.738
0.279 0.467 0.674
0.390 0.583 0.765
Notes: This table reports the variance decomposition for excess reallocation of bonds by 16 regions (9 provinces
and 7 metropolitan cities).

Table 2.7: Variance Decomposition for Bond Reallocation by 16 Regions

50

1981-2012
1981-1996
33%
50%
66%
33%
50%
66%
Seoul
National
0.308 0.370 0.435
0.034 0.066 0.110
Region-specific 0.239 0.306 0.379
0.117 0.243 0.401
Idiosyncratic
0.222 0.289 0.362
0.476 0.642 0.780
Kyunggi
National
0.129 0.166 0.206
0.016 0.036 0.065
Region-specific 0.039 0.074 0.139
0.019 0.053 0.147
Idiosyncratic
0.650 0.723 0.781
0.741 0.864 0.925
Incheon
National
0.087 0.112 0.140
0.018 0.045 0.107
Region-specific 0.062 0.162 0.310
0.047 0.164 0.387
Idiosyncratic
0.562 0.705 0.803
0.438 0.655 0.825
Busan
National
0.003 0.007 0.014
0.006 0.015 0.033
Region-specific 0.039 0.109 0.242
0.230 0.399 0.583
Idiosyncratic
0.740 0.872 0.942
0.370 0.550 0.715
Kyungnam National
0.003 0.007 0.014
0.021 0.053 0.111
Region-specific 0.088 0.182 0.325
0.141 0.362 0.524
Idiosyncratic
0.661 0.803 0.896
0.370 0.514 0.721
Ulsan
National
0.048 0.067 0.090
0.034 0.063 0.103
Region-specific 0.537 0.602 0.671
0.661 0.723 0.784
Idiosyncratic
0.249 0.319 0.386
0.127 0.181 0.239
Daegu
National
0.045 0.064 0.087
0.040 0.098 0.192
Region-specific 0.325 0.438 0.562
0.288 0.400 0.535
Idiosyncratic
0.355 0.484 0.600
0.229 0.373 0.536
Kyungbuk
National
0.079 0.107 0.139
0.107 0.159 0.231
Region-specific 0.121 0.235 0.393
0.070 0.209 0.409
Idiosyncratic
0.484 0.641 0.751
0.333 0.517 0.697
Jeonbuk
National
0.002 0.005 0.011
0.035 0.058 0.086
Region-specific 0.671 0.730 0.787
0.735 0.788 0.838
Idiosyncratic
0.201 0.258 0.317
0.094 0.138 0.185
Kwangju
National
0.026 0.041 0.060
0.074 0.143 0.250
Region-specific 0.085 0.191 0.351
0.119 0.231 0.394
Idiosyncratic
0.591 0.751 0.861
0.359 0.503 0.646
Jeonnam
National
0.075 0.109 0.149
0.035 0.059 0.097
Region-specific 0.488 0.554 0.621
0.429 0.533 0.642
Idiosyncratic
0.248 0.315 0.382
0.261 0.365 0.469
Daejeon
National
0.078 0.110 0.147
0.179 0.334 0.506
Region-specific 0.342 0.363 0.383
0.257 0.399 0.536
Idiosyncratic
0.482 0.520 0.555
0.157 0.221 0.287
Chungbuk
National
0.010 0.021 0.036
0.152 0.266 0.412
Region-specific 0.711 0.748 0.788
0.060 0.136 0.264
Idiosyncratic
0.181 0.220 0.257
0.238 0.403 0.600
Chungnam National
0.076 0.100 0.127
0.004 0.011 0.024
Region-specific 0.014 0.038 0.091
0.041 0.099 0.197
Idiosyncratic
0.775 0.833 0.875
0.766 0.866 0.929
Notes: This table reports the variance decomposition for excess reallocation of total credit
firms by 14 regions (7 provinces and 7 metropolitan cities).

1999-2012
33%
50%
66%
0.453 0.546 0.630
0.048 0.109 0.200
0.194 0.280 0.373
0.163 0.215 0.269
0.195 0.332 0.489
0.289 0.437 0.560
0.003 0.009 0.018
0.074 0.254 0.527
0.448 0.718 0.899
0.075 0.117 0.164
0.056 0.153 0.313
0.517 0.674 0.783
0.288 0.358 0.421
0.054 0.110 0.208
0.379 0.481 0.572
0.161 0.216 0.279
0.068 0.150 0.276
0.423 0.569 0.688
0.014 0.029 0.049
0.273 0.442 0.631
0.322 0.512 0.679
0.034 0.060 0.094
0.112 0.283 0.491
0.424 0.627 0.789
0.022 0.041 0.064
0.546 0.628 0.711
0.233 0.315 0.397
0.091 0.139 0.195
0.049 0.113 0.240
0.550 0.676 0.771
0.148 0.202 0.257
0.335 0.435 0.535
0.237 0.337 0.441
0.003 0.008 0.017
0.092 0.223 0.431
0.550 0.755 0.885
0.119 0.173 0.227
0.669 0.730 0.790
0.069 0.091 0.115
0.303 0.378 0.453
0.046 0.107 0.215
0.325 0.439 0.542
for chaebol-affiliated

Table 2.8: Variance Decomposition (Chaebol-affiliated firms)

51

1981-2012
1981-1996
1999-2012
33%
50%
66%
33%
50%
66%
33%
50%
66%
Seoul
Natoinal
0.586 0.604 0.622
0.008 0.019 0.037
0.324 0.344 0.364
Region-specific 0.161 0.200 0.241
0.213 0.401 0.597
0.270 0.328 0.390
Idiosyncratic
0.156 0.196 0.232
0.367 0.564 0.746
0.272 0.333 0.387
Kyunggi
National
0.743 0.759 0.775
0.004 0.010 0.020
0.167 0.184 0.202
Region-specific 0.007 0.020 0.045
0.034 0.103 0.234
0.058 0.144 0.267
Idiosyncratic
0.183 0.208 0.229
0.743 0.872 0.941
0.549 0.671 0.749
Incheon
National
0.463 0.480 0.496
0.142 0.188 0.243
0.185 0.199 0.214
Region-specific 0.031 0.067 0.139
0.330 0.437 0.546
0.100 0.214 0.366
Idiosyncratic
0.379 0.445 0.481
0.219 0.327 0.447
0.434 0.587 0.699
Busan
National
0.363 0.380 0.398
0.311 0.382 0.445
0.087 0.099 0.111
Region-specific 0.275 0.325 0.380
0.151 0.256 0.369
0.459 0.538 0.619
Idiosyncratic
0.238 0.293 0.343
0.221 0.333 0.446
0.284 0.363 0.443
Kyungnam
National
0.563 0.583 0.603
0.006 0.013 0.024
0.055 0.065 0.075
Region-specific 0.085 0.123 0.173
0.028 0.084 0.218
0.244 0.359 0.498
Idiosyncratic
0.242 0.290 0.328
0.758 0.888 0.946
0.439 0.577 0.691
Ulsan
National
0.684 0.702 0.719
0.259 0.301 0.344
0.351 0.369 0.387
Region-specific 0.026 0.071 0.129
0.034 0.108 0.243
0.017 0.048 0.123
Idiosyncratic
0.167 0.220 0.259
0.434 0.550 0.631
0.504 0.569 0.604
Daegu
National
0.727 0.744 0.761
0.086 0.113 0.144
0.096 0.108 0.119
Region-specific 0.039 0.067 0.102
0.038 0.105 0.254
0.097 0.179 0.291
Idiosyncratic
0.148 0.183 0.212
0.600 0.745 0.828
0.604 0.715 0.792
Kyungbuk
National
0.702 0.718 0.734
0.028 0.057 0.094
0.280 0.301 0.323
Region-specific 0.004 0.009 0.024
0.523 0.609 0.700
0.011 0.058 0.147
Idiosyncratic
0.239 0.259 0.279
0.231 0.319 0.400
0.548 0.621 0.667
Jeonbuk
National
0.492 0.509 0.525
0.107 0.134 0.164
0.169 0.181 0.193
Region-specific 0.327 0.357 0.389
0.324 0.437 0.559
0.552 0.601 0.650
Idiosyncratic
0.101 0.132 0.162
0.294 0.415 0.531
0.171 0.219 0.268
Kwangju
National
0.224 0.239 0.254
0.292 0.338 0.385
0.087 0.097 0.108
Region-specific 0.505 0.553 0.601
0.098 0.204 0.350
0.570 0.623 0.681
Idiosyncratic
0.158 0.207 0.254
0.298 0.438 0.536
0.223 0.279 0.332
Jeonnam
National
0.420 0.436 0.452
0.330 0.409 0.476
0.001 0.003 0.005
Region-specific 0.063 0.111 0.188
0.039 0.105 0.236
0.113 0.207 0.347
Idiosyncratic
0.374 0.448 0.497
0.317 0.425 0.521
0.650 0.789 0.883
Daejeon
National
0.674 0.691 0.706
0.081 0.103 0.127
0.048 0.055 0.063
Region-specific 0.047 0.070 0.105
0.024 0.102 0.306
0.246 0.401 0.567
Idiosyncratic
0.198 0.232 0.260
0.578 0.765 0.847
0.379 0.544 0.699
Chungbuk
National
0.767 0.784 0.800
0.507 0.590 0.659
0.249 0.265 0.281
Region-specific 0.006 0.020 0.050
0.030 0.076 0.152
0.038 0.112 0.245
Idiosyncratic
0.153 0.183 0.206
0.203 0.282 0.367
0.488 0.618 0.686
Chungnam
National
0.783 0.797 0.811
0.339 0.396 0.458
0.498 0.520 0.541
Region-specific 0.041 0.061 0.085
0.058 0.147 0.293
0.107 0.164 0.228
Idiosyncratic
0.117 0.139 0.159
0.251 0.386 0.500
0.259 0.319 0.366
Kwangwon National
0.682 0.700 0.717
0.500 0.565 0.623
0.248 0.265 0.282
Region-specific 0.029 0.058 0.098
0.114 0.172 0.239
0.060 0.129 0.233
Idiosyncratic
0.195 0.233 0.265
0.163 0.231 0.305
0.504 0.607 0.667
Jeju
National
0.102 0.112 0.123
0.331 0.376 0.421
0.031 0.037 0.044
Region-specific 0.379 0.486 0.594
0.321 0.410 0.489
0.268 0.328 0.407
Idiosyncratic
0.293 0.400 0.507
0.132 0.203 0.286
0.554 0.634 0.694
Notes: This table reports the variance decomposition for excess reallocation of total credit for non chaebolaffiliated firms by 16 regions (9 provinces and 7 metropolitan cities).

Table 2.9: Variance Decomposition (Nonl-affiliated firms)

52

Var(yi jt ) = (βinj )2Var(Ft ) + (βirj )2Var( fitr ) +Var(εi jt )

θinj =

(βinj )2Var(Ft )
Var(yi jt )

(2.5)

(2.6)

Tables 2.5 to 2.7 respectively contain the variance decomposition for excess reallocation of
total credit, loans and bonds in the 16 regions.5 Table 2.5 uncovers evidence that the common
factor explains a large fraction of the volatility in regional excess reallocation rates. It accounts
for more than 40% of the variation in the flow in the 12 regions (Seoul, Kyunggi, Incheon, Busan,
Daegu, Kyungnam, Kwangju, Jeonbuk, Jeonnam, Daejeon, Chungbuk, and Chungnam). In rural
or underdeveloped areas (Kangwon and Jeju), the idiosyncratic components account for a large
fraction of the variation in the reallocation rates. An interesting finding is that the idiosyncratic
component explains a large fraction of the variation in Ulsan. The city is the most industrialized
city in which several large companies including Hyundai Auto and Hyundai Shipping are located.
In addition, the common factor is the most strong driving force of regional excess loan reallocation
rates (see Table 2.6). By contrast, Table 2.7 shows that idiosyncratic components capture the
greatest share of excess bond reallocation fluctuations.
Tables 2.8 and 2.9 respectively report the variance decomposition for excess reallocation based
on chaebol-affiliated firms and non-affiliated firms. It is noteworthy that the idiosyncratic components are the main driving force for credit reallocation of chaebol-affiliated firms, while the
common factor plays a dominant role in driving credit reallocation of non-affiliated firms.
The variance decomposition results for the pre-crisis period (1981-1996) and the post-crisis
period (1999-2012) are also reported in Tables 2.5 to 2.9. Note that the results for the sub-sample
periods and the whole sample period are somewhat different, because factors and factor loads are
separately estimated for the two periods and the volatility of the observable variables are different
in the two periods. When we look at excess reallocation and gross reallocation in the two periods
separately, the fluctuation in the two variables are not huge, hence different results came out.
5 The results for gross reallocation are available on request.

53

2.6 Conclusion
We explored the evolution of credit reallocation from the geographical location perspective over
the 1981–2012 period. The hypothesis that has motivated this chapter is that the 1997 financial
crisis and the subsequent corporate and financial reforms can change the driving forces behind
evolution of credit reallocation across non-financial firms in Korea. We employed a Bayesian
dynamic latent factor model to decompose credit reallocation fluctuations into national, regionalspecific and idiosyncratic allocation components.
We uncovered evidence that the common factor explaining regional reallocation flows of total credit across all the regions increased after the crisis. It was highly positively correlated with
national excess credit reallocation, while it was negatively correlated with national gross reallocation. The factor turned out to be correlated with other credit flows. It was positively correlated
with national credit destruction, while it was negatively correlated with national credit creation and
net credit growth. However, it did not exhibit clear cyclicality; it was mildly counter-cyclical. This
pattern also applied to loan reallocation.
Some interesting patterns emerged when we investigated the roles played by the national,
region-specific and idiosyncratic components in driving the fluctuation of regional credit reallocation rates. When it comes to regional excess reallocation rates of total credit and loans, the
common factors explained a large fraction of the volatility in the rates. By contrast, idiosyncratic
components contributed a large fraction of the volatility of regional bond excess reallocation rates,
while the common factor played only a minor role in explaining the volatility in the flows. When
we examined the variance decomposition for excess reallocation based on chaebol-affiliated firms
and non-affiliated firms. It was noteworthy that the idiosyncratic components were the main driving force for credit reallocation of chaebol-affiliated firms, while the common factor played a
dominant role in driving credit reallocation of non-affiliated firms.
The analysis raises interesting topics for future research. Using other categories rather than
geographical location would be useful to shed lights on what the driving forces behind credit reallocation are. For example, using the 2-digit industry category is very interesting to explore the
54

evolution of credit reallocation, because industry sectors in the Korean economy exhibit asymmetric and heterogeous responses to aggregate shocks.

55

CHAPTER 3
RELIGION AND BANK PERFORMANCE: EVIDENCE FROM CREDIT UNIONS IN
KOREA

3.1 Introduction
Religion is an important institution and culture that affects economic performance (Weber
1930). The economics of religion literature shows that religion has influences on economic growth,
a countrys legal system and economic attitudes (Iannaccone, 1998; Barro and McCleary, 2003,
2006; Guiso et al., 2003; Stulz and Williamson, 2003). In the economics of education, a lot
of attention has been paid to the effect of Catholic high school attendance on students academic
achievement (Bryk et al., 1993, Altonji et al, 2005; Cohen-Zada and Elder, 2009). Likewise, religion may influence bank performance. To my knowledge, the empirical banking literature has not
explored the relationship between religion and bank performance. To fill the gap in this literature,
this article uses data on credit unions in Korea for the period 2000 to 2007 to investigate the effects
of religion on bank performance. A credit union, based on a region, is a member owned depository institution controlled by its members. Some credit unions are owned by religious institutions
such as the Catholic Church, protestant churches and Buddhist temples. Unique characteristics
of religion may affect the performance of the Religious Credit Unions (hereafter RCUs), such as
loan default rate and profit. This article is organized as follows: Section 1 provides the research
motivation; Section 2 explores the specific mechanisms by which RCUs may achieve better performance; Section 3 introduces data and methodologies; Section 4 shows empirical results and
Section 5 concludes.

56

3.2 Sources of Better Performance of RCUs
The RCUs have unique features, which create specific mechanisms by which they can achieve
better performance. Above all, potential clienteles of the RCUs are not random because their
potential customers are members of their religious institution who have not joined credit unions
yet. In fact, a religious institution is a strong common bond, which is so advantageous that it can
reduce the cost of assessing the creditworthiness of potential borrowers as well as the marketing
costs. Rather than attracting customers through standard marketing channels, they can achieve
this through private channels, which contribute to a reduction in costs. However, non RCU should
put efforts to gather customers living within an operation area. In addition, RCUs can screen and
monitor closer than other credit unions because they have several channels to acquire information
about their customers. For instance, the employees and customers of a RCU regularly attend the
same religious institution and some participate in religious activities such as volunteering groups
and prayer meetings. Through these channels, they can acquire extensive information, such as
economic attitudes, sincerity and family history. Hence, RCUs are able to utilize richer information
in making decisions on loan approval, withdrawal and renewal. But ordinary credit unions can
hardly get this type of information. On the borrower side, borrowers may have strong incentives
to repay their RCU loans. Borrowers who are in default on the RCU loans face the more severe
damage than those of ordinary credit unions because their reputation in a religious society would
be marred and their religious activity can be affected by loan default. In other words, borrowers
have a reputational incentive to pay back their loans to RCUs. Thus, it is possible that disutility
of RCU loan default is big enough to make an expected net utility of loan default less than that
of repayment. These features that facilitate RCUs to make more profits and suffer less from loan
default are not only reflected in RCUs in itself but also work through specific channels. First,
the number of members per employee accounts for the advantage of rich soft information, closer
screening and monitoring because the more employees are the more private information they get.
As this variable increases, the advantage that results in lower loan default rate weakens. Next,
sales and advertising costs have a different meaning for RCUs and non RCUs. RCUs, thanks to
57

Total Sample
(n=5,063)
Default rate 5.592
ROE
0.064
ROA
0.004
Size
9.806
Capital
0.076
S1
262.045
S2
31.210

RCUs
Non RCUs
(n=4,603) (n=460)
4.008
6.146
0.06
0.065
0.006
0.004
8.952
9.891
0.112
0.073
148.342
273.408
30.787
31.252

t-test
-9.237
-0.737
6.047
-23.432
16.429
-19.577
-1.057

Table 3.1: Summary Statistics

non random clienteles, do not need to spend marketing costs as much as non RCUs do. Saving
the costs is an important source of profits. The lower this ratio is, the more profits RCUs make.
However, the opposite is true for ordinary credit unions because they should spend marketing costs
to gain customers and to make profits.

3.3 Data and Model Specification
The data on credit unions come from the Financial Supervisory Service which gathers annual
reports from credit unions. Data about regional economies are obtained from the Bank of Korea.
There are three types of credit unions: regional, occupation and work place. I exclude credit
unions based on jobs and work places because the performance of these types is correlated with
which occupation or work place they are based on. The sample includes regional credit unions
which include RCUs. The RCUs are also based on a region as religious institutions are. Thus,
religion is the sole difference between a RCU and an ordinary credit union which are located in the
same region. The sample starts from 2000 to avoid the effects of the financial crisis that occurred in
1997. For a similar reason, the sample ends in 2007 because massive savings bank failures started
in 2008 and have lasted until 2012. Eventually, the final sample consists of 5,063 observations
of 816 credit unions among which 76 credit unions are RCUs: 35 Catholic, 36 protestant and 5
Buddhist. The model specification is as follows.

58

Yit = β0 + β1 Xit + β2 religion + β3 Sit ∗ religion + β4 region j + β5 yeart + εit

(3.1)

where Yit is loan default rate, ROE or ROA for credit union i at year t. A set of Xit includes
natural logarithm of total assets (Size) and core captial divided by total assets (Capital). A set of Sit
comprises the number of members per employees (S1) and proportion of marketing costs (S2) and
religion is a dummy variable that equals 1 for RCUs, otherwise 0. Region j includes real GRDP
growth rate and business bankruptcy rate for region j. Yeart represents year fixed effects and εit
is an error term. Table 3.1 reports summary statistics for the variables used in this study and t test
results.

3.4 Empirical Results
The empirical results are reported in Table 3.2. Column 1 shows that RCUs suffer from troubled
loans 5% less than ordinary credit unions. As the number of members per employee increases, loan
default rate also increases for both types of credit unions. However, it increases more for RCUs
than others because RCUs advantage of soft information weakens. This finding can be interpreted
that RCUs keep low loan default rate through relation oriented banking in which employees play
key roles. However, the extent that the positive effect of religion is offset by an increase in the ratio
is relatively small. Holding other things fixed, the positive effect is completely offset when S1
increases by 593.89. Given the fact that the range of the variable is from 22.16 at the 1% percentile
to 492.43 at the 99% percentile, RCUs in most cases benefit from the positive effect of religion,
although the effect differs depending on the variable.
In addition, an increase in sales and advertising costs helps non RCUs lower loan default rate
because these costs include acquiring customer information and monitoring. Interestingly, it has no
effect on lowering troubled loans for RCUs that rely on soft information acquired through private
channels. For robust check, I transform the loan default rate into a proportion by dividing it by
100 and do fractional logit analysis using a QMLE approach. Column 4 indicates qualitatively

59

Default Rate

ROE

ROA

Default Rate
(Fractional Logit)
Size
-0.772***
-0.001
0.126*** -0.007***
(0.005)
(0.886)
(0.000)
(0.000)
Capital
-0.011
-0.004*** 0.038*** -0.0004**
(0.698)
(0.000)
(0.000)
(0.000)
Religion
-5.345***
0.090*** 0.619*** -0.012***
(0.000)
(0.000)
(0.000)
(0.000)
S1
0.005***
0.000*
-0.0002** -0.0001***
(0.000)
(0.063)
(0.017)
(0.000)
S1*Relgion
0.009***
0.000
0.0001
0.0001***
(0.004)
(0.913)
(0.705)
(0.005)
S2
-0.080**
0.001**
0.008*** -0.0004***
(0.083)
(0.040)
(0.000)
(0.027)
S2*Religion
0.078**
-0.002*** -0.018*** 0.0004
(0.032)
(0.000)
(0.000)
(0.237)
GRDP (%)
-0.054**
-0.000
0.004
-0.0002
(0.015)
(0.755)
(0.108)
(0.484)
Bankruptcy (%) 2.275***
-0.054*** -0.201*** 0.028***
(0.000)
(0.006)
(0.002)
(0.000)
Constant
13.259***
0.051
-1.278***
(0.000)
(0.335)
(0.000)
Observations
5,063
5,063
5,063
5,063
0.231
0.064
0.246
0.095
R2
Log likelihood
-850.01
Notes: *, ** and *** denote statistical significance at the 10%, 5%, and 1%
level. The numbers in parentheses indicate p value. column 4 indicates average
marginal effects.
Table 3.2: Empirical Results

same results to Column 1. Turning to profitability, Column 2 and Column 3 show that RCUs
enjoy higher profitability relative to non RCUs. Interestingly, the proportion of marketing costs has
negative effects on profits of RCUs, whereas it has positive effects on non RCUs profits. Consistent
with Section 2, non RCUs should spend on marketing costs to attract deposits and to make profits;
by contrast, RCUs should save the costs to improve profitability, which is possible due to the
non random potential clientele. Holding other things fixed, the positive effect of religion on ROA
(ROE) eventually disappears when a RCU increases the proportion of marketing costs by 34.39
(45.00)%. Given that the proportion ranges from 12.20 at the 1% percentile to 66.67 at the 99%

60

percentile, most RCUs benefit the positive effect of religion irrespectively of their expenditure on
marketing.

3.5 Conclusion
This article found evidence that RCUs do not only suffer less from troubled loans but they also
enjoy higher profits relative to ordinary ones. The features intrinsic to RCUs, such as non random
potential clientele, rich soft information and large disutility of loan default, are likely to be what
enables RCUs to outperform. The findings of this study have important policy implications on
microfinance institutions and cooperative banks, particularly in emerging economies where these
institutions play important roles in their banking industries; a policy promoting cooperative banks
based on a religious institution can reduce social costs derived from loan defaults and let their
members share more profits. This paper is a preliminary step to explore the relationship between
religion and bank performance. Later, I will add other possible variables which can mitigate a
potential endogeneity problem or a omitted variable problem.

61

BIBLIOGRAPHY

62

BIBLIOGRAPHY

[1] Allen, F., Babus, A., Carletti, E., 2010. Financial crises: theory and evidence, Annual Review
of Financial Economics 1, 97-116.
[2] Altonji, J.G., Elder, T., Taber, C.R., 2005. Selection on observed and unobserved variables:
assessing the effectiveness of Catholic schools. Journal of Political Economy 113, 151184.
[3] Anas, A., Arnott, R., Small, K.A., 1998. Urban spatial structure. Journal of Economic Literature 36, 1426-1464.
[4] Bank of Korea, 2003. Financial Stability Report, Bank of Korea, April.
[5] Banker, R.D., Chang, H., Lee, S., 2010. Differential impact of Korean banking system reforms on bank productivity. Journal of Banking and Finance 34, 1450−1460.
[6] Barro, R.J., McCleary, R.M., 2003. Religion and economic growth across countries. American Sociological Review 68, 760781.
[7] Barro, R.J., McCleary, R.M., 2006. Religion and economy. Journal of Economic Perspective
20, 4972.
[8] Bazdresch, S., 2013. The role of non-convex costs in firms’ investment and financial dynamics. Journal of Economic Dynamics and Control 37, 929−950.
[9] Beck, T., Levine, R., Loayza, N., 2000. Finance and the sources of growth. Journal of Financial Economics 58, 261−300.
[10] Berger, A.N., Clarke, G.R.G., Cull, R., Klapper, L., Udell, G.f., 2005. Corporate governance
and bank performance: A joint analysis of the static, selection, and dynamic effects of domestic, foreign, and state ownership. Journal of Banking and Finance 29, 2179−2221.
[11] Bernanke, B., Gertler, M., Gilchrist, S., 1999. The financial accelerator in a quantitative
business cycle framework. Handbook of Macroeconomics 1, 1341−1393.
[12] Bertola, G., 1993. Models of economic integration and localized growth. In:Torres, F., Giavazzi, F. (Eds.), Adjustment and Growth in the EMU. Cambridge University Press, Cambridge, 159-179.
[13] Blalock, G., Gertler, P., Levine, D.I., 2008. Financial constraints on investment in an emerging market crisis. Journal of Monetary Economics 55, 568−591.
[14] Bordo, M.D., Haubrich, J.G., 2010. Credit crises, money and contractions: An historical
view. Journal of Monetary Economics 57, 1−18.
[15] Borensztein, E., Lee, J., 2002. Financial crisis and credit crunch in Korea: evidence from
firm-level data. Journal of Monetary Economics 49, 853−875.
63

[16] Borensztein, E., Lee, J., 2005. Financial reform and the efficiency of credit allocation in
Korea. Journal of Policy reform 8, 55−68.
[17] Bryk, A.S., Lee, V.S., Holland, P.B., 1993. Catholic schools and the common good. Harvard
University Press, Cambridge.
[18] Caballero, R., Hammour, M., 2001. Institutions, restructuring, and macroeconomic performance. In: Advances in Macroeconomic Theory, Dreze, J.(Ed.), Palgrave New York.
[19] Caballero, R., Hammour, M., 2005. The costs of recessions revisited: A reverse-liquidationist
view. Review of Economic Studies 72, 313−341.
[20] Caballero, R., Hoshi, T., Kashyap, A., 2008. Zombie lending and depressed restructuring in
Japan. American Economic Review 98, 1943−77.
[21] Chang, H., 2006. The East Asian development experience: the miracle, the crisis and the
future. Third World Network.
[22] Chen, K., Irarrazabal, A., 2012. Misallocation and the recovery of manufacturing TFP after a
financial crisis. Working paper, Emory University.
[23] Claessens, S., Kose, M.A., Terrones, M. E., 2012. How do business and financial cycles
interact? Journal of International Economics 87, 178−190.
[24] Cohen-Zada, D., Elder, T., 2009. Historical religious concentrations and the effects of
Catholic schooling. Journal of Urban Economics 66, 6574.
[25] Covas, F., Den Haan, W., 2011. The cyclical behavior of debt and equity finance. American
Economic Review 101, 877−899.
[26] Cruicini, M.J., Kose, M.A., Otrok, C., 2011. What are the driving forces of international
business cycles? Review of Economic Dynamics. 14, 156-175.
[27] Davis, S.J., Faberman, R.J., Haltiwanger, J.C., 2006. The flow approach to labor markets:
New data sources and micro-macro links. Journal of Economic Perspectives 20, 3−26.
[28] Davis, S.J., Haltiwanger, J.C., 1992. Gross job creation, gross job destruction, and employment reallocation. Quarterly Journal of Economics 107, 819−863.
[29] Davis, S.J., Haltiwanger, J.C., 1999. On the driving forces behind cyclical movements in
employment and job reallocation. American Economic Reivew 89, 1234-1258.
[30] Dell’Ariccia, G., Garibaldi, P., 2005. Gross credit flows. Review of Economic Studies 72,
665−685.
[31] Dell’Ariccia, G., Igan, D., Laeven, L., Tong, H., 2012. Policies for macrofinancial stability:
How to deal with credit booms. IMF Staff Discussion Note, Washington, D.C.
[32] Dunchin, R., Ozbas, O., Sensoy, B., 2010. Costly external finance, corporate investment, and
the subprime mortgage credit crisis. Journal of Financial Economics 97, 418−435.

64

[33] Eisfeldt, L., Rampini, A., 2006. Capital reallocation and liquidity. Journal of Monetary Economics 53, 369−399.
[34] Eisfeldt, L., Muir, T., 2013. Aggregate issuance and savings waves. Working paper, University of California, Los Angeles.
[35] Galindo, A., Schiantarelli, F., Weiss, A., 2007. Does financial liberalization improve the allocation of investment? Micro-evidence from developing countries. Journal of Development
Economics 83, 562−587.
[36] Gilchrist, S., Sim, J., Zakrajsek, E., 2013. Misallocation and financial market frictions: some
direct evidence from the dispersion in borrowing costs. Review of Economic Dynamics 16,
159−176.
[37] Gourinchas, P.O., Obstfeld, M., 2012. Stories of the twentieth century for the twenty-first.
American Economic Journal: Macroeconomics 4, 226−65.
[38] Gourio, F., Kashyap, A.K., 2007. Investment spikes: New facts and a general equilibrium
exploration. Journal of Monetary Economics 54S1, 1−22.
[39] Guiso, L., Sapienza, P., Zingales, L., 2003. Peoples opium? religion and economic attitudes.
Journal of Monetary Economics 50, 225282.
[40] Haggard, S., Lim, W., Kim, E., 2010. Economic Crisis and Corporate Restructuring in Korea: Reforming the Chaebol. Cambridge Asia-Pacific Studies, Cambridge University Press,
Cambridge. United Kingdom.
[41] Herrera, A.M., Kolar, M., Minetti, R., 2011. Credit reallocation. Journal of Monetary Economics 58, 551−563.
[42] Iannaccone, L.R., 1998. Introduction to the economics of religion. Journal of Economic Literature 36, 14651496.
[43] IMF, 2004. Are credit booms in emerging markets a concern" In: World Economic Outlook,
April 2004: Advancing Structural Reforms, Washington, D.C.
[44] Joh, S., 2003. Corporate governance and firm profitability: evidence from Korea before the
economic crisis. Journal of Financial Economics 63, 287−322.
[45] Kashyap, A., Stein, J., Wilcox, D., 1993. Monetary policy and credit conditions: evidence
from the composition of external finance. American Economic Review 83, 78−97.
[46] Kim, M., Maksimovic, V., 1990. Debt and input misallocation. Journal of Finance 3,
795−815.
[47] Korajczyk, R., Levy, A., 2003. Capital structure choice: macroeconomic conditions and financial constraints. Journal of Financial Economics 68, 75−109.
[48] Kose, M.A., Otrok, C., Whiteman, C.H., 2003. International business cycles: world, region,
and country-specific factors. American Economic Review 93, 1216-1239.
65

[49] Kose, M.A., Otrok, C., Whiteman, C.H., 2008. Understanding the evolution of world business
cycles. Journal of International Economics 75, 110-130.
[50] Lang, W., Nakamura, L.I., 1995. Flight to quality in banking and economic activity. Journal
of Monetary Economics 36, 145−164.
[51] Lee, J., Rhee, C., 2007. Crisis and recovery: what we have learned from the South Korean
experience? Asian Economic Policy Review 2, 146−164.
[52] Levy, A., Hennessy, C., 2007. Why does capital structure choice vary with macroeconomic
conditions? Journal of Monetary Economics 54, 1545−1564.
[53] Lim, K., 2010. Structural fundamentals of Korean corporations: this time was different. KDI
Conference.
[54] Liu, D.C. 2013. The evolution of excess job reallocation in the U.S. Journal of Macroeconomics 36. 188-206.
[55] Martin, R., Sunley, P., 1998. Slow convergence? The new endogenous growth theory and
regional development. Economic Geography 74, 201-227
[56] Mendoza, E., Terrones, M., 2012. An anatomy of credit booms and their demise. Working
Papers Central Bank of Chile 670.
[57] Midrigan, V., Xu, D.Y., 2013. Finance and misallocation: Evidence from plant-level data.
American Economic Review, forthcoming.
[58] Neumeyer, P.A., Sandleris, G., 2010. Understanding productivity during the Argentine crisis.
Business School Working Papers, Universidad Torcuato Di Tella.
[59] Nilsen, J.H., 2002. Trade credit and the bank lending channel. Journal of Money, Credit and
Banking 34, 226−253.
[60] Oliner, S.D., Rudebush, G.D., 1995. Is there a bank lending channel for monetary policy?
Economic Review, FRB of San Francisco 2, 1−20.
[61] Park, Y., Lee, J., 2003. Recovery and sustainability in east Asia. In: Dooley, M., Frankel, J.
(Eds.), Managing Currency Crises in Emerging Markets.
[62] Rajan, R., Zingales, L., 1995. What do we know about capital structure? Some evidence from
international data. Journal of Finance 50, 1421−1460.
[63] Stock, J. H., Watson, M.W., 2003. Has the business cycle changed and why? In: NBER
Macroeconomics Annual 2002, Volume 17, 159-230, MIT Press.
[64] Stulz, R.M., Williamson R., 2003. Culture, openness, and finance. Journal of Financial Economics 70, 313349.
[65] The Economist, 2012. Deleveraging. The bad, the Good, and the Ugly. May 2nd.

66

[66] Törnqvist, L., Vartia, P., Vartia, Y.O., 1985. How should relative changes be measured?
American Statistician 39, 43−46.
[67] Weber, M., 1930. The protestant ethic and the spirit of capitalism. Harper Collins, New York.
[68] World Bank, 2000. East Asia: recovery and beyond.
[69] Wurgler, J., 2000. Financial markets and the allocation of capital. Journal of Financial Economics 58, 187−214.

67