< '-< A
*nLu‘Tﬁ'n

 

A

'o

3
Eu

‘5

-: t v
.1 \ $4]
'9:

, A-ﬁt

J 5
1’ ii}

a
3

($.2“§." :
F31: 3‘

RM:

5...;

w
”owe-f
"‘3'" "

.n

 

Q n’
‘35;

i.-
5‘13!

THESIS

. .4
I ' ’
at J L! H

b035/Qéa

This is to certify that the
dissertation entitled

THREE ESSAYS IN FINANCIAL LIBERALIZATION

presented by
Iva Krasteva Petrova

has been accepted towards fulﬁllment
of the requirements for the

Ph.D. d in Economics

A

 

 

V/ W6? Professor’s Signature

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

 

 

LIBRARY

Michigan State

University

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

76512391021105

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJClRCJDatoDuepes-sz

THREE ESSAYS IN FINANCIAL LIBERALIZATION
By

Iva Krasteva Petrova

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Economics

2004

ABSTRACT
THREE ESSAYS IN FINANCIAL LIBERALIZATION
By

Iva Krasteva Petrova

Financial liberalization throughout the world is associated with a rapid increase in
lending through banks’ subsidiaries abroad. “Financial Liberalization and the
Representation Choice of a Foreign Bank” uses Horstmann and Markusen’s (1996)
framework to develop a theoretical model of banking that explains the capital allocation
choice of a foreign investor in a host bank market. A foreign bank has two approaches to
maintain borrower-lender relationships with borrowers in the host country. It can invest
through the domestic banking system and delegate monitoring responsibility to the local
bank management, or it can open a subsidiary in the domestic market and internalize its
monitoring responsibilities. The paper argues that the risk of negligent monitoring by the
domestic bank’s management may encourage foreign investors to increase their
ownership in a host country banking system. Additional incentives for the foreign
investor to enter the banking market include monitoring cost advantage of the foreign
investor and small capital endowment of the host economy. The paper ﬁnds that foreign
bank entry is beneﬁcial in a highly productive host economy with an incumbent bank that
can sustain rising pressure of monitoring costs.

Financial liberalization resulted in signiﬁcant credit dollarization in developing
and transition economies. “Credit Dollarization in Transition Economies: Is it Firms’

or Banks’ “Fault”?” investigates the respective contributions of banks and ﬁrms to the

dollarization of credit. Whether dollarization comes mainly from the ﬁnancial sector or
from the real sector has important policy implications. First, it indicates whether there
are balance sheet mismatches in the economy. Second, it shows in which sector of the
economy they are concentrated, and which sector (if any) policy makers should target in
order to contain the negative effects of exchange rate depreciation. This paper uses a new
dataset for twenty-two transition economics for the period 1990 - 2001. It separates the
contributions of banks and ﬁrms to credit dollarization by grouping potential
determinants into bank- and ﬁrm-speciﬁc factors. Empirical results provide evidence that
bank currency matching is the main driving force of credit dollarization. However,
currency mismatches tend to be concentrated in the real sector, and this indirectly
exposes the economy to ﬁnancial crises.

“Currency Mismatches and Banking Crises” is a continuation of the second
dissertation chapter. If banks do not match the currency structure of their assets and
liabilities, they face exchange rate risk. However, if the currency structure of their assets
and liabilities is identical, banks pass the currency risk onto their borrowers. In the end,
banks will bear the costs indirectly because they face a higher default risk if the domestic
currency depreciates. This paper uses a simple theoretical framework to Show that bank
runs are determined by currency mismatches in the banking and the real sectors, and
unexpected exchange rate depreciation. The econometric results present some evidence
that the persistence of banking crises in transition economies is reinforced by too much
lending in foreign currency. There is also evidence that banks face direct exchange rate

risk even though their open currency positions are very limited.

ACKNOWLEDGEMENTS

I commence these acknowledgements with a thank you to my major advisor, Dr.
Rowena A. Pecchenino, for her guidance and encouragement throughout the past ﬁve
years. It is difﬁcult to express my immense gratitude for her support and dedication
throughout my Ph.D. program. I also wish to thank my dissertation committee, Dr. Susan
J. Linz, Dr. John D. Wilson, Dr. Jeffrey M. Wooldridge, and Dr. Wei-Lin Liu for their
invaluable advice.

I am very much indebted to my friend and co-author, Alina C. Luca. This study
took many positive turns after her critical proofreading and during many of our daily
discussions.

I am also grateful to the participants of the 2002 Red Cedar Conference organized
by the MSU Department of Economics, the 15th European Advanced Studies Institute in
Regional Science in Eksjo, Sweden, June 10-16, 2002, and the 8th Meeting of the Latin
American and Caribbean Economic Association in Puebla, Mexico, October 9-12, 2003.
Their comments and suggestions had a big impact on this ﬁnal product.

A special thanks goes to the economics graduate advisor Dr. Thomas D.
Jeitschko, to Dr. Carl E. Liedholm, and all my professors and friends at Michigan State
University for their constant help and priceless advice. I also want to say thank you to
my old friends from the Bulgarian National Bank, Kalin Hristov and Nikolay Nenovski.
Their contagious passion for economics and encouragement persuaded me to seek Ph.D.

studies in the US.

iv

Finally, I am grateful to my parents, Nada and Krastu Krastevi, for supporting all
my endeavors. My husband, William H. Shields, deserves as much credit for this work
as I do for his proofreading, formatting, patience, devotion, and babysitting over the last
year. I dedicate this dissertation to my one-year old son Peter who, despite the little time

we spent together, still recognizes his mom.

PREFACE

Financial liberalization in this dissertation concerns the banking system and
indicates that banks are decentralized and allowed to exist as proﬁt maximizing entities in
attracting deposits, extending loans, and providing information and other ﬁnancial
services. The most recent and striking examples of ﬁnancial liberalization occurred in
the economies of Eastern Europe and Central Asia. Two-tiered banking systems were
created from a centralized bank, the exchange rate, interest rates, and capital ﬂows were
liberalized, and banks were allowed to provide a range of ﬁnancial services rather than
simply direct credit to state-owned enterprises.

As ﬁnancial liberalization created new proﬁt opportunities for foreign and
domestic investors, it also raised interesting questions about investors’ opportunities to
reduce market risk. This dissertation focuses on investors’ choices associated with two
types of uncertainty: (i) project related -since foreign investors know little about local
ﬁrms; (ii) macro risk related to the stability of the domestic currency.

The ﬁrst type of uncertainty raises the issue of how domestic borrowers should be
evaluated. Should foreign investors employ local knowledge and delegate evaluation of
the ﬁnanced projects to the domestic banking system? Or should they internalize the
agency costs by opening their own subsidiaries and using their skills in project
monitoring?

The second question relates to the currency of denomination (domestic or foreign)
in which credit to domestic borrowers is extended. Since foreign ﬁnancing is mostly

denominated in foreign currency, the host economy absorbs the exchange rate risk.

vi

However, is it banks or borrowers who bear this risk? Is the banking system safer if
ﬁrms rather than the banks cover the exchange rate risk?

The ﬁrst dissertation essay, “Financial Liberalization and the Representation
Choice of a Foreign Bank”, explores the question of how foreign investors choose to
channel capital to a host economy in ways that minimize their information and entry
costs. This paper focuses on the foreign banks’ choice between correspondent banking
and opening a subsidiary in a host economy. Precisely what encourages foreign banks to
channel investments through the local banking system or with subsidiaries is a question
motivated by an interesting economic phenomenon «there is a clear difference in the
pattern of banks’ investments in developed and emerging economies. Correspondent
banking is the main channel for foreign banks’ investment in developed countries, while
emerging economies see a larger capital inﬂow from local subsidiaries of foreign banks
than from correspondent banking. However, developed economies have very low
barriers to entry in the banking system as compared to developing economies. The
question then is why do we see multinational banks entering emerging economies’
banking systems and being reluctant to do the same in developed economies?

The choice between internalization and delegation has been explored in detail for
the case of a manufacturing company, especially in relation to the behavior of
multinational corporations. A multinational company may consider it more costly to
initiate production in a host country than an incumbent company because of stringent
entry requirements or due to an informational disadvantage about the market. On the
other hand, the multinational company may delegate representation to a local contractor

and exploit his exclusive information about the market. However, delegation creates the

vii

risk that the agent may choose to serve his own interests ﬁrst. A cost-beneﬁt analysis of
the two options may prove that opening a subsidiary in the host country yields higher
proﬁt for the foreign company than contracting with a local agent. Thus, the foreign
company has an incentive to enter a host market with ownership. Similar to a
multinational corporation, a foreign investor (bank) faces the choice to lend to the host
country using correspondent relationships with a domestic bank or opening a subsidiary
bank. However, current multinational banking theory is unable to explain this choice.

This paper addresses the following questions. First, how do foreign investors
choose the organizational form of representation in a host country, speciﬁcally between
subsidiary and correspondent banking? Second, what is the effect of this choice on
investment and output in the domestic country? Third, what welfare gains are generated
by the host economy from foreign bank entrance and correspondent banking?

The paper ﬁnds that the foreign investor will choose to enter the domestic
banking market when agency costs are relatively higher than competition costs, set up
costs are low, and the host economy has a small capital endowment. However, if the
foreign investor is at disadvantage in monitoring the borrowers, he will prefer
correspondent banking. The analysis shows that correspondent banking with asymmetric
information leads to more volatile capital ﬂows than correspondent banking with perfect
information. This is consistent with the fact that correspondent banking capital flows are
more volatile in emerging (where asymmetric information is a serious problem) than in
developed economies. Foreign entrance brings higher volume of capital ﬂows than
correspondent banking, but the paper is inconclusive whether these ﬂows are more

volatile or not.

viii

The paper also concludes that the domestic economy beneﬁts from foreign bank
entrance if the foreign investor does not exploit a monitoring advantage that rapidly
depletes domestic banks’ proﬁts and if the domestic economy enjoys high capital
productivity.

The second and third dissertation chapters “Credit Dollarization in transition
economies: Is it Firms’ or Banks’ “Fault”?” and “Currency Mismatches and Banking
Crises” redirect attention from capital ﬂows to the currency of denomination of domestic
credit. As emerging (including transition) economies lack domestic capital, they often
turn to international markets to ﬁnance production. Years of exchange rate instability
have also created persistently high deposit dollarization. Thus, banks channel resources
to domestic borrowers that are increasingly denominated in foreign currency and oﬁen
exceed the capacity of the tradable sector to absorb them. These resources create a
currency imbalance that increases exchange rate risks for banks and ﬁrms and potentially
raise the costs of government intervention during ﬁnancial crises.

Currency mismatches increase ﬁnancial fragility and make economies more prone
to banking crises. Because of high bank liquidity and solvency risks, banks face a higher
premium on foreign and domestic capital. This brings multiple and self-enforcing
problems to dollarized economies. On one hand, the risk premium worsens ﬁrms’
balance sheets and tightens their borrowing constraints, so overall credit ﬁnance and
economic growth are lower than in countries where currency mismatches do not exist.
On the other hand, foreign and domestic depositors are more sensitive to news about
exchange rate movements in economies with vast currency mismatches. The reason is

that depreciation directly and indirectly brings banks’ proﬁts down. Thus, dollarized

ix

economies tend to experience abrupt capital reversals and slow recovery from economic
downturns.

In “Credit Dollarization in transition economies: Is it Firms’ or Banks’ “Fault”?” I
emphasize that factors explaining credit dollarization also explain why grossly
disproportionate currency mismatches exist in small open economies with
underdeveloped ﬁnancial systems. I take a close look at the domestic credit market and
examine the risk sharing forces behind credit dollarization. Is credit dollarization the
outcome of a unilateral decision made by domestic banks? Or is it driven by domestic
ﬁrms’ integration in the international marketplace?

Currently, only two empirical studies, Arteta (2002) and Barajas and Morales
(2003), analyze the determinants of credit dollarization. I extend the existing work on
credit dollarization by using a comprehensive, newly-assembled dataset for twenty-two
transition economics from Central and Eastern Europe and Central Asia, for the period
1990 — 2001. I estimate and compare the respective contribution of ﬁrms and banks to
the ﬁnancial dollarization phenomenon by grouping potential determinants into bank- and
ﬁrm-speciﬁc factors. Bank-speciﬁc factors include indicators of asset and liability
currency matching. As ﬁrm-speciﬁc factors, I use measures of real dollarization and
access to alternative ﬁnancing sources.

I estimate a reduced form equation, and use the pooled OLS estimator as a
benchmark, and ﬁrst difference and ﬁxed effects estimators to control for unobserved
heterogeneity. Empirical results provide evidence that credit dollarization in transition
economies is determined by banks’ optimization decisions (asset and liability

management variables). Banks match the currency of denomination of their assets and

liabilities, that is, deposit dollarization drives credit dollarization, and net foreign assets
represent a substitute for dollar loans to domestic ﬁrms. There is limited evidence of
ﬁrms hedging the currency of their debt and production returns.

However, as long as ﬁnancial dollarization outpaces real dollarization, and this
seems to be the case in transition countries, there is a currency mismatch somewhere in
the economy. I show that banks pass the exchange rate risk to ﬁrms. This decreases
banks’ exposure to currency risk, but it increases their exposure to default risk, and
ultimately the exposure of the economy to ﬁnancial crises.

The third dissertation chapter explores the link between currency mismatches and
banking crises in the countries of Eastern Europe and Central Asia. This paper builds on
the results from the second dissertation essay that identify the sector bearing the
concentration of exchange rate risk. It explores whether the concentration of exchange
rate risk through mismatched balance sheets has increased the probability of banking
crises and whether the transfer of risk away from the banking sector has reduced the
threat to the banks. If banks in transition economies have avoided direct exchange rate
risk, were they ultimately able to prevent the loss of net worth due to exchange rate
ﬂuctuations?

The paper models the case of small open economies, which are unable to borrow
externally in their own currency and have poor domestic currency credibility. Depositors
determine the currency composition of banks’ deposits and ﬁrms receive revenues in
foreign currency that depend on world prices and external demand. The equilibrium level
of credit dollarization depends on ﬁrms’ and banks’ expected revenues in foreign

currency and their ability to absorb exchange rate risk. A Diamond-Dybvig (1983) —

xi

style model describes the link between banks’ and ﬁrms’ optimal currency mismatches to
poor bank net worth and bank runs.

Currently a few studies have empirically explored the implications of
dollarization for ﬁnancial stability. Arteta (2003) and De Nicolo et al. (2003) look for a
direct causality from dollarization to bank losses and crises. Arteta (2003) does not ﬁnd
evidence that dollarization (deposit or credit) has any effects on banking stability or that
banking and currency crises are more costly if banks face higher dollarization ratios. De
Nicolo et a1. (2003) ﬁnd a direct relationship between dollarization and banking stability
but they only consider the effects of deposit dollarization. Instead, I argue that
dollarization is not harmful unless it creates a mismatch in either ﬁrms’ or banks’ balance
sheets and that mismatches prompt banking crises only in interaction with large
unexpected changes in the exchange rate.

The econometric analysis ﬁnds strong evidence that banks in transition economies
are exposed to default risk related to currency mismatches in ﬁrms’ balance sheets. With
an overall measure of banks’ foreign currency positions that includes foreign assets and
liabilities, the results show that in addition to default risk, banks also face direct exchange
rate risk, even though their open positions are generally very small. There is evidence
that foreign exchange and trade liberalization has increased, while institutional
development of transition economies banking systems has reduced, the likelihood of
banking crises.

The paper’s ﬁndings have an important policy implication with respect to
countries’ choice of exchange rate regime, or alternatively about their choice of

mechanisms to protect banks from currency risk. The results provide support for the

xii

Calvo and Reinhart’s (2002) “fear of floating” idea — the fact that governments are
unwilling to let the exchange rate move independently if the economy is heavily indebted
in foreign currency because of the high economic cost of exchange rate depreciation.
However, the tendency to reduce exchange rate ﬂuctuations may induce ﬁrms and banks
to maintain even larger currency imbalances, thus raising ﬁnancial crises costs further
yet.

The paper also shows that restrictions on open currency positions do not eliminate
the exchange rate risk for banks altogether, since banks transfer the unhedged exchange
rate risk onto their borrowers and face a higher default risk. This suggests that a market
approach to deal with exchange rate risk may work better to reduce the probability of

banking crises than banking regulations and ﬁxed exchange rate.

xiii

TABLE OF CONTENTS

PART ONE. FINANCIAL LIBERALIZATION AND THE REPRESENTATION

CHOICE OF A FOREIGN BANK ..................................................................................... 1

1. Introduction ............................................................................................................. 1

2. The Model ............................................................................................................... 5

2.]. Domestic Agents and a Foreign Investor ........................................................ 6

2.2. Timing of the Events and Equilibrium .......................................................... 10

2.3. Loan Contracts with Correspondent Banking ............................................... 11

2.4. Agency Contract with the Correspondent Domestic Bank ........................... 15

2.5. Entrance of the Foreign Investor in the Domestic Banking Market ............. 24

2.6. Foreign Entrance and Capital Flows ............................................................. 28

3. The Foreign Investor’s Choice .............................................................................. 31

4. Welfare Implications ............................................................................................. 33

5. Summary and Conclusions ................................................................................... 37

REFERENCES ................................................................................................................. 4O

APPENDICES .................................................................................................................. 42

Appendix 1. Proof of Proposition 1 ......................................................................... 42

Appendix 2. Proof of Proposition 2 ......................................................................... 48

Appendix 3. Proof of Proposition 5 ......................................................................... 53

Appendix 4. Proof of Statement ............................................................................... 54
PART TWO. CREDIT DOLLARIZATION IN TRANSITION ECONOMIES: IS IT

FIRMS’ OR BANKS’ “FAULT”? .................................................................................... 56

1. Introduction ........................................................................................................... 56

Continued on next page

xiv

- 2. Conceptual Framework and Main Hypotheses ..................................................... 62

2.]. Firms ............................................................................................................. 64

2.2. Banks ............................................................................................................. 65

3. Empirical Analysis ................................................................................................ 69

4. Robustness Tests ................................................................................................... 73

5. Conclusions and Future Extensions ...................................................................... 79
REFERENCES ................................................................................................................. 81
APPENDIX ....................................................................................................................... 84
PART THREE. CURRENCY MISMATCHES AND BANKING CRISES .................. 93
1. Introduction ........................................................................................................... 93

2. A Simple Model .................................................................................................... 99
2.1. Demand Driven Credit Dollarization (8:0) ................................................ 102

2.2. Bank Currency Asset and Liability Matching ............................................ 104

3. Econometric Model and Estimation Methods ..................................................... 105

4. Data ..................................................................................................................... 107
4.1. Variables Speciﬁed in the Theoretical Model ............................................ 108

4.2. Other Determinants of Banking Crises ....................................................... 109

5. Estimation Results .............................................................................................. 112
5.1. Alternative Speciﬁcations ........................................................................... 115

5.2. Endogeneity ................................................................................................ 120

6. Policy Implications and Conclusions .................................................................. 121
REFERENCES ............................................................................................................... 123
APPENDIX ..................................................................................................................... 126

XV

Table 2-1
Table 2-2
Table 2-3
Table 2-4
Table 2-5
Table 2-6
Table 2-7
Table 2-8
Table 3-1
Table 3-2
Table 3-3

Table 3-4

Table 3-5
Table 3-6
Table 3-7

Table 3-8

LIST OF TABLES

Data Deﬁnitions and Sources .................................................................... 84
Country and Dollarization Data Coverage ................................................ 86
Summary Statistics .................................................................................... 87
Correlation Matrix .................................................................................... 88
Basic Model Estimation Results ............................................................... 89
Extended Regressions ............................................................................... 90
Basic Model with Household Deposit Dollarization ................................ 91
Extended Regressions with Household Deposit Dollarization ................. 92
Data Deﬁnitions and Sources .................................................................. 126
Banking and Currency Crises in Transition Economies ......................... 129
Overall Effect of Real Depreciation and Currency Mismatches ............ 134
Effect of Real Depreciation and Currency Mismatches in the Existence of
a Currency Crisis ................................................................................... 137
Effect of Nominal Depreciation and Currency Mismatches ................... 140
Effect of Total Bank Currency Mismatch ............................................... 143

Overall Effect of Real Depreciation and Currency Mismatches—-
Alternative Speciﬁcation ....................................................................... 144

Effect of Real Depreciation and Currency Mismatches in the Existence of
a Currency Crisis - Alternative Speciﬁcation ....................................... 147

xvi

LIST OF FIGURES

Figure 1-1: Correspondent Banking with Asymmetric Information ................................. 18
Figure 1-2: Correspondent Banking with Incentive-Compatible Contract ....................... 22
Figure 1-3: Foreign Bank Entrance ................................................................................... 30
Figure 3-1: Annual and Sample Variable Means ......................................................... 131

Figure 3-2 Effect of Depreciation on the Probability of Banking Crises at Different
Levels of Currency Mismatch in the Banking Sector ............................... 150

Figure 3-3 Effect of Depreciation on the Probability of Banking Crises at Different
Levels of Currency Mismatch in the Real Sector ..................................... 151

Figure 3-4 Effect of Real Depreciation on the Probability of Banking Crises at
Different Levels of Total Currency Mismatch in the Banking Sector ...... 152

xvii

PART ONE. FINANCIAL LIBERALIZATION AND THE REPRESENTATION

CHOICE OF A FOREIGN BANK

1. Introduction

Liberalization of ﬁnancial services throughout the world fostered a rapid
expansion in international banking. With proﬁt opportunities rising across borders many
banks increased their investments in other countries. This increase took a variety of
forms --acquisition of existing banks in the host country, opening a new branch or
subsidiary, or simply channeling funds through the host banking system and extending
direct loans to residents in the host country. With so many patterns of capital mobility, it
is important to understand what motivates foreign banks’ choice of representation in a
host economy and the economic impact of this choice.

This paper is focused on the foreign banks’ choice between correspondent
banking1 and opening a subsidiary in a host economy. What encourages foreign banks to
channel investments through the local banking system or with subsidiaries is a question
that is motivated by an interesting economic phenomenon. There is a clear difference in
the pattern of banks’ investments in developed and developing countries. Correspondent
banking is the main channel for foreign banks’ investment in developed countries, while
developing economies see a larger capital inﬂow from local subsidiaries of foreign banks
than from correspondent banking. However a survey by Barth et al. (2001) indicates that

developed economies have very low barriers to entry in the banking system compared to

 

’ For the purposes of this paper, correspondent banking relates to agency arrangements in which a bank
provides services to another bank for a fee. These services include extending investment credit, credit lines
and trade credit, and gathering of information. Also, correspondent banking relates to loans extended by a
foreign bank to a domestic bank, where the domestic bank uses the loans to further grant credits to
borrowers.

developing economiesz. Then why do we see multinational banks entering developing
economies’ banking systems and being shy to do the same in developed economies?

There is some anecdotal evidence about the behavior of multinational banks that
show that host countries’ institutional characteristics and information (dis)advantages
may motivate these banks to choose the form of representation in the host country. The
International Finance Corporation (IFC) extends loans to small and medium enterprises
in host countries primarily through the domestic banking system. The IFC uses
intermediaries to extend long-term loans to these enterprises because direct lending
carries high transaction costs. Clearly, delegated lending reduces the costs of loan
administration in this case and this channel of investment seems most natural, given that
domestic banks have easier access to information about local producers than foreign
investors. However, HSBC claims that the reason this British bank maintains a wide
international presence through subsidiaries is to take advantage of local knowledge about
the host economy. Thus, some foreign investors prefer to internalize information costs
since those costs may be lower than the costs of delegation.

The choice between internalization and delegation has been explored in detail for
the case of a manufacturing company3, especially in relation to the behavior of
multinational corporations (Horstmann and Markusen, 1987, 1996). A multinational

company may consider it costly to initiate production in a host country because of

 

2 For example, Barth et al. (2001) show that the rate of denial of foreign bank applications is 2 percent in
developed economies, compared to 29 percent in developing and emerging markets. Only 8 percent of
developed countries impose restrictions on foreign ownership of domestic banks and 4 percent impose
restrictions on foreign entrance. In developing and emerging economics, the respective numbers are 31 and
17 percent.

3Severinov (1999) tackles the question of the organizational form a producer might choose in outsourcing
production of inputs. The producer may contract directly with supplier(s) or may delegate some
contracting responsibilities to one of the suppliers. Mookherjee and Tsumagari (2001) study a similar
problem assuming that the producer may delegate contracting to an informed intermediary rather than a
supplier.

stringent entry requirements or because of informational disadvantage about the market
compared to an incumbent company. On the other hand, the multinational company may
delegate representation to a local contractor and exploit his exclusive information about
the market. However, delegation creates the risk that the agent may choose to serve his
own interests ﬁrst. A cost-beneﬁt analysis of the two options may prove that opening a
subsidiary in the host country yields higher proﬁt for the foreign company than
contracting with a local agent; thus, the foreign company has an incentive to enter a host
market with ownership.

Similarly to a multinational corporation, a foreign investor (bank) faces the choice
to lend to the host country using correspondent relationships with a domestic bank or
opening a subsidiary bank. However, the theory of multinational banking faces a large
gap in explaining this choice. Since a major proportion of international capital
movements are channeled through the banking system, foreign banks’ investment choices
may have signiﬁcant impact on capital accumulation, output, and welfare in the host
country.

The choice of the foreign bank is likely to have larger effects in countries with
underdeveloped ﬁnancial systems than in developed countries. For example, in transition
economies where the scope of the ﬁnancial system is limited to banking, and domestic
savings provide an insufﬁcient supply of investment funding, multinational banks offer a
major source of ﬁrm ﬁnance. However, because of ﬁxed costs and layered asymmetric
information, loans from the subsidiary of a foreign bank, and loans administered by a
correspondent domestic bank each potentially have different implications for cross border

capital movements.

This paper addresses the following questions. First, how do foreign investors
choose the organizational form of representation in a host country, speciﬁcally between
subsidiary and correspondent banking? Second, what is the effect of this choice on
investment and output in the domestic country? Third, what welfare gains are generated
by the host economy ﬁ'om foreign bank entrance and correspondent banking?

The paper’s approach is related to the internalization theory of the existence of
multinational corporations, which states that a foreign company would “prefer outright
ownership of complementary assets, rather than bearing the costs of contracting in the
open market“. The model presented in this paper borrows a framework from Horstmann
and Markusen (1996) of a multinational corporation’s choice between an agency contract
and a subsidiary, and employs it in a two-period general equilibrium model where a
foreign investor chooses between an agency contract with a domestic bank and opening a
subsidiary bank. The investor makes this choice so as to minimize his agency costs and
costs of entering the host banking system. An agency cost is created by a two-tier
asymmetric information problem. This paper applies a form of Townsend’s (1979) costly
state veriﬁcation in a model of investment ﬁnance and, similarly to Diamond (1984,
1991), uses the costly state veriﬁcation concept to study the incentives for a ﬁnancial
intermediary in delegated monitoring. Costs of entering the domestic banking system are
comprised of a ﬁxed set up cost and losses associated with competition.

The paper ﬁnds that the foreign investor will choose to enter the domestic
banking market when agency costs are relatively higher than competition costs, set up

costs are low, and the host economy has a small capital endowment. The paper also

 

‘ Williams (1997), p. 73

concludes that the domestic economy beneﬁts from foreign bank entrance if the foreign
investor does not exploit a monitoring advantage that rapidly depletes domestic banks’
proﬁts and if the domestic economy enjoys high capital productivity.

The paper is organized as follows. Section 2 describes a general equilibrium
model with three types of domestic agents (depositors, project owners, and a bank owner)
and a foreign investor. Section 3 derives the conditions under which the foreign investor
makes his capital allocation choice. Section 4 analyzes the implications of the model for
the domestic economy’s welfare. Section 5 presents a summary and conclusions of the

paper’s ﬁndings.

2. The Model

This section develops a two-period general equilibrium model incorporating
asymmetric information in the bank-borrower and the bank-foreign investor relationships.
The asymmetric information problem introduced in the model’s borrower-lender
relationship leads to costly state veriﬁcation, as described by Townsend (1979).
Townsend shows that if a principal in a bilateral exchange contract does not obtain free
knowledge about the agent’s output, the optimal contract must be incentive compatible.
He shows that the set of transfers between principal and agent, and the veriﬁcation region
of output must create sufﬁcient incentives for the agent to reveal the actual output.
Townsend also proves that stochastic veriﬁcation dominates deterministic audit. The
model presented in this paper assumes that the domestic bank must put forth costly effort
to receive a signal about the borrower’s true output. Since effort iS assumed to be
proportional to the probability of receiving a positive signal, this type of veriﬁcation

procedure is very similar to stochastic auditing.

The model combines the asymmetric information effects outlined in Townsend
(1979) and Diamond (1984, 1991) to argue why correspondent banking is costly to a
foreign investor and uses the framework of Horstmann and Markusen (1996) to present
the optimal agency contract. It then goes on to describe the capital allocation choice of

the foreign investor.

2.1. Domestic Agents and a Foreign Investor

There are three types of agents in the domestic economy: n depositors (savers), m
project owners (ﬁrms), and a single bank owner (banker). Depositors, project owners,
and the banker consume only in the second period. Project owners and the banker are
risk-neutral and have no initial capital endowments. Firms’ investment is ﬁnanced
through loans from the bank. The bank raises capital from foreign investors and by
selling deposit contracts. Domestically, only depositors have positive initial endowment,
(0°. The second period consumption of a depositor brings utility given by

U(cd), U'(cd) > 0, U"(cd) < 0, U'(0) =oo, U700) = 0.

The depositors, as well as all other agents, have access to a safe storage

technology, which transforms with certainty a unit of period 1 good into rs units of period

2 good. They can also deposit their savings with the bank, which pays an interest rate rd.

Depositors maximize expected utility subject to the budget constraints
d
c = rdd+ rss

0
d+s=m

where d denotes deposits and s denotes savings in storage.
Firms are perfectly competitive. They borrow from the bank to ﬁnance
investment and produce a homogeneous ﬁnal good using a decreasing—returns-to-scale

technology:

y=€F(ki), F(0)=0, F'(ki)>0, F'(ki)<0. F'(0)=°°. F'(°°)=0

where ki is the amount of capital utilized in the production process by ﬁrm i,

while a is a productivity coefﬁcient introducing uncertainty in the outcome of production.
With probability 0, the project yields F(ki) and thus 3 equals one. With probability (1-0),

the project brings no output, s=0. The realization of e is independent across projects and
the risk of the projects is either high or low as explained in detail below. The expected

proﬁt derived from a single project, which also equals second period consumption, is

E(Ilif) = 0F(ki) - Oriki
ri is the price of capital paid by ﬁrm i and it repays a loan only in case of success.

The assumption of “limited liability” is necessary so that the project owner has
nonnegative consumption.

Actual output is exclusive information for the borrower. So, while ﬁrm i

costlessly observes the realization of a, the lender would have to put some effort (e;) to

reveal this information. However, if the lender discovers that the ﬁrm is dishonest, he
chooses the most severe punishment by seizing the whole output of the borrower.

The auditing procedure of the lender has low efficiency in the sense that as the
lender devotes time and resources to monitor the ﬁrm, his effort only results in a signal

about the actual output. If the realized output is 0, then the signal is 0 with probability

one. If the investment project is successful, then the signal is 1 with probability p(e,) and
0 with probability I-p(e,-). Therefore, if the project owner misreports the output, the

lender will be able to collect only p(eL-) 6F(k,) on average. The efﬁciency of the output

verifying procedure depends positively on effort and further it is assumed that effort takes

values between 0 and l and p(e,)=e,-5. The cost of monitoring is a convex function of

effort: C(e), C'(e) Z 0 and C"(e) Z 0.

Generally, the bank owner is an agent who, like the project owners, has no initial
capital endowment and whose delegated role is to monitor the projects. While
depositors, or any other lender, could monitor the projects themselves, the total cost of
auditing all projects by all depositors would exceed the total output derived from

domestic savings. Also, if all depositors verify the yield of every project, the total

m
auditing cost is n ZC(ef1) and is larger than the auditing cost for the bank owner,
i=1

m
ZC(e,I’), given ed]- = ebj. Thus, depositors ﬁnd it beneﬁcial to delegate the output
i=1

veriﬁcation to the bank and to take advantage of the scale economies in project
monitoring.

Certainly, the bank owner has an incentive to misreport the project returns to his
depositors. However, Since the number of projects is sufﬁciently large, the bank is

perfectly diversiﬁed: it receives a certain positive return on the projects and, as in

 

5 Another way to think about the auditing effort is to assume that the bank owner is endowed with a unit
of labor and that the veriﬁcation technology is linear in the labor factor.

Diamond (1984, 1991), the moral hazard cost of delegation is reduced to zero".

A foreign investor (bank) considers extending credit to projects in the domestic
economy, for which he could use a correspondent relationship with the domestic bank, he
could initiate a subsidiary in the domestic country, or extend direct loans to the project
owners. However, the projects may have low (state 0': l) or high (state 0= 2) expected

return, and the foreign investor does not have enough information to discern the state of

the average project. Ex-ante, he knows that state a = 1 occurs with probability ,0], and
state 0= 2 with probability p2, p2=1-p1. The foreign investor is aware that the average

productivity of the projects is higher when o=2, 62 > 61, and he also knows those

productivities.

As in McKinnon and Pill (1998), the model assumes that the bank has an
informational advantage over depositors and the foreign investor, and only the banker
knows the average quality of the projects at t=0. Thus, if the foreign investor chooses to
delegate monitoring to the domestic bank, the bank owner will have an incentive to
misreport the return on investment. With asymmetric information, the foreign investor
will have to offer a contract that at least compensates the domestic bank for the project
veriﬁcation costs, if the bank is to report honestly.

If the foreign investor chooses to open a subsidiary bank in the host country, he
pays a ﬁxed cost, G, for licensing and coverage of other expenses pertaining to domestic
banking regulation. However, by doing so, he learns the quality of the projects and can

compete with the domestic bank in extending credit and monitoring borrowers. The

 

6 Krasa and Villamil (1991) show that delegation dominates direct investment even when the number of
projects is limited but sufﬁciently large.

. . . 1 Fl
Investor’s cost of monItorIng, CF (e ), may or may not be the same as the cost of the

domestic bank.7

2.2. Timing of the Events and Equilibrium

At an initial time r=0, the foreign investor calculates the expected rates of return
from correspondent banking and from the subsidiary, and makes an investment decision.
For the foreign bank to consider any investment option in the domestic country, the
investment should yield an expected gross return at least as high as the world rate of
return r*. Aﬁer the decision is made, the state of the economy is revealed to the domestic
bank and the foreign subsidiary, if the subsidiary exists.

If the foreign investor expects a higher rate of return from delegating to the
domestic bank, at time t=1, he offers a contract to the correspondent specifying the
amount of the investment and a rate of return for the foreign investor in each state of the
portfolio return (which also determines the monitoring effort of the bank).

The domestic bank raises capital from the foreign investor and from selling
deposits, and allocates the capital to the investment projects. The bank offers loan
contracts to the project owners that detail the amount of a loan, interest rate, and
veriﬁcation procedure. These loan contracts must be incentive compatible to induce the
borrowers to report zero output only when there is indeed no output.

At time i=2, each ﬁrm reports to the bank and repays the loan if it has revenue.

Otherwise, the bank initiates bankruptcy of the ﬁrm and puts effort to verify if the ﬁrm

 

7 Direct lending and subsidiary banking are discussed simultaneously in this paper. Direct lending can be
interpreted as a case where the foreign investor pays a cost G to ﬁnd out the average quality of the projects
and most likely has a monitoring disadvantage over the domestic bank and a foreign subsidiary.

10

has reported truthfully. The bank then pays gross interest to the depositors and the foreign
investor.

If the foreign investor expects to have higher proﬁt from banking with a
subsidiary in the host country, at time i=1 he pays the ﬁxed cost, learns the quality of the
projects, and enters into competition with the domestic bank for borrowers. The two
banks then extend loans to the project owners.

A competitive rational expectations equilibrium for this economy is characterized

by: a set of consumption allocations for the depositors, project owners and the bank

owner %7, cg, c1f , cg, cb }; a set of capital allocations for the project owners
{k,, km} and the bank owner {K}; set of savings allocations for the depositors
{d,, d”; 3,, Sn}; interest rate on deposits, on loans extended by the domestic

bank and the foreign subsidiary, and interest rate paid to the foreign investor

{(d, rb, r”, RI; and veriﬁcation effort of the bank owner and the foreign subsidiary

{eb, en }, such that depositors, project owners, the bank owner and the foreign investor

maximize expected utility or proﬁts, and goods, deposit, and loan markets clear.

2.3. Loan Contracts with Correspondent Banking

A main assumption in the described model is the monopoly position of the

domestic banks. If the bank is competitive, it will have no incentive to proﬁt from its

 

8 This assumption is not unrealistic since most emerging markets have very concentrated banking
systems. Tests by Mathieson and Schinasi (2001)) show that Latin American countries tend to have
monopolistic banks, while the banking structures in Poland, Czech Republic and Turkey are either
monopolistic or monopolies.

ll

information advantage over the foreign investor, because any proﬁt will be eroded in the
loan contract.

To present the optimal loan contract offer of the domestic bank, it is useful ﬁrst to
focus on the bank’s ﬁnancing. In this case, the bank owner is a monopolist in the deposit

market. As long as he offers a certain deposit rate that is at least as high as the storage

rate rs, savers will inelastically provide their whole initial endowment to the bank. Thus,
. . . . . . Kd o
the total ﬁnancmg the bank 18 able to rarse from domestic deposrtors IS = nd= M) =
.00, where .00 is the economy-wide endowment of period 1 good. Depositors will agree
. . . . d .
to put their savmgs In the bank only If r as. The lowest cost for the bank 1S to pay

depositors exactly rs per unit deposited.

Since the focus of this paper is on the choice of the foreign investor to enter the
banking market of the host country, it is useful to simplify the behavior of the domestic
bank. It is assumed that in autarky the bank will use the storage technology and will not
grant Ioansg, since the marginal cost of the veriﬁcation effort at zero exceeds its marginal

d

beneﬁt even when the projects have a high expected return, 62F[K—Js(l- 62)c'(0+).
m

Thus, because of the small size of the projects, the monitoring cost is prohibitive. As the
bank does not lend, the bank proﬁt will be zero in autarky.
Foreign investment increases the size of the domestic projects and makes lending

and output veriﬁcation proﬁtable for the bank, even when the average quality of the

projects is in state 1. Since domestic depositors cannot be paid less than rs, the foreign

 

9 For example, the bank may invest in a risk-free government bond.

12

investor will bear the veriﬁcation cost that arises from the inﬂow of capital.
Since the bank owner does not have free information about the outcome of the
projects, he offers incentive compatible loan contracts that maximize bank proﬁt. Each

contract speciﬁes the amount of capital used in a project and the interest rate on a loan.

The bank owner takes as given the rate of return R, and the effort ei, that he agrees upon

with the foreign investor, as well as the total domestic deposits and the minimum interest

rate on deposits. The maximization problem of the bank also determines the demand for

foreign capital, KF 1.
M F! d '"
Max 26th - RK — rsK - 711-6908;)
KFI,{k1,...,km},{r1,...,rm} i= i=
S-t Hkai) ' 90kt Z (1 'ei)9F(ki) (1C)
61%)- 65-192 0 (IRF)

Egg-k, — 12K” — rSKd — §(1-9)C(e,-)20 (IRB)

,= ,=
Ek, =K F’ + K d
i:

The ﬁrst constraint is incentive compatibility. It requires the proﬁt from each
project to be higher if the project owner repays the loan, rather than if he claims that the
output is zero when it is actually positive. If the project owner misreports, the bank will
start bankruptcy proceedings and will try to verify the outcome of the project. In this
case, the bank will seize the whole output if it appears that the borrower was dishonest.

However, since the signal received by the bank is not perfect, the bank will collect the

13

output only with probability eilo. This veriﬁcation procedure is almost identical to

stochastic auditing -- where the banker receives a perfect signal, but chooses to verify the
output randomly instead of every time the ﬁrm misreportsl 1.

Because of the imperfect signal, the ﬁrm earns proﬁt and the second constraint
can be dropped. The fact that a perfectly competitive ﬁrm earns proﬁt is only a result of
the monopoly position of each project owner in a market for information. With costly
state veriﬁcation, the loan contract generates gains because a is independent across
projects and its realization is not public knowledge. The proﬁt, (I-e) 6Fﬂr), is the rent to
the project owner’s exclusive information about his own revenue.

The third constraint guarantees that the bank owner will have nonnegative
consumption. The last term in the bank’s proﬁt function is the monitoring cost covered

by the foreign investor. As all projects are identical ex ante, the bank tailors the same

contract for each ﬁrm. In symmetric equilibrium (ki=k and ri=r, i =1 ...m), the ﬁrst order

conditions of the bank maximization problem yield three equations solving for k, r and

K”:

e,6F’(k)=R (I)
e,6Fﬂc)=0rk (2)
K” =mk-Kd (3)

Market clearing requires the demand for foreign capital KFI equal the quantity

 

'0 Monitoring with imperfect signal is introduced in Diamond (1991b). A noisy signal about the quality
of the projects that depends on bank effort (screening intensity) is also used by Gehrig (1989) in a model
with adverse selection, and by Boadway et. al. (1999) in an adverse selection model where social workers
need to determine the type of the agents applying for social beneﬁts.

” The main difference is that with stochastic veriﬁcation the cost always depends on the probability of
veriﬁcation even if C(e ) is assumed to be a constant. Also, the assumption of stochastic veriﬁcation is not
very realistic. The bank may be required to initiate an audit of every ﬁrm that is in default on a loan.
However, the quality of the audit, which depends on effort, can hardly be regulated.

l4

provided by the foreign investor, K0. Thus, every project receives k = m-](K0 + Kd) =

-l . . . . . .
m K, where K IS the total Investment In the economy. The market cleanng condition and

the equal distribution of capital among all projects imply that the effort the bank must put

forth for a loan contract to be incentive compatible is an increasing function of the

interest rate paid to the foreign investor, e; = [6F ’(k)]-IR. Also, the bank devotes the

same effort in monitoring every project, e,- = 8.

Since the investor entrusts monitoring to the bank, the effort requirement is a
clause in the offer of the foreign investor to the domestic bank and, as shown below,
solves the foreign investor’s proﬁt maximization problem.

The maximum proﬁt for the bank is given by:

me9F(k) — RK" - m(1-I9)C(e) - rst
me9F(k) — e6F’(k)K0 - m(1-9)C(e) - rst.

Hb

2.4. Agency Contract with the Correspondent Domestic Bank

The foreign investor suffers an informational disadvantage compared to all
domestic agents because he does not know the state of the domestic bank’s portfolio
return. To see how this imposes a moral hazard cost on the investor, initially suppose

that he knows the average return. Then, in a ﬁrst best contract delegating loan

management to the bank, he proposes investment K0, requires an interest rate R, and

implicitly demands monitoring effort e, so as to maximize proﬁt -- given that this proﬁt
outweighs the opportunity cost of funds, r. per unit of capital. Also, the domestic bank

must be better off allocating foreign funds than staying in autarky.

15

Max RKO — r’K"
R, K0

0 d
s.t. mec91{-K—+—K——]- RKO — (l — 6)mC(e) — rst 2 0
m

R

e: o d
6F{K +K ]
m

 

This problem is equivalent to

o d
M... .6F{£_~:£<_}.o -. K0

K0,e m
o d o d
s1. me0F[-1(—+I(—] — eI9F[-I-<—:K—]K° -(1- 0)mC(e) - rst 2 o
m m

The ﬁrst order conditions are:

 

 

w.r.t. K" {3917“ a / ”0K 0 ](1 — ,1)— r‘ + eI9F'(K° /m) = o (4)
m
w.r1. e : 9F’(K° /m)K°(l — A) + m[6F(K° /m) — (1 — 6)C'(e)]& = 0 (5)
w.r.t. xi. :meBFLgiK—d] — e6F'[ K0 + Kd )Ko - (l — 0)mC(e) - r’Kd = 0 (6)
m m

The ﬁrst order conditions of this problem Show that the bank earns no proﬁt, the
marginal cost of the bank veriﬁcation effort equals the expected project output, and the
marginal beneﬁt of the foreign investor’s capital equals the world cost of capital.

Proposition 1

Under perfect information, the optimal contract speciﬁes higher capital (KO),

interest rate (R), and effort (e) in state o=2 compared to state o=l.

Proof. See Appendix 1

l6

Equations (4) and (5) imply that as projects become safer on average (and Orises)
the optimal veriﬁcation effort must be higher. This result occurs because veriﬁcation
becomes cheaper on the margin, while the expected product of the investment increases.
Also, the foreign investor enjoys an increase in the marginal beneﬁt of effort as the
productivity variable enhances the effect of effort on the interest rate R. As effort
increases, the gross return for the foreign investor should be higher, as well. In addition,
when facing a perfectly elastic supply of capital, the foreign investor has an incentive to
invest more capital when the quality of the domestic bank’s portfolio is good.

Now consider the proﬁt of the domestic bank in each state 0 of its portfolio
b . . .
return, 170 . Also consrder the proﬁt when o=2, the forergn Investor does not know the

state and the banker claims that the average project has poor quality. This is the banker’s

proﬁt from cheating, Hzcb.

0 d 0 d
11?, = meaﬂaFL—K" + K ]~e0t90 F{———Ka + K JK; — (l — 60)mC(ea)- rst = 0
m

m

o d o d
Hg, _ W, {11115;}..3191 F£m£_]kf_ (,_62,,,C(,1 1]-,st
m m 62

(l - 01)MC(€1) - (l - 62)m({e1 312-] > 0

Since the veriﬁcation cost decreases, as the projects become safer, the bank can

earn proﬁt if it receives capital K10, and puts effort e161/62 when its portfolio has high

expected return. Figure 1-1 illustrates this idea.

17

Figure 1-1: Correspondent Banking with Asymmetric Information

K° l—IZFI

 

  

 

b
“R
K20
0 LEGEND
K1 . Correspondent banking
with perfect information
‘ Cheating
C

 

 

 

 

6191/92 e1 e2

b b . . b
171 and [72 represent the domestic bank’s zero—proﬁt curves In each state. 172C

shows the positive proﬁt of the domestic banker when he wrongfully claims that the

projects have low quality. Notice that in this case the domestic bank chooses contract
(K 10, R1) but puts lower monitoring effort than e1,

With the potential for moral hazard, the foreign investor constructs proﬁt-
maximizing contracts that will reveal the domestic bank’s portfolio state through self-
selection on the side of the bank. A contract speciﬁes the capital received by the

domestic bank, and interest rate paid to the foreign investor (implicitly specifying the

effort since e = [9F ’(k)]-IR) in each state. The foreign investor solves the problem

18

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Max
Kf,K1°,R1,R2
8.!
I {Kurd} I I I
m
m
R, 0 d —K0 —(l—01)mC R1 d -rSKd
F{K1+K ] 01F'[K—l————- +K J
K m i K m 1
N
( {199+de \ f
m
m
2R2 0 d —K" —(1—6,)mC 1:2 d —-rSKd (ICBI)
F,[K2+K ] 61F,[K2+K J
K m J x m J
\
m
m
R2 0 d -—K" —(1—62)mC R2 d -r"Kd
F,[K2+K ] 62F,[K_2___+K ]
K m i K m i
/ {K10+Kd] \ K W
m
m
ZR] 0 d —K" ( 62)mC 1:: d —rSKd (1032)
F,[K1+K ] 62F[K1+K J
K m J K m )
( K1 +Kd \ f \
K0 R1 3 d
R, K—(l—61)mC d -rK 20 (IRBI)
:[Kf’ +Kd] 61F,[Kf’+K J
I i K m )
(mF[K2+Kd] \ K \
m
R2 d -K" —(l-—02)mC R2 d 4st 2.0 (1RB2)
F,[K§+K ] 02F,[K§+K J
t m J K m )
p1(R1K10 — r’Kf’)+ p2(R2K§’ - r’K§)2 Max{0,II]FI } (IRFI)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l9

 

p,(R,Kf’ —r’1<f’)+ pZIRng —r’K§)

Fl . . . .
where 171 In the IRF I constraint IS the Investor’s proﬁt ﬁom the ﬁrst best contract

when the average project has poor quality. The IRFI constraint implies that if 171F1> 0,

there is a possibility for either pooling or separating equilibria, depending on the value of
p1. The separating equilibrium is discussed ﬁrst.

As shown by the incentive compatibility constraints, the state contingent contracts

are constructed so that the banker is better off reporting a good quality portfolio rather
than receiving K10, paying interest rate R], and putting effort e101/62. Similarly, the

banker has no incentive to misreport a low quality portfolio. The foreign investor’s
problem is essentially costly state veriﬁcation. The cost for the foreign investor
(compared to the perfect information contract) is translated into overinvestment and
underpricing of capital ﬁnancing high quality projects and underinvestment in low
quality projects.

The foreign investor also faces an interim moral hazard problem since a speciﬁed

amount of capital K0 and interest rate R implicitly determine the bank owner’s effort.

Thus, by choosing an incentive compatible agency contract, the banker reveals his loan
portfolio, as well as his own future actions.

In the ﬁrst best contract, the banker earns zero proﬁt. Even with asymmetric
information, there is no reason for the foreign investor to allow the bank to proﬁt in the
bad state, since this will only reduce his proﬁt. Therefore, constraint IRBl is binding.
However, if the foreign investor does not know the state, the banker will misreport the
high quality portfolio and earn a positive payoff. Therefore, in the incentive compatible

contract, the bank should be able to gain some information rents, or otherwise, the banker

20

will cheat. Thus, the bank individual rationality constraint in state 0= 2 is not binding.
The foreign investor will raise the bank’s proﬁt only enough to make the banker
indifferent between truthfully announcing state 2 and lying, and thus ICBZ is binding.

Also, it can be shown that ICBl is not necessary since the bank never lies about state 1.

. . . . . K 0 + K d
WIth two binding constrarnts and srnce R = eOF —— , the problem of the
m

. . . . . 0 0
foreign Investor can be reduced to maxrmrzatron over X] , K 2 , e 1, and 62.

o d o d
M... pIIelolr'Ifl—ijxr—r‘Kr]+p2[e262F'[5ﬂ—]K§We]
K10, K10 , e1, e2 m m

SJ.

0 d o d
m6292F[-1—(2—;—K—] — e202F'[-K—2%£—JK§ - (1 — 02 )mC(ez ) - rst

o d 0 d
= me19]F[§—l—;i—] — elaF'[£;—K-]Kf’ — (1 — 02 )mc[e2 3:] - rst (ICBZ)

o d o d
me161F[§l—+—I—(——] - elaF'[5‘—+—I<—]Kf — (1 — 6])mC(e1)- rst = 0 (IRBI)
m m
o d * 0 d *
p,[..a,p[£1_:£_]a _. Kg].p,[.,g,p[er_ls_]1.g -. Kg]
m m
2 MaxIOIIlFI } (IRFI)

Proposition 2

. . . . . . IC .
In state 0 =1, the capital specrﬁed In the mcentrve compatible contract, K1o , Is
. . . . . . . oIC .
smaller than In the contract wrth perfect Information, while 111 state 6 =2, capital Kg 18

. . IC IC . . . .
higher. The effort of the domestic bank, el and e2 , IS always smaller In the Incentive

21

compatible contract than in the perfect information contract.
Proof. See Appendix 2

Proposition 2 is illustrated in Figure l-2.

Figure 1-2: Correspondent Banking with Incentive-Compatible Contract

0

K “2“

nzblC

 

LEGEND
O Correspondent banking
with perfect information

 

 

I Correspondent banking
with incentive-
compatible contract

 

 

 

 

 

 

 

Obviously, in state 2 the foreign investor would have to reward the domestic bank
for revealing the quality of its projects by investing more capital at a reduced interest rate.
The downward pressure on interest rates will give the bank incentive to lower its effort
and save on monitoring costs. Thus, when the quality of the projects is good, there is
excessive capital inﬂow compared to correspondent banking with perfect information.
The quality of the monitoring deteriorates in this situation, and it becomes more likely

that the project ﬁnancing will not be repaid.

22

Since the incentive compatible contract only makes the bank indifferent between
truthfully revealing the state and cheating, the foreign investor will also try to make
cheating unattractive. Therefore, he will try to bring the new isoproﬁt curve of the
domestic bank as close as possible to the zero proﬁt. The foreign investor could only
achieve this if he lowers the amount of capital invested in low quality projects. Since the
decrease in capital will lower the marginal beneﬁt of effort, the bank will not be as
stringent in monitoring the projects.

Compared to a correspondent banking contract with perfect information,
asymmetric information will cause overﬁnancing of good quality projects and
underﬁnancing of low quality ones. However, the domestic bank will reduce its
monitoring effort in each state.

The expected proﬁt of the foreign investor from correspondent banking with
separating equilibrium is:

01C

K + Kd . K0
[(___ , Irma- 4.]
01C d
+ mp2[65C02F[£;—+—1£—]-(1 — 02 )C(e5c)
m

]_r‘.’fi]_rst (7)

 

m

-(1-01)C(eic)+(1-92)C[e{c61
92

If p1 is in the neighborhood of 1, then there is a pooling equilibrium. If e1 is the

value of effort and K10 is foreign investor’s capital that maximize proﬁts with symmetric

information between the foreign investor and the domestic bank, then:

23

d

0 d I. 0 K
HFIP=II1Fl=m elglp[.1£1_t§_]-(1_gl)c(el)_r EL—rs— >HFIIC
m m m

[P . . . . . . . .
where 17F 1S foreign Investor’s proﬁt In pooling equilibrium.

Thus, if state a = l is very likely, the foreign investor may just as well offer the

ﬁrst best contract and let the bank always report a low quality loan portfolio”.

2.5. Entrance of the Foreign Investor in the Domestic Banking Market

If the foreign investor chooses to open a subsidiary in the host country and invest
directly, he pays the ﬁxed cost G and discovers the projects’ expected return. It is
assumed that the foreign subsidiary does not enter the deposit market and will be

ﬁnanced only with the capital available to the foreign investor from the world capital
market, K0”.

The foreign investor’s cost ﬁmction is aC(e), a>0. a<1 implies a cost advantage
in monitoring. For example, the investor may use foreign staff, which is more
experienced and carries long-term knowledge acquired at the investor’s subsidiaries
abroad. If a>1, the domestic bank is relatively more efﬁcient in monitoring the projects
than the foreign investor. This paper considers the case when the foreign investor has an

efﬁcient monitoring technology but faces very high costs of acquiring information about

 

'2 It is assumed that 171Fl>0. Otherwise, the market will close whenever the domestic bank announces a
= 1 and there is no asymmetric information problem.

'3 The model will not change signiﬁcantly if the foreign investor draws resources from domestic savers.
The amount of capital invested in the domestic economy, and the monitoring effort will be the same. The
number of ﬁrms served by the foreign investor becomes ambiguous and we avoid the ambiguity by
imposing this simpliﬁcation.

24

the average quality of the projects. In this case, he will prefer to open a subsidiary and
collect information in the host country. If he can acquire enough information about the
quality of the projects at a low cost, he will prefer to grant direct loans to project owners
in the country. These two cases are analyzed simultaneously, as the only difference is in
the amount of the entrance cost G.

Since both foreign and domestic capital ﬁnance the ﬁrms, the projects will be
large enough to make monitoring worthwhile, even for the domestic bank which, as in
autarky, is funded only from deposits.

The foreign investor and the domestic bank grant loans to different ﬁrms. Since a
ﬁrm will choose a debt contract that maximizes its proﬁt, in equilibrium, the domestic
and the foreign bank will offer contracts that yield equal proﬁt for the ﬁrm. Firms’ proﬁt
equalization across the debt contracts determines the number of borrowers that will be

each bank’s customers.

. . . . . F1 F1
Suppose the equIIIbrIum number of customers of the foreign bank 13 m , m

I/\

. . . F1 F]
m. Then the foreign banker wrll choose an Interest rate r on a loan, amount k per
. . oFI . . F1
borrower (or the total amount of Investment 18 K ), and veriﬁcation effort e , so as to

. . . FI .
maxrmrze proﬁt given the number of customers, m , the world cost of capital r*, and the

ﬁxed cost G; and subject to the usual minimum proﬁt constraints and the incentive

compatibility constraint for the borrower.

25

F1 Filiaxo F1 mnélrnkn—mF1(1—6)0C(eF1)-r*KO-G
r ,k ,K,e

s.t 9F(kFI ) sink” 2 (1 - eF] )9FIRF’ ) (1C — F1)
6F(kFI)6rF’kFI 20 (IRF-FI)
mFI kFI 2K0

As before, the incentive compatibility constraint is binding and the expected

interest rate equals the expected average product for which the veriﬁcation is successful,

rFI— — e F (kFI)/kFl The foreign investor’s capital lS evenly distributed among all of his

customers, k” K0 H mm ,and the marginal product of capital equals the world price of
capital
eF’aF'IkF’ )=r’ (8)
The marginal beneﬁt of the auditing effort equals its marginal cost

BFIkFI )= (1 - 6)aC'(eF’ I (9)

A similar ﬁrst order condition is obtained for the domestic bank

61F(kb)=(l- I9)C'(eb) (10)

where kb=Kd/mb, mb is the number of domestic bank customers (mb+mFI=m),
and eb is the effort this bank puts into monitoring.

A ﬁrm derives proﬁt 17f F] = (l-eFl) 0F(kFl) from a contract with the foreign

subsidiary and 17b = (1 -eb) 0F(kb) from a contract with the domestic bank. The ﬁrm

will choose a contract with the foreign subsidiary only if its payoff is higher than the

payoff from a contract with the domestic bank. As long as the proﬁt of the ﬁrm increases

26

in the amount of capital (and decreases in the number of projects ﬁnanced by the same

bank)”, there will be equilibrium in the loan market. The ﬁrm will gain the same payoff

from either contract only if (1 -eFI) 0F(kFI)= (1 -eb) 6F(kb) or if

(I - e“ LC'IeF’ )= (I - e” )C'Ie”) (1 1)

When a=1, eFI=eb, and (K0F1)/mFI = [(1 /mb = (K0 F1 +Kd)/m. This implies that

the number of projects ﬁnanced by the foreign investor is m” =m K0 FI)/(K0 F1 +Kd).

Using (8)-(1 1), it can be shown that as the monitoring efﬁciency of the foreign investor

. F] b F] b F1 I .
decreases (a Increases), k , k , e , e , m and KP become smaller, provrded that the

second order condition for foreign investor’s proﬁt maximization is satisﬁed:

— (l - t9)aC"(eF’ )eF’aF"(kF’ )— [9F'(kFl )Iz > o.
In addition, when a>1, the domestic bank puts more effort into monitoring than

. . b . F1 b
the foreign Investor, en<e and ﬁnances larger pr0jects, k < k .

As monitoring becomes costlier on the margin, the foreign investor will reduce
both his effort and the total capital inﬂows. Since his capital share becomes smaller, in
equilibrium, fewer ﬁrms would receive ﬁnancing from the foreign subsidiary. If the
investor has a cost advantage, he will gain larger market share while putting more effort
and investing more capital.

At the initial time t = 0, the foreign investor’s expected proﬁt from direct

investment becomes

 

” The condition is that [-C’(e) + (I-e)C"(e)] > o.

27

KOFI Kd KOFI Kd
HF’D’ =mf’p1 ef’eiF ————‘ + -(1-61)0CI81F')- r" —‘ +
m m
KOFI Kd KoFI Kd
+mf’p2 eZF’aZF —2—+—— —(1—62)ac(e§’)—r* —2—1— —G (12)
m m

Compared to the ﬁrst best with correspondent banking, this proﬁt differs in the
ﬁxed cost, and the efﬁciency coefﬁcient. As a increases, the investor’s proﬁt becomes
smaller. This proﬁt is additionally reduced because the foreign investor serves only a

section of the market. However, compared to the proﬁt from the agency contract, 11””,

HF”), is not burdened by a delegation cost.

2.6. Foreign Entrance and Capital Flows

When the foreign investor chooses to open a subsidiary, he eliminates three types
of costs that he would incur in correspondent banking: the cost of delegated monitoring,
information rents paid to the domestic bank, and repayment of interest to the domestic
depositors for using their funds in lending. The delegation cost occurs because the
domestic bank must earn at least a zero proﬁt in order to provide services to the foreign
investor. This cost is even bigger with asymmetric information. In this case, the foreign
investor suffers some misallocation of capital because he must guarantee that the contract
with the domestic bank is incentive compatible. The information rents represent a direct
payment to the domestic bank for providing information services to the foreign investor.

The agency costs (delegation costs and the information rents) determine the

amount of capital that is invested in the domestic economy through correspondent

28

banking. Should entry of a foreign subsidiary occur (i.e. for G low enough), then in a
benchmark case with a=1, the capital channeled through correspondent banking is always
smaller than the capital imported by the foreign subsidiary.

Proposition 3

Given a=1, the correspondent banking contract with perfect information always

speciﬁes a smaller capital investment and a larger monitoring effort than the ones
. . . . . 0 OH FI
determined by a subsrdiary of the foreign Investor, K <K and e>e .

Proof. The ﬁrst order conditions (8) and (9) in the problem of the foreign
subsidiary satisfy the equilibrium condition (Al.12), which is described in Appendix 1.
Therefore, the foreign subsidiary’s allocation lies on the locus of equilibrium points

determined in correspondent banking. Moreover, since

oFI d
,F(£__K-J_,
m

the foreign subsidiary’s allocation is the same as that in correspondent banking,
when the foreign investor earns zero proﬁt. Thus, this allocation is to the leﬁ of the
allocation in the perfect information contract.

Proposition 4

Given a =1, in separating equilibrium the correspondent banking contract with

asymmetric information speciﬁes lower capital than the one determined by the foreign

. . . IC FI IC FI
subSIdrary In every state, K10 < K10 and K20 <K2o .

Proof. In state 0 =1, the incentive compatible contract allows smaller investment
than the perfect information contract, implying an additional underinvestment compared

to the entrance of the foreign subsidiary.

29

In correspondent banking with asymmetric information, the foreign investor must
earn positive proﬁt in state a = 2, or there will be no separating equilibrium. Therefore,
the capital-effort allocation given by this contract must be to the right of the foreign
subsidiary’s allocation.

Propositions 3 and 4 are demonstrated on Figure 1-3. In correspondent banking
with perfect information, the domestic economy faces a moderate amount of capital
inﬂow in both states. Asymmetric information increases the volatility of capital inﬂows
since the projects enjoy excessive ﬁnancing in the good state, while the projects are
underfunded in the bad state. The foreign subsidiary may or may not increase the

volatility of capital, but will bring more of it in every state.

Figure 1-3: Foreign Bank Entrance

K0

Ilib=II lbIC [:1
H2
nzblc
KZOF I
. LEGEND
Kzolc // / b A Entrance of foreign

 

 

 

 

 

 

 

 

 

 

 

 

 

 

K I OF I // “2 subsidiary
o _--"__/_..—-——-—-/ / . Correspondent banking
K2 -- with incentive-
0 1 compatible contract
K1 _,
—-——--"""" . Correspondent banking
with perfect information
e
ellC 61F el e2F! 621C 32

30

Since there is an inverse relationship between the foreign investor’s monitoring
efﬁciency and the capital inﬂow, the conclusions above are reinforced when a<1. In this
case the equilibrium points presented on the graph will drift up and to the right, thereby

increasing both capital and monitoring effort.

3. The Foreign Investor’s Choice
The foreign investor chooses the investment option that delivers the highest proﬁt,
so his selection depends on the severity of the agency cost, the set up cost G, and the

credit market concentration in duopoly that depends on the relative capital endowments
of the two banks.

The difference between the proﬁt in direct investment and in correspondent
. ID] IIC . .
banking, [1F - 17F shows how large the agency cost 18 relative to the entrance costs.

The agency cost has two components: a cost from misallocation of capital and effort, and

information rents:

Agency Cost = pzmkl —I91)C(e{C)—(l —02)C(e{C01 /62 )1
+ Pi [II l1:101 (Klan 'elFl 'm)_ I111.701 (KlolC ,e{C,m,a = 1)]

+ p2 [n 2F1131 (K3171 , €57 ' m)_ [-1510] (K2010 , 650' m, a = 1)]

where 115”” (KgF 1 ,e51 ,m)0'=l,2 is the maximum proﬁt of the foreign

subsidiary, given that it ﬁnances all domestic projects, while

115”” IKgIC,ecI,C,m,a = I) a = 1,2 is the proﬁt of the foreign subsidiary evaluated at the

allocation in correspondent banking with asymmetric information and a=I.

p2m[(1 — 6| )C(e11C )— (1 - 62 )C(e1I C61 / I92 )] represents the information rents.

31

The entrance cost consists of reduction in proﬁt due to competition with the

domestic bank and the ﬁxed cost of setting up the subsidiary:

Entrance Cost = plmlb [ﬁler] F(k1FI )— (1 — I91 )aC (61F I )— 1".le I I
+ p2m§[62e{1F(k{1)— (1 — 62 )aC(e2FI )— r‘kpf‘l ]+ G
The entrance cost becomes larger if the domestic economy owns abundant

resources that can be used by the bank to produce loans. Kd reduces the number of

projects that can be served by the foreign investor, but does not affect the total capital
invested in the country or the equilibrium monitoring effort.
The monitoring efﬁciency coefﬁcient reduces both the agency cost and the

entrance cost. A larger a lowers 1'1ng 1 (K3157 ,2? ,m) but does not affect

1'1ng 1 (Kglc,e,l,c,m,a =1) and thus it may even mitigate the information rents. Also, an

increase in 3 reduces the agency cost by a larger proportion than the entrance cost.
Therefore, it is more likely that an investor who faces monitoring disadvantage will
choose correspondent banking, while an investor with strong monitoring expertise will
prefer direct lending or opening a subsidiary.

If the host economy has negligible capital endowment, small barriers to foreign
entry, and a fairly underdeveloped ﬁnancial system (a<I), the foreign investor would
prefer to open a subsidiary. This is because the moral hazard costs play a signiﬁcant role
and the foreign investor has monitoring advantage.

A country with small capital endowment, high barriers to entry and an
underdeveloped ﬁnancial system, but a transparent real sector (6, and 62 are either known

by the foreign investor or are almost equal), will encourage the foreign investor to grant

32

direct loans rather than open a subsidiary or use correspondent banking.
A country with abundant capital resources and a relatively more developed
ﬁnancial system (a>1) would see inﬂow of capital mostly through correspondent

banking.

4. Welfare Implications
Firm proﬁts and the interrnediation services provided by the domestic bank are
the main sources of welfare gains for the host country. Because of this, the results of the
model suggest that the stringency of barriers to entry in the domestic banking system will
depend on the efﬁciency of capital absorption by the ﬁrms and the monitoring efﬁciency
of the domestic banking sector.

In autarky, the total welfare in the host country is measured by the utility of the

depositors since no other domestic agent has positive consumption, ”/1 = nU(rsd). With

foreign investment, regardless whether it is correspondent banking or a subsidiary, there
is lending and the ﬁrm owners earn proﬁts. In addition, with correspondent banking, the

domestic banker raises income by providing agency services to the foreign investor.

. [C [C .
Thus, the total welfare In state 0 =1 and state 0 =2, W1 and W 2 , when the domestic

bank is entrusted with monitoring is

01C d
WIIC = nUIr‘m°)+ m(1 - e,’C )91F[£—+—K—] (13)
m

01C (1
WZIC = nU(rsw°)+ m(1- egc )92F[——K1 + K ]
m

+m[(1-o,)c(e{C)— (tandem/62)] (13')

33

If the foreign investor sets up a subsidiary, the domestic bank earns proﬁts only
from the projects it ﬁnances and not from agency. Also, it is assumed that the entire
ﬁxed set up cost, G, or some of it, is revenue for the domestic economy. This revenue is
distributed as a lump sum transfer to the domestic agents and brings additional welfare
gain AW(G). Thus, the total welfare of the domestic economy when the foreign investor

opens a subsidiary is

W: = nUIrsco 0I+ mII- egI90 (kb I
+ m3[t9,,e3F (k3I— (1- 6,, )CIe3II— rst + AW”(G), a = 1,2 (14)
Suppose that there is a government that determines G in a way that maximizes

total domestic welfare, and that the government does not know the average quality of the
projects at t=0, when the ﬁxed cost is announced to the foreign investor. Also, assume
that G is transferred to the domestic agents in the form of a public good that yields linear
utility to everyone. The maximum G the government can set without discouraging
foreign entrance is

G : p1 I—IIFIDI(K10FI,81FI m)-1_IFIDI(K101C811C HUI

+[02[HF11)1(K0R1£2 F] m)__ I—IZF11)1(1,(01C82 [C m)]
—p]m]b[6]e1nF(le I (l 61)aC(e1F 1)— r le 1]
—p2m1[02e§7 FIk2 )- (1- 92)aC(e2F ’)— r k” I
+ szI(1- oiicreIC1-(1— 6216(ei’cei /02)]+ ,st
At G -- set so that the foreign investor is indifferent between correspondent
banking and entrance -- the expected welfare of the domestic economy with a foreign

subsidiary is greatest and it equals

34

15(W” ) = I’ll/(rswo ) + p1m(1- ef)91FIki’I+ 10sz - 83 WWI"? I

+pimi’IaierFIki’I—(I-axref’k

+pzm’z’lezeé’FIké’I—(I-62)C(e$’)]

+ pl[n1171D1(KloFI,elFl,m)_anIDI(KlolC,ellC,m)]

+ p2 ”5101(K3F1,e§‘1,m)_H§‘IDI(K201C,e£C,m)]

— piml’IBIef’Fka’I— (1 — 6i)aC(eiF’)- r’kf’]

—p2mI’I62e§’F(k2”I—(1-02)acref’)—r’k{’l

+p2m[(1—6110(e{C1—(1~62)C(e{C61my] (15)
The expected welfare with correspondent banking does not depend on the

government revenue

01C d
E(W’C ) = nU(rsa)0 ) + p1m(1 - e{C )01F[K'—+I£—I

m

+ mmtI—e5C162FIK2
+ szI(1- 6i )Crei’C) ~11 - 9,)c(e{Ca, #221] (16)
Proposition 5
Given a=1, the host economy is unambiguously better off allowing entrance of the
foreign subsidiary if ellc >e1“ and the government collects the maximum entrance cost.
Proof. See Appendix 3.
Proposition 5, states that if the foreign subsidiary does not put pressure on
monitoring costs, then foreign entrance is overall beneﬁcial for the economy. It is
important to notice that the government collects the whole amount of the entrance cost.

If this is not possible, ﬁrms’ and the bank’s duopoly proﬁts may not be enough to

35

compensate for the loss in income from bank’s agency services”. The domestic

economy may still gain from entrance if its production technology is very elastic with
respect to capital, so the additional capital brought by the foreign subsidiary will boost
both ﬁrms’ and the bank’s revenues. Also, the domestic bank should not lose much of its
market share.

If the foreign investor puts more effort than the correspondent bank when the
quality of the projects is low, the cost of borrowing increases for the projects owners.
This also puts pressure on the monitoring cost of the domestic banks, thus making foreign
entrance less desirable.

Welfare effects become more ambiguous once the monitoring efﬁciency of the

foreign investor is taken into account. Appendix 4 shows that as long as

 

F'(k)k+ C(e)
F(k) eC'(e)

>1 the maximum welfare with a foreign subsidiary increases as the
monitoring efﬁciency of the foreign investor increases. Therefore, as long as domestic
output has high elasticity with respect to capital and the monitoring cost has low elasticity
with respect to effort, the host economy may beneﬁt from foreign bank participation.
More capital is utilized in the country and project owners enjoy signiﬁcant proﬁts, while
the competitiveness of the domestic bank is not banned.

If this condition is not satisﬁed, an efficient foreign bank would easily deplete the

domestic bank’s proﬁt and foreign bank entry may not be as good an option for the

country as correspondent banking.

 

'5 This would become even more evident if the bank was earning proﬁts in autarky. Autarky proﬁt is the
reservation proﬁt that the domestic bank would demand and receive in a correspondent relationship.
However, there is no such guarantee when the foreign subsidiary is allowed to enter, so the domestic bank
may suffer losses from competition as well.

36

5. Summary and Conclusions

This paper attempts to formally describe the organizational form (correspondent
banking, and foreign bank entrance) of a foreign investor (bank) in a host country. The
main assumptions of the paper include a layered asymmetric information problem and a
possibility that the foreign investor has a different cost structure from the domestic bank.

The model yields important results about the amount, channel, and welfare effects
of capital ﬂows, depending on the characteristics of the host economy and the foreign
investor. A very trivial outcome comes from the interaction of the foreign and the
domestic bank. As described by, Gehrig (1998), Claessens et al. (1998), Weller (1999,
2000) «who focus on foreign bank entrance and bank competition only -- foreign banks
reduce duopoly proﬁts and raise concerns about systemic risk. In the current model, this
comparison could be made between correspondent banking and entrance. Entrance
reduces the incumbent bank’s market share and eliminates its income from agency
services and thus it attenuates the host economies’ proﬁts in the banking system.

Trivially, the paper concludes that entrants with high monitoring costs and small
capital contributions may earn lower proﬁts than domestic banks (as in developed
economies), while entrants who have more efficient monitoring would proﬁt relatively
more than the incumbent bank (as in developing economies). Such results are
empirically shown by Claessens et. al. (1998).

More interesting are the conclusions of the current model about the amount of
capital ﬂows. In correspondent banking the economy is prone to overborrowing and

pronounced boom-bust cycles. Extensive literature -- Schneider and Tomell (2001),

37

MacKinnon and Pill (1999), Burnside et al. (2001) -- ascribes overborrowing to the
existence of formal or informal bailout guarantees. In this paper, as in Gianetti (1998),
bailout guarantees are not necessary to obtain the same result. In the model presented
here, overborrowing arises because the domestic bank would only agree to prudently
check the outcome of the projects in the good state if the foreign investor rewards it with
additional proﬁt. To provide this additional proﬁt, the foreign investor must pour more
capital into the projects in the good state and squeeze investments in the bad state.
However, those exaggerated cycles are not necessarily bad since, ex-ante, the welfare of
the host economy may be higher than in the perfect information case”.

Another important result of the model is that the welfare of the host economy is
affected not only by the amount of capital inﬂow but also by the effort exerted in the
banking system. Since a very efficient foreign entrant would extract the proﬁts of the
domestic project owners and will force the domestic bank to suffer high monitoring costs,
the host country has a perverse incentive too avoid such entrants even if they bring a lot
of capital”. The host country beneﬁts from knowledgeable foreign bankers only if the
domestic projects have very high capital productivity. This productivity will generate
signiﬁcant proﬁts for both the real and the banking sector. Also, the domestic banks
must be able to resist the pressure of rising monitoring costs, and their marginal costs
must not be increasing rapidly. From a policy perspective, the host economy may

consider that large amount of capital inﬂow is not necessarily associated with more

welfare. Thus, the country may want to discriminate between potential foreign investors.

 

'6 In the perfect information case, the domestic bank does not provide information services to the foreign
investor and does not earn proﬁts.

'7 Interestingly, the Economist (Sept. 14‘“, 2002) reports that banks entering the Central European
banking markets «the markets that went through the most extensive ﬁnancial liberalization over the last
decade« are not necessarily the most well established foreign banks.

38

The paper also presents important implications of the model with respect to the
channel of capital ﬂows. McCauley et al. (2002) show that developed countries tend to
receive foreign capital mostly through correspondent banking, while developing countries
receive through local subsidiaries of foreign banks. This conforms with the result that
foreign investors who have relatively more inefﬁcient monitoring technology than the
host country bank will prefer correspondent banking, while efﬁcient investors will enter
the banking market directly.

The paper raises some interesting questions that require further exploration. If it
is true that foreign banks become multinational to “follow-the-customer” and that they
tend to skim borrowers off from the domestic credit market, a model with heterogeneous
borrowers may give a more comprehensive idea about the real effects of multinational
banking. Additional extension could be the introduction of a multi-period model, which
may bring other key factors in cross-border banking into play, such as the terms of the
contracts, exchange rate variations, learning, etc. Finally, an empirical test of the model
could make an important contribution to the understanding of the investment channels in

international banking.

39

REFERENCES

Aizenman, J. 1998. "Capital Mobility in a Second Best World « Moral Hazard with
Costly Financial Interrnediation." NBER Working Paper 6703.

Bain, E., J. Fung, and 1. Harper. 1999. "Multinational Banking." Melbourne Business
School Working Paper 99-5.

Barth, J., G. Caprio, and R. Levine. 2001. "The Regulation and Supervision of Banks
Around the World: A New Database." World Bank Working Paper 2588.

Bemanke, B. and M. Gertler. 1989. "Agency Costs, Net Worth, and Business
Fluctuations." American Economic Review, 79:1, pp. 14-31.

Boadway, R., N. Marceau, and M. Sato. 1999. "Agency and the Design of the Welfare
Systems." Journal of Public Economics, 73: 1999, pp. 1-30.

Burnside, C., M. Eichenbaum, and S. Rebelo. 2001. "Hedging and Financial Fragility in
Fixed Exchange Rate Regimes." European Economic Review, 45:7, pp. 1151-93.

Claessens, S., A. Demirguc-Kunt, and H. Huizinga. 1998. "How Does Foreign Entry
Affect the Domestic Banking Market?" World Bank Working Paper 1918.

Diamond, D. 1984. "Financial Interrnediation and Delegated Monitoring." Review of
Economic Studies, 51, pp. 393-414.

Diamond, D. 1991a. "Financial Interrnediation as Delegated Monitoring: A Simple
Example." Federal Reserve Bank of Richmond Economic Quarterly, 82/3, pp. 51-
66.

Diamond, D. 1991b. "Monitoring and Reputation: The Choice between Bank Loans and
Directly Placed Debt." Journal of Political Economy, August, pp. 689-721.

The Economist. 2002. "Survey: Rogue Trader, Rogue Parent." Vol. 364: 8.

Gehrig, T. 1998. "Screening, Cross-Border Banking, And The Allocation of Credit."
Research in Economics, 52, pp. 387-407.

Gertler, M. and K. Rogoff. 1990. "North South Lending and Endogenous Domestic
Capital Market Inefﬁciencies." Journal of Monetary Economics, 26:2, pp. 245-66.

Giannetti, M. 2003. "On the Causes of Overlending: Are Guarantees on Deposits the
Culprit?" CEPR Discussion Papers 4055.

Gordon, R. and L. Bovenberg. 1996. "Why is Capital So Immobile Internationally?
Possible Explanations and Implications for Capital Income Taxation." American
Economic Review, 86:5, pp. 1057-75.

40

Greenwood, J. and S. Williamson. 1988. "International Financial lntermediation and
Aggregate Fluctuations Under Alternative Exchange Rate Regimes." Journal of
Monetary Economics, 23:3, pp. 401-31.

Horstmann, I. and J. Markusen. 1987. "Licensing Versus Direct Investment - a Model of
Intemalization by the Multinational Enterprise." Canadian Journal of Economics,
XX:3, pp. 464-81.

Horstmann, I. and J. Markusen. 1996. "Exploring New Merits: Direct Investment,
Contractual Relations and the Multinational Enterprise." International Economic
Review, 37:1, pp. 1-19.

Krasa, S. and A. Vilamil. 1992. "Monitoring the Monitor: An Incentive Structure for a
Financial Intermediary." Journal of Economic Theory, 57, pp. 197-221.

Mathieson, D. and G. Schinasi. 2001. "International Capital Markets: Developments,
Prospects, and Key Policy Issues." International Monetary Fund, August 2001.

McCauley, R., J. Ruud, and P. Wooldridge. 2002. "Globalizing International Banking."
BIS Quarterly Review, March 2002, pp 41-51.

McKinnon, R. and H. Pill. 1998. "International Overborrowing: A Decomposition of
Credit and Currency Risks." World Development, 7, pp. 1267-82.

Mookherjee, D. and M. Tsumagari. 2003 "The Organization of Supplier Networks:
Effects of Delegation and Interrnediation." Forthcoming in Econometrica.

Severinov, S. 1999. "Optimal Organization: Centralization, Decentralization or
Delegation?" mimeo, University of Wisconsin.

Schneider, M. and A. Tomell. 2000. "Balance Sheet Effects, Bailout Guarantees, and
Financial Crises." NBER Working Paper 8060.

Townsend, R. 1979. "Optimal Contracts and Competitive Markets with Costly State
Veriﬁcation." Journal of Economic Theory, 21, pp. 265-93.

Weller, C. 1999. "The Connection Between More Multinational Banks and Less Real
Credit in Transition Economies." Center For European Integration Studies
Working Paper B8 1999.

Weller, C. 2000. "Banking On Multinationals: Increased Competition from Large Foreign
Lenders Threatens Domestic Banks, Raises Financial Instability." Economic
Policy Institute Issue Brief # 142.

Williams, B. 1997. "Positive Theories of Multinational Banking: Eclectic Theory Versus
Intemalization Theory." Journal of Economic Surveys, 11:1, pp. 71-100.

41

APPENDICES

Appendix 1. Proof of Proposition 1

From (4) and (5) it follows that

o d\ o d
div-[w K01.-.*...F:[L:K_I
m m
A: )K d
0
eBF' E—j—K— KO/m
k m

o d
,FILLILIKo

= m - (Al. 1)

- [m magi-156:] - eF'IE‘L—KdIKO - m(l — 6)C’(e)j

m m

 

 

 

 

Since A must be no less than zero

0 d o d
..F~[£.:£_IKo/m-.*..9p[1<_i].o (.12)
m m

and
m m

m6F[£0;K«d-] - eF'IﬂIKO — m(1 — 0)C’(e) < 0 (A13)

. . . . . . , K 0 +Kd *
Also, If the foreign Investor 1S to have posrtrve proﬁt, e6F —— —r 2 0,
m

it follows form (4) in the FCC. that A S l and combined with (5), this implies:

6F[-IL;K:I — (1 - 6)C'(e) s o (Al.4)

The proﬁt of the foreign investor is

42

KO—r K0

o d
HF1=RK°-r*K0=et9F'[—K +K I *
m

After ﬁrll differentiation

0 d o d o d
An = .FIIwIKo...I..p~[L<—+K_Ito 1....FIL:£_I-.*I.K

m m m

o d
..FIK—iIKoA

m

It follows from (A12) and the concavity of F(k) that

 

 

 

 

 

o d
—6F'[——K +K IK"
m
g— = d d > 0 (A15)
8 0 0
I..F~[K_~:1<_]Ko/....FILLLI-:I
m m
‘11:] = d 1 d . <0 (A16)
0 0
0'“ I..F~[1<+_KIKo/..AFIK_:LLI-:I
m m
o d
_F[K_:K_IK0
m
521% = , 0 d d > o (A1.7)
0
I..F~[1<_:_I<_.Ixo/m...FIK_:_I<_I-rI
m m
and
o d o d o d
0F £_+_K__ K0 0F'(K0/m) 51L KO/m+gp' §__+_K_
de m m m
6182 _ KO‘I‘Kd KO'I’Kd * 2 (A18)
e0F" ————— Ko/m+e6F' —— —r
m m

43

Since 6F "( K 0 / m )K 0 / m + 0F '( K 0 / m) may be either positive or negative, the
isoproﬁt curves of the foreign investor may be either convex or concave. In any case,

they are monotone increasing, and with steeper slope in state o=2.

Figure 1-4: Foreign Investor’s Isoproﬁt Curves

K° Fl 0

II _, K l_[2FI I-IIFI

 

 

 

 

/ . .

The proﬁt of the domestic bank is

K°+Kd

I'Ib = me6F[
m

I—RKO—mtl—9)C(e)-rSKd

o d o d
= me6F[-I-<——+£—I - e6F'[£—+£—]KO — m(1- 6)C(e) — rSKd

m m

After ﬁrll differentiation

0 d o d
dubIMFIQLIAIHIK--.AWIAIA
m

m

o d o d o d
-I..F~[£:£<_IKo/.I.KoIMFIELIAIELIW...ch...
m m m

It follows from (A l .3) and the concavity of F(k) that

44

Also,

 

 

 

 

 

dK" m
de = K0 Kd >0 (141.9)
et9F'[ + IKo/m
m
0
de= K0 le >0 (A1.10)
dI'I e0F'[ + IKo/m
m
o d o d
0 meF[K +K ]— F’[K +K IKO+mC(e)
dK m m
d6 - 0 d <0 (Al.11)
eBF" K +K KO/m
m
d2K0 = —m((1—o))C'(e)
2 o d
.. ..F~[£_:_£<_]Ko/m
m

o d o d
mgFILgLI _ gpr[§__+_’f_]go - ,0 _ 9“,)
m

m
o d
.2.F~[LiIKo/m

 

[in 6F[§::£i] — 0F'[—K—0;—1£:]KO — m(l - 6)C'(e) + m(l — 9)C"(e)]

m
d 2 0
0
e29F'[K—iK—-]KO / m

 

m
assuming C" 2 0 (Al. 12)

The isoproﬁt curves of the domestic bank are monotone increasing, convex, and

with smaller slope in state o=2.

45

Figure 1-5: Domestic Bank’s Isoprofit Curves
KU K0

III

 

 

 

 

Figure 1-6: Symmetric Information Equilibrium

KO

 

 

 

 

 

 

 

b b .
The curves III and IIZ are the zero proﬁt curves for the domestic bank. The

46

F1 F1 . . .
curves [[1 and “2 demonstrate the highest proﬁt that the forergn Investor could

achieve in each state, given zero proﬁt for the domestic bank. The isoproﬁt curves of the

foreign investor and the domestic bank are tangent in equilibrium.
The point (K 10, e1) determines the equilibrium contract (K 10, R1) when the bank
has poor portfolio. In the good state the isoproﬁt curve of the foreign investor is steeper,

while the isoproﬁt curve of the domestic bank is ﬂatter at K10. Therefore, the

equilibrium capital K 20 is higher when the domestic bank has good loan portfolio.

Since the zero-proﬁt curve for the domestic bank is ﬂatter in the good state, the

equilibrium effort, eg, and the interest rate paid to the foreign investor, R 2, must be

higher, as well.

To guarantee existence of the equilibrium when the isoproﬁt curves of the foreign
investor are convex, the slope of the isoproﬁt curve of the domestic bank must be smaller
than the slope of the isoproﬁt curve for the foreign investor for small e, and larger for

large e.

47

Appendix 2. Proof of Proposition 2

The ﬁrst order conditions for the problem with asymmetric information are

d
w.r.t.e1: PII:91F'[K 1—:—;K—]Kf]

+(—Ai+izImAFI—’g—g’f—I— AFIK‘—;——K— IKr— m(— —61)C(ei)I

—21Im(1—oi)c'(ei)- m( -e.>c[e —-II=o (A21)
d o d
w.r.t. K10. p1[8191F{K——10———— + K ————]K10 /m — r* + e161F'[£L_+K_]]
m m
K0 + Kd
-(— 2.1 +2, e161F" 4;— Kf/m = 0 (A22)
0 d
w.r.t.e7_ : p2[02F’[£2—;£—]K§]
o d o d
.i,I.g,F[£2_;LI-.,FI.I<_2;-1LIK,-..(i-.,)c(.,)I (A...)
o d o d
wK, .. .,I.,.,F~[£2.:L<_IK,/m-.* ”IE—ill
m m
o d
-i,I.,.,F~[K2_+K_IK,/.I=o (A...)
m

Where A.) and 22 are the multipliers of ICBI and IRBZ constraints, respectively.

From (A21) and (A22) it follows that

48

m
K10 +Kd
m

[—r*+e191F'[-Ki;-£:]:'[MB1F[K—l——+Kd)“91F[-K—-——l:1KdK10]"’"1‘(‘91)C'(el)]

 

e101F"[————— Klo/m

[e 61F[K—1"—— ;Kd]Kr/m][m swf‘l" ,jf ]-m(1-al)c'(el)]

+* d
e161,}r "(L—1 +K ]K10/m
m

 

-g[m(1-al)c'(e1)-mo-ez)c'[e,22]]:(1 (A25)

Since the multipliers must be positive and
" K10 +Kd 0
816117 —— K1 /m<0
m

in state 0:1, A1.12 is negative in the equilibrium with asymmetric information

and thus capital and effort must be smaller than the ones obtained in the perfect
information equilibrium (K101 C<K1 oand e1 IC<e1).

To prove the second part of the proposition, we use the second order conditions
for the maximization problem with perfect information. The determinant of the bordered

Hessian is:

-L L Lee—L 0L 0 )+L,,(LM0LK,—L 0 oL,,)20

Axoi 1x0 M K e

Where Li]- are the cross derivatives of the Lagrangean from the perfect

information problem, i,j=l, K”, e.

Since

49

——>OandLAe

m

K°+Kd K0
m

Limo = -e6F [

m m

o d o d o
= mgp[1<_:£_] _ 9F1[£_+1<_]§_ _ ma _ WW, < 0
m

it is sufﬁcient to have

Lee—L L )<o

L
( 1K0 [111:0 K08

and

(LAKOLKE — onx" Lle)< O

The ﬁrst order conditions with respect to K0 and e from the perfect information
problem provide the locus of equilibrium combinations of those two variables from the
unrestricted objective function. We need to ﬁnd the slope of the relationship between K0
and 2. Full differentiation of these ﬁrst order conditions yields:

onx" dK" + onede + I'M" d). = 0

onedKo + Leede + L led/1 = O

or

(Limo 1‘“ ' [31x0 [’er P3 = (onx0 Lie — szo L“? )‘IKO

Therefore, dKo/de<0. This implies that the locus of equilibrium points lies on a

negatively sloped curve.

Thus, if the proﬁt of the domestic bank is to increase above zero, the proﬁt of the
foreign investor should drop down, more capital should be invested in the domestic

economy and the domestic bank should put less monitoring effort. This is what occurs in

50

state 0:2 with asymmetric information.

o d 0 d
b F] "16F Li -0F’ _K__+_I(_
[dKOJ _[dK0J _ m

de de 0 ‘1
egF-JhK—JKo/m

JKO - m(l — 6)C'(e)

m

m

o d
-9FJQLJK0
m

o d 0 d :
eewJK—iJKo/mwFJwaJ

 

J—

m m

" o d o d o d
.9F~[.I<_+_K_JKo/m..gFrJﬁaJJJngFJLiJWa_a)cv(.)J
m

m m

 

o d o d i o d
JegKJLtLJKo m + egF'JLﬂLJ- ”JJQFJLLJKo “"J
m m m

o a' f o d
JeQFJK—fLJJJOF' £_:£_JK0J
m K m

- o d o i d‘ a d *
JeaFJﬁLJKo/mprm -waJKFJLLJKW/J

m m m
/

 

The sign of this difference is determined by

F o d o d
_ ,QFJLLJqJJQKJLLJKJJ
m m

—— K0 /m + e0F'[§—+§—d—J - r*J[m0F[£:i-£:J - "1(1- 9)C'(6)J =

m

z KJLLJJKJLJKJAJKmon

m m

 

—— K0 /mJJm 9,:[537251] — m(1 — 0)C’(e J (Al. 12)

5]

When e=0, (Al.12) is negative because of the concavity of F (k). When e=1, the

term is positive as long as C'(1)—> oo . This proves the existence of equilibrium.

52

Appendix 3. Proof of Proposition 5
When a=1, kFI=kb and eFI=eb. Therefore, to proof the proposition, it is enough

to show that

0F] (1
plm-(I elb_1_m_)91F[K1+ K ]+p1[I-I1FIDI(K10FI ,elFI ,m)— HIFIDI(K101C,eIIC,m)J

01C d
>p1m(1 e{C)91F[K———m— +K J (A3.1)

After some algebraic manipulations, (A3. I) is transformed into

0F] d 01C d
m91[F[K1+K ]_F(K1 +K ]J_rt(K10F[-Klolc)

m m
>m(1-61)[C(eln)- C(e{C)J

Let KIFI - KllC = h. Then

K10“ +K" K10“ —h+K" .. , K10” +Kd ..
mal F———— -F -r 109,1: ————h—rh

 

m m m
I K10F1+Kd KIOFI'i'Kd
=61F —— h— (91F —— h=0
m m

Since m(1- -61 )[C(e1FI )- C(ell C)J<O only 1f e11C>e1F this condition guarantees

. . . . [C
that the domest1c welfare Will be h1gher under entrance 1n state 0 =1. In state 0 =2, 82

F1 . . . .
>e2 1s always sat1sﬁed and entrance 1s always beneﬁc1al.

53

Appendix 4. Proof of Statement
We need to derive the conditions under which the welfare in the host economy
decreases in a. Therefore we need to establish the relationship between the monitoring
efficiency, a, and the entrance cost, G, the proﬁt of the domestic bank,
m3 J93e3FJk3 )— (1 - 0,, )C(e3)J, and the proﬁt ofthe project owners, (1 - e3 )6,,F(k3 J.
The terms in G that are affected by a, are given by the proﬁt of the foreign
investor. Using the envelope theorem, the proﬁt of the foreign investor decreases in (1,

therefore G also decreases in a.

. . b .
Next, we use the prev1ously ment1oned result that de /da<0. Smce

(1-e3)9,,F(k3)=(1-e3)(1— 9,, )C'(e3 ) and given that — C'(e3 ) +eC"(e3 ) > 0, it
immediately follows that the proﬁt of the project owners increases in e" and decreases in

0.

Also, since kg = K d / mg , then dm,b7 = -(m,b,/kg M3, and since

F,F(Fg)= (l _F,)o(eg), then 0,,F'(k3}ik3 = (1-9,)c~(eg Fb

After full differentiation of the proﬁt of the domestic bank, assuming d60=01

- (mg/k3 )(1 16016"
6,,F'Jk3)

ng F15?) BaegF'lkr’ik’: - ,FgF(Fg)+ <1-6F>C<e3> e3

___m3(1—t9,, )C'(e3)t9,,e3F F(k3) F'(k3)r3_ +_____ bC(e3)
19,,FJ/t3 )r3 F(k3) e3C (e3)d

"Therefore, the proﬁt of the domestic bank increases in the effort and decreases in

(eg)le,e3F(k3)- (1- Face?) e3+m3lea(1 9010760) F,

 

 

 

54

aif

l

F(ngg + cog)

 

F(k3) e3C'(e3)

J»

55

PART TWO. CREDIT DOLLARIZATION IN TRANSITION ECONOMIES: IS IT

FIRMS’ OR BANKS’ “FAULT”?’8

1. Introduction

This paper analyzes the respective contributions of banks and ﬁrms to the
phenomenon of credit dollarizationl9 in transition economies. We seek to provide
answers to the following questions: Why do domestic banks in these countries lend in
foreign currencies? Why do domestic ﬁrms borrow domestically in foreign currencies?
Is credit dollarization mainly supply or demand-driven? Ultimately, we would like to
contribute to a better assessment of the sectoral balance sheet risks associated with
ﬁnancial dollarization.

An issue that currently draws economists’ attention is whether credit dollarization
increases the riskiness of ﬁrms’, banks’, and governments’ balance sheets. In a small
open economy which is dependent on foreign capital and external demand for its exports,
credit dollarization determines the distribution of currency risk across different sectors of
the economy. Overall, an economy may possess dollar earning assets similar to its dollar
bearing liabilities and still face widespread defaults as a result of unexpected currency
depreciation if one sector bears disproportionately large currency risk. As Allen et al.
(2002), Delgado et al. (2002), and others suggest, the concentration of currency

imbalances in one sector may lead to a spill-over of ﬁnancial difﬁculties throughout the

 

'8 This paper is co-authored with Alina C. Luca, Department of Economics and International Business,
Drexel University.

'9 Financial dollarization occurs when a signiﬁcant share of residents’ domestic ﬁnancial contracts
(including deposits and loans) are denominated in foreign currencies, not necessarily the US dollar. Dollar
deposits and loans denote all foreign currency deposits and loans. In addition, deposit dollarization and
credit dollarization refer to the phenomenon of denominating deposits and loans in foreign currencies.

56

economy.

As emerging (including transition) economies lack domestic capital, they often
turn to international markets to ﬁnance production. Years of exchange rate instability
have also created persistently high deposit dollarization. Thus, banks channel resources
to domestic borrowers that are increasingly denominated in foreign currency and oﬁen
exceed the capacity of the tradable sector to absorb them. These resources create an
imbalance that increases market risks for banks and ﬁrms and potentially raise the costs
of government intervention during ﬁnancial crises.

The domestic credit market determines whether banks or ﬁrms will bear the
exchange rate cost of the foreign currency liabilities. Given relative market power and
banks’ limited hedging opportunities, bank currency matching activities may determine
credit dollarization, thus transferring the exchange rate risk onto the borrowers. As a
result, a depreciation of the local currency deteriorates borrowers’ balance sheets, reduces
their creditworthiness, and intensifies asymmetric information problems (see Mishkin
(1996)). Balance sheet worsening triggers ﬁnancial accelerator effects, ﬁrms experience
reduction in credit supply and production interruption, and the economy is likely to suffer
a recession. If ﬁrms drive credit dollarization, banks work to cover the exchange rate
risk. Depreciation may increase bank exchange rate losses but banks do not necessarily
face additional default and information costs. Thus, economic costs associated with
ﬁrms’ and banks mismatches seem to be different and possibly higher if ﬁrms rather than
banks are ﬁrst hit by the exchange rate depreciation.

In any case, currency mismatches increase ﬁnancial fragility and make economies

more prone to banking crises. Because of high bank liquidity and solvency risks, banks

57

face a higher premium on foreign and domestic capital. This brings multiple and self-
enforcing problems to dollarized economies. On one hand, the risk premium worsens
ﬁrms’ balance sheets and tightens their borrowing constraints, so overall credit ﬁnance
and economic growth are lower than in countries where currency mismatches do not
exist.

On the other hand, foreign and domestic depositors are more sensitive to news
about exchange rate movements in economies with vast currency mismatches. The
reason is that depreciation directly and indirectly brings banks’ proﬁts down. Thus,
dollarized economies tend to experience abrupt capital reversals. Levi-Yeyati (2003)
points out that balance sheet currency mismatches amplify the cyclicality of capital ﬂows
and slow down the recovery from an economic downturn. Many governments attempt to
alleviate this problem by smoothing exchange rate ﬂuctuations —— a policy that Calvo and
Reinhart (2002) named “fear of ﬂoating”. However, the tendency to reduce exchange
rate ﬂuctuations may induce ﬁrms and banks to maintain large currency imbalances, thus
raising ﬁnancial crises costs even further.

To summarize, we believe that the factors explaining credit dollarization also
explain why grossly disproportionate currency mismatches exist in small open economies
with underdeveloped ﬁnancial systems. We take a close look at the domestic credit
market and examine the risk sharing forces behind credit dollarization. Is credit
dollarization the outcome of a unilateral decision made by domestic banks? Or is it
driven by domestic ﬁrms’ integration in the international marketplace?

This work draws on literature concerning the determinants of currency

composition of banks’ portfolios and corporate debt. Calvo (2001, 2002) proposes bank

58

currency matching as a possible explanation for the high levels of dollar loans in
emerging economies, given that deposits are highly dollarized. Ize and Levy Yeyati
(2003) also argue that banks hold foreign exchange matched positions, not only because
of regulatory requirements, but also because they are “generally adverse to assuming risk
directly” (pg. 8). However, most studies focus on ﬁrms’ hedging behavior. A portfolio of
loans denominated in more than one currency can provide an improved risk and cost
position for ﬁrms (see Kawai (1981), Cotner (1991)). A number of empirical papers
provide evidence that ﬁrms borrow in dollars for hedging reasons — larger ﬁrms, and
firms with exporting activities, borrow more in foreign currencies”.

The abundance of studies on ﬁrms’ currency matching incentives stems from the
assumption that ﬁrms are risk averse, while banks can fully diversify. Thus, credit
dollarization depends on the covariance of ﬁrms’ revenues with the exchange rate.
However, when it comes to explaining ﬁnancial crises a bulk of the balance sheet
imbalances literature (Krugman (1999), Cespedes et al. (2000), Gertler et al., (2003),
Aghion et al. (2001)) focuses on the mismatch between the size of the tradables sector
and foreign currency borrowing. Only a few papers (see Burnside et al. (2000), Choi and
Cook (2002)) assume explicitly that banks bear the exchange rate risk by maintaining a
mismatched balance sheet. Thus, the questions of whether supply or demand factors
determine credit dollarization and which sector faces a higher exchange rate risk are still
Open.

Currently, only two empirical studies, Arteta (2002) and Barajas and Morales

 

2° See Aguiar (2002), and Martinez and Werner (2002), for evidence for ﬁrms from developing
economies, Keloharju and Niskanen (2001), Kedia and Mozumdar (2002), and Giraldi and Hamaui (1991),
for evidence for ﬁrms from developed economies.

S9

(2003), analyze the determinants of credit dollarization. Arteta (2002) evaluates the
effect of the exchange rate regime on credit dollarization and essentially tests whether
banks’ hedging incentives (deposit and loan currency matching) are distorted by
exchange rate policies. Arteta’s assumption is that supply determines credit dollarization
and ﬁrms’ hedging incentives are not considered. Overall, the paper provides evidence
that a more ﬂexible exchange rate regime tends to amplify bank currency mismatches as
credit dollarization does not change (or it decreases), while deposit dollarization
increases. Barajas and Morales (2003) use data on bank- and ﬁrm-speciﬁc factors, in
addition to macroeconomic variables, for fourteen Latin American economies to explore
possible biases introduced by domestic monetary policy into the decisions by banks and
ﬁrms to lend and borrow in dollars. Their results indicate that both bank asset and ﬁrm
liability allocation decisions are important determinants of dollarization. They also ﬁnd
limited evidence that greater volatility in the exchange rate decreases the level of credit
dollarization, at least in the short run.

This paper extends the existing work on credit dollarization, ﬁrst, by using a
comprehensive, newly-assembled dataset for twenty-two transition economies from
Central and Eastern Europe and Central Asia, for the period 1990 — 2001.

We estimate and compare the respective contribution of ﬁrms and banks to the
ﬁnancial dollarization phenomenon by grouping potential determinants into bank- and
ﬁrm-speciﬁc factors. Bank-speciﬁc factors include indicators of asset and liability

currency matching. As ﬁrm-speciﬁc factors, we use measures of real dollarization2| and

 

2' We deﬁne real dollarization as the denomination of the returns and costs of production for ﬁrms in
foreign currencies. That is, ﬁrms export the ﬁnal good and/or use imported intermediate goods in the
production of the ﬁnal good.

60

access to alternative ﬁnancing sources. In addition, we include speciﬁc indicators of
overall hedging opportunities, liberalization and deregulation of the foreign exchange
market, uncertainty and lack of credibility of domestic policies, and cost of foreign
capital.

We estimate a reduced form equation, and use the pooled OLS estimator as a
benchmark. We then extend the analysis and use ﬁrst difference and ﬁxed effects
estimators to control for unobserved heterogeneity. The main results seem to be robust
across speciﬁcations.

Empirical results provide evidence that credit dollarization in transition
economies is determined by banks’ optimization decisions (asset and liability
management variables). Banks match the currency of denomination of their assets and
liabilities, that is, deposit dollarization drives credit dollarization, and net foreign assets
represent a substitute for dollar loans to domestic ﬁrms.

There is limited evidence of ﬁrms hedging the currency of their debt and
production returns. A higher ratio of exports to GDP increases, while existence of
alternative ﬁnancing sources (speciﬁcally non-bank trade credit) reduces credit
dollarization. More generally, the more integrated is the economy in the international
goods market, the higher is credit dollarization. This result suggests that integration of
international goods and ﬁnancial markets, real and ﬁnancial dollarization respectively,
move in the same direction. It questions the previous (counterintuitive) ﬁnding (see
Arteta (2002) and Barajas and Morales (2003)) that the openness of the economy has a
negative effect on credit dollarization.

The paper also provides evidence that lack of monetary policy credibility and

61

macroeconomic uncertainty (measured by the central bank’s net foreign reserves) raise
dollarization levels.

To summarize, banks in transition economies hedge against exchange rate risk by
holding matched foreign exchange positions. However, as Ize and Parrado (2002) point
out, as long as ﬁnancial dollarization outpaces real dollarization, and this seems to be the
case in transition countries, there is a currency mismatch somewhere in the economy.
We show that banks pass the exchange rate risk to ﬁrms. This decreases banks’ exposure
to currency risk, but it increases their exposure to default risk, and ultimately the
exposure of the economy to ﬁnancial crises.

The remainder of the paper is organized as follows. Section 2 reviews the
theoretical background and presents the main hypotheses for the bank- and ﬁrm-related
effects on credit dollarization. Section 3 outlines the econometric model and the
estimation methods and results, while section 4 describes several robustness tests.

Section 5 presents the conclusions.

2. Conceptual Framework and Main Hypotheses
To assess the extent to which credit dollarization is mainly a ﬁrm or bank
phenomenon, we separate potential determinants into 1) factors that affect only the
supply of foreign currency credit (relative to domestic currency credit), 2) factors that
affect only the demand for foreign currency credit (relative to domestic currency credit),
and 3) factors that affect both demand and supply of credit.
We develop a theoretical framework similar to Ize and Levi-Yeyati (2003) that

employs the idea of a minimum variance portfolio. However, deposits in this model are

62

predetermined and the focus is on the credit market. Presumably both ﬁrms and banks
aim at achieving a minimum variance portfolio allocation. While the assumption of
ﬁrms’ risk aversion is common in the literature, it is usually considered that banks can
spread out risks completely. We assume that at low levels of ﬁnancial development
(missing forward markets) agents’ ability to diversify risks is limited. Thus, both ﬁrms
and banks are risk averse.

This is a static short-run model. F irms’ long-run production decisions are
determined in advance by the availability of capital and ﬁrms make short-run ﬁnancing
choices given a desired level of output. This assumption is consistent with the
observation that loans in transition economies are of primarily short duration.

Firms trade both as exporters of ﬁnal product and as importers of inputs. They
also sell their output and buy inputs in the domestic market. Thus, ﬁrms’ proﬁts are
imperfectly correlated with the depreciation of the exchange rate.

Banks have access to international capital markets and they can both import and
export capital. Banks are perfectly competitive.

Domestic savings are in the hands of domestic consumers. In the short run,
consumer’s deposits with banks are ﬁxed and the proportion of foreign and domestic
currency deposits is predetermined.

These features of the model imply that credit dollarization depends on the world
interest rate, the level of domestic deposit dollarization and foreign assets and liabilities
of banks, the variability of the exchange rate depreciation, and the correlation between
export prices and the exchange rate. F irms’ and banks’ relative risk aversion and market

competitiveness determine the strength of dependence between ﬁrm and bank factors and

63

credit dollarization.

2.1. Firms

Firms are identical and uniformly distributed between 0 and 1. In period 0, ﬁrms
buy a unit of input at price w=1 in the domestic and world markets. The input is used in
proportion to capital, which is ﬁxed and completely determines output, Y, in the short

run. In period 1, ﬁrms sell their output at price p, which is correlated with the exchange

. . . . e
rate deprec1at1on, a, as some of the output 18 1n tradable goods. We assume E(£)=a ,

Var(£)=0'52, E[p]=pe, Var[p]=0'p2, Cov[p(&),£]=0'gp.
Firms maximize expected utility of proﬁts subject to the budget constraint

w=L +Lf. Their proﬁt function is

I'If=pY-w—rLL—rLf(l+£)Lf

where L denotes domestic currency loans, Lf— foreign currency loans, and rL and '

rLf denote the interest rates on domestic and foreign currency loans.
With a coefﬁcient of risk aversion, 7, ﬁrms have the following preferences:

EJIIf J— $- 7Var[l'If J

F irms’ foreign and domestic currency loan demands depend on the covariance
between output prices and the exchange rate depreciation, the variability of the exchange
rate depreciation, and ﬁrms’ risk aversion. Upon reworking ﬁrms’ ﬁrst order conditions,

loan demands become

64

Lf _ —[r[(l+£eJ—rL]-l-}'0'gpl’rLf

2
2
7’0ng

 

L = 1— 1.f

Given a positive spread between domestic and foreign currency loans, ﬁrms
increase borrowing in foreign currency if they have large revenues in hard currency, if
they are not very risk averse, and if the depreciation rate is stable. The expected
exchange rate depreciation increases the cost of borrowing in foreign currency and lowers

the quantity demanded.

2.2. Banks

N banks compete for depositors and borrowers. Banks attract deposits in
domestic and foreign currency from domestic consumers and extend loans in domestic
and foreign currency to domestic producers. Banks also have access to the international
and domestic interbank market, where they can attract additional deposits in domestic

and foreign currency. Each bank’s proﬁt function is
Hb = rLl+ r1{r(l+r:)lf +r(1+.5‘)mf + rIm —de-rg(l +£)df
where 1 denotes domestic currency loans, /- foreign currency loans, d — domestic

currency deposits, and c/— foreign currency deposits. The net position of a bank in the

interbank market (domestic and international) is mf +m = df +d-If —l. The total

amounts of domestic and foreign currency loans, deposits, and interbank funds are L=Nl,

Lf=N/, D=Nd, Df=Naf, M=Nm, and A/=Nn/.

6S

Let 6 be the coefﬁcient of banks’ risk aversion. Banks maximize expected utility

from proﬁts
EJHbJ—éoVarJH” J

subject to the budget constraint: mf + m = df + d — If — I.
From the ﬁrst order conditions, the interest rate on loans and deposits in foreign
currency must equal the world interest rate. The rate on domestic currency loans and

deposits compensates for the expected domestic currency depreciation and uncertainty.

f f

rL :rD =r
rL=rD:rl
and

rLf(l+se)—rL =5r20'3[1f+mf —de
The total supply of foreign currency loans is then
f( . — 1
Lf = NIrL 1+28) rL +[Df_Mf]
6Q] 03
Since the expected rate of devaluation has only an indirect effect through the

spread, it disappears when we substitute for the spread in the foreign currency loan

demand

Lf= N7 “er”, 5 [Df_MfJ
Ny+6 agr Ny+6

The ﬁrst term on the right hand side shows the effect of ﬁrm determinants on loan
dollarization. The higher the correlation between output prices and the exchange rate the

more likely it is that ﬁrms will borrow in foreign currency. Firms with exporting

66

activities, and thus returns denominated in foreign currencies, are hedged against
currency risk if they also borrow in foreign currency. The higher are the exports, relative
to domestic production, the more dollarized the economy should be. This result is
consistent with studies using ﬁrm level data which ﬁnd evidence that the productive

22 High pass-through

structure matters for the corporate debt currency composition.
(domestic prices that increase in parallel with the depreciation rate) has the same effect as
exports and should increase credit dollarization, as well.

The ﬁrst difference of loan dollarization with respect to the world interest rate r is
negative. The world interest rate increases the cost of foreign currency deposits for
banks. With predetermined supply of foreign currency deposits, banks have to pass this
cost onto the borrowers. Thus, if output prices are not correlated with the exchange rate,
loan dollarization is also predetermined (by the foreign currency constraints of the
economy) and the world interest rate has no effect on it. Only the spread increases
proportionately to the foreign currency deposits. If output prices are correlated with the
exchange rate, the increase in the world rate reduces the relative price of output. F irms
substitute domestic currency for dollar loans and the premium on foreign currency loans
is not as large. Thus, given that ﬁrms are risk averse and their output prices are
correlated with the depreciation rate, the world interest rate has only a substitution effect
and reduces loan dollarization. This also results from the fact that output is

predetermined by the amount of capital.

The loan dollarization equation shows that an increase in the variance of the

 

22 Aguiar (2002) and Martinez and Werner (2002) use data on Mexican ﬁrms, and show that exporting
ﬁrms borrow more in foreign currencies. Kedia and Mozumdar (2002) ﬁnd that US ﬁrms performing
foreign operations borrow more in foreign currencies. Keloharju and Niskanen (1997) also ﬁnd evidence
for the hedging explanation using data on Finnish ﬁrms. Giraldi and Hamaui (1991) show that the
currencies used to settle international trade determine the currency composition of debt for Italian ﬁrms.

67

exchange rate reduces dollarization. Exchange rate volatility affects demand and supply
in opposite directions depending on the interest rate spread. The ﬁnal effect is negative
because output prices and the exchange rate are imperfectly correlated. In the extremes
when the correlation is zero, or the exchange rate factors out from the output prices
(perfect correlation), the volatility does not affect credit dollarization.

Banks inﬂuence loan dollarization through currency matching of assets and

liabilities. An increase in deposit dollarization Df increases credit dollarization as banks

match the currency of denomination of their deposits and loans. An increase in net

foreign assets Mr has a negative effect on credit dollarization as banks substitute foreign

currency loans to domestic ﬁrms with foreign assets, for a given level of dollar deposits.

The effect of bank concentration on credit dollarization is not straightforward.
Concentration emphasizes banks’ matching behavior. A more concentrated banking
system tends to reduce credit dollarization unless banks have high dollar liabilities that
they need to offset. As the number of banks increases, the currency risk spreads among
them perfectly and they behave as if they are risk neutral. Thus, the main determinant of
credit dollarization becomes real dollarization”.

Risk aversion shapes the relative importance of banks’ and ﬁrms’ determinants.
Matching is relatively more important if banks’ risk aversion is larger than ﬁrms’ risk

aversion.

 

23 Catﬁo and Terrones (2000) present a similar argument that higher concentration in the banking

industry leads to lower equilibrium dollarization levels. The higher banks’ market power, the more they
will exercise it in lending. Naturally, banks’ market power is higher for borrowers that do not have access
to direct borrowing from abroad (denoted below as “external borrowing”). These are borrowers in the non-
traded goods sector, or borrowers that do not have traded collateral. Thus, they are more likely to demand
domestic currency credit. Therefore, concentration tends to increase lending in domestic currency and
reduce lending in dollars.

68

3. Empirical Analysis
Following the equilibrium level of credit dollarization derived in the theoretical

model, we estimate the reduced-form equation
Credit dollarization“: a, Bank variables it+ B. Firm variablesit+ y. Controlsit+ uit.

Credit a’ollarization denotes the ratio of foreign currency loans to total bank
loans. Bank variables stands for the set of variables related to the banking sector,
described further in this section and in more details in Table 2-1 in the Appendix, which
identify the supply-side of credit dollarization. Firm variables denotes the set of
variables related to the real sector, which measure the demand-side of credit dollarization.
Controls measures all other factors that affect both demand and supply of dollar credit, or
which capture country-speciﬁc effects. Finally, u is a disturbance term.

We use pooled OLS (POLS) as the benchmark. We then extend the analysis and
use the ﬁrst difference (FD) and ﬁxed effects (FE) estimators to control for missing
country—speciﬁc effects and check the robustness of our results to alternative
speciﬁcations. The results obtained using the random effects (RE) estimator are also
reported. However, we believe that missing country-speciﬁc effects are more likely to be
correlated with our regressors, so we mainly rely on the results obtained using the FD and
FE estimators.

We have constructed a new unbalanced panel dataset for twenty-one transition
economies ﬁ'om Central and Eastern Europe and Central Asia covering the period 1990 -
2001. Data on credit dollarization are annual, and were collected from various central

bank publications. For some countries, data starts in the early 19905. However, for most

69

countries, data covers the second half of the 19908 (see Table 2-2). The other variables
come from various sources, mainly the International Monetary Fund’s International
Financial Statistics, the World Bank’s World Development Indicators, and the European
Bank for Reconstruction and Development Transition Reports (see the Table 2-1).

Credit dollarization is the ratio of foreign currency credit to total credit, from
domestic banks to domestic (non-bank) enterprises. As bank variables, measures of
deposit dollarization and bank net foreign assets are used. Deposit dollarization denotes
the ratio of foreign currency deposits to total deposits held by residents at domestic
banks. Bank net foreign assets denote the ratio of domestic banks’ net foreign assets, to
total deposits held by residents at domestic banks.

As ﬁrm measures we use exports, expressed as the ratio to GDP, correlation of
export prices and the exchange rate, and correlation of domestic prices and the exchange
rate. Based on the theoretical model, we need to control for the exchange rate volatility
and the foreign interest rate. However, since we use measures of the stock (rather than
the ﬂow) of loans, 24 deposits, and bank foreign assets, we also include the rate of
devaluation of the exchange rate to control for the valuation effects. Standard errors in
the POLS, FE, and FD estimations are adjusted for heteroskedasticity and serial
correlation, to further account for any persistence problems.

Table 2-3 reports descriptive statistics of the dependent and main explanatory
variables used in the estimations. The average level of credit dollarization in transition

economies is 38.2 percent, with a minimum of 1.7 percent in the Slovak Republic in

 

2‘ We use stock measures due to data availability. However, we argue that this is a sensible measure,
since bank loans in transition economies have a shorter maturity than in most developed economies. For
example, the average credit period for bank loans (denominated in either domestic or foreign currency) for
a sample of large ﬁrms for the period 1990 - 2001 is 70 days. The median credit period is even lower: 49.5
days.

70

1992, and a maximum of 84.8 percent in Albania in 2001. In general, the former Soviet
Union and the South-Eastern European countries tend to be more dollarized then the
(more developed) Central European countries.

In Table 2-4, a simple data analysis shows positive correlation of deposit and
credit dollarization (0.582, P-value of 0.000). Net foreign assets, on the other hand, do
not seem to be correlated with credit dollarization (-0.042, P-value of 0.607). On the ﬁrm
side, exports seem to be negatively correlated with the level of ﬁnancial dollarization (-
.121, P—value of .131 for credit dollarization, and -.l76, P-value of .020 for deposit
dollarization), which is the opposite of what one would expect. A higher correlation of
domestic prices and exchange rate is associated with more dollar loans, as expected (.198,
P-value of .019), while the correlation of export prices and the exchange rate is not (.099,
P-value of .243).

The estimation results are summarized in Table 2-5.25 Regardless of the
estimation method, bank asset and liability management are the main determinants of
dollarization. Deposit dollarization has a large positive effect, while net foreign assets
have a negative effect on credit dollarization. A one percentage point increase in deposit
dollarization increases credit dollarization by .75 - .83 percentage points with the POLS
and FE estimators, and by .44 - .57 percentage points with the FD and RE estimators. A
one percentage point increase in net foreign assets decreases credit dollarization by .15 -

.23 percentage points. It seems to be the case that domestic banks use dollar deposits to

 

25 In addition to the regressors described before, we use a dummy variable in the POLS estimation, which
takes the value one for Croatia, zero otherwise. According to Bonin (2001), most loans in Croatia are
indexed to the German mark/Euro. These loans are equivalent to foreign currency loans for the purpose of
this paper, but are instead recorded as domestic currency loans. This explains why in Croatia the average
level of deposit dollarization is 68.6 percent, while credit dollarization is only 27.3 percent. Alternatively,
we estimate the model without data on Croatia, and obtain very similar results for all our speciﬁcations.

71

“create” dollar loans, and foreign assets as a substitute for domestic dollar loans.

There is also evidence of a positive effect of exports on credit dollarization.
Initially, with the POLS estimator, exports have a positive, but insigniﬁcant effect. This
result and the initial negative correlation between exports and credit dollarization (see
Table 2-4) suggest that this variable is correlated with unobserved country speciﬁc
effects, and proxies for the overall level of development of the economy. Once we
control for unobserved country effects, exports have a positive and very signiﬁcant effect
on credit dollarization. Thus, our results overturn the counterintuitive ﬁndings of
previous papers (Arteta, 2003, and Barajas and Morales, 2003) that a higher level of trade
and openness of the economy decrease the levels of dollar loans. These studies seem to
suggest that partial dollarization is mainly an outcome of uncertainty. We show that
dollarization is also driven by integration in the international good markets, and we
expect it to persist even after uncertainty has been curbed.

The other ﬁrm-related measures are very insigniﬁcant, but have also very large
standard errors, which suggest possible measurement problems. The exchange rate
volatility and the foreign interest rate have the negative sign suggested by the model for
three out of four estimations, although they are not- signiﬁcant. The rate of devaluation of
the exchange rate is positive and mostly signiﬁcant. It probably captures a valuation
rather than a substitution effect.26

So far, we have identiﬁed the demand and supply side of the market for dollar
loans, and have shown that both ﬁrm and bank variables are determinants of credit

dollarization. Since the ultimate goal of the paper is to say something about the relative

 

2" The level and volatility of the rate of devaluation of the exchange rate are highly correlated (.587, P-
value of .000). However, no results change when one is dropped from the estimations.

72

importance of the two sets of factors, we compare the joint signiﬁcance of ﬁrm and bank
variables. In all cases, the F -statistic (or Chi-square, for the RE estimation) is higher for
the bank factors relative to ﬁrm factors. Currency matching is therefore the main driving

force of dollarization.27

4. Robustness Tests

There are two methodological issues that need to be further considered when
estimating the above equation: the identiﬁcation problem and endogeneity.

In order to be able to correctly identify and separate demand and supply factors,
we need (at least) one demand-speciﬁc measure that does not affect the supply of foreign
currency credit, and (at least) one supply-speciﬁc measure that does not affect the
demand for foreign currency credit. Based on the theoretical model, we use measures of
comovement of export (and domestic) prices and the exchange rate, as well as the ratio of
exports to GDP. The comovement measures are very imprecise, and have no signiﬁcant
effects. Arguably, the exports, which turn out to have a strong positive and signiﬁcant
effect on credit dollarization might just measure an overall openness of the economy, and
capture some bank effect as well. In the next set of estimations, reported in Table 5, we
add a measure of the non-bank external trade credit, which we believe identiﬁes the ﬁrm
sector. On average, trade credit used by ﬁrms in transition economies is 3.23 percent of
GDP, compared to total bank credit which is 19.85 percent of GDP. Firms that have

access to alternative sources of external ﬁnance are expected to rely less on credit from

 

27 One issue to point out is that while some year dummies are also signiﬁcant, the country dummies are
always very signiﬁcant. This might indicate that dollarization levels vary widely across countries. Second,
it might also be the outcome of country initial conditions and characteristics that are not explicitly
measured. And third, it is the cross-section rather than the time dimension of the panel dataset that we
explore, and not surprisingly the country dummies will be more signiﬁcant than the time dummies.

73

domestic banks. Hence, we expect ﬁrms’ trade credit from abroad to have a negative
effect on credit dollarization, since we believe that external loans are a substitute for
foreign currency loans from domestic banks.

We also use additional controls: an index of foreign exchange and trade
liberalization, which should have a positive effect on credit dollarization, and a dummy
variable for the existence/ development of forward exchange market, which presumably
should reduce credit dollarization, as both banks and ﬁrms could use forward exchange
contracts to hedge against exchange rate risk.28 The ratio of net foreign reserves (held by
the central bank) to GDP is used, together with the volatility of the devaluation rate, to
proxy for the expectations about the future exchange rate. A higher level of reserves will
signal higher sustainability of domestic currency/ less uncertainty with respect to the
exchange rate (see also Ize and Parrado (2002)).

The new set of results is presented in Table 2-6. Trade credit has a negative and
signiﬁcant effect (with the FE, FD, and RE estimators) on credit dollarization, as
expected, suggesting that ﬁrms use external trade credit as a substitute for dollar credit
from domestic banks. This is consistent with Barajas and Morales (2003) who show that
external credit is a substitute for the domestic foreign currency credit in Latin American
economies.

The effects of other ﬁrm and bank variables do not change: more exports still lead
to more loans denominated in dollars, while deposit dollarization still increases and net

foreign assets reduce credit dollarization. Note that bank variables have a higher joint

 

2“ Burnside et al. (2000) argue that the forward market can be used to alleviate the bank currency
mismatches, and banks can borrow in dollars, and lend in domestic currency. A similar argument applies
to exporting ﬁrms: they can borrow in domestic currencies, and hedge against the currency risk using
derivatives.

74

signiﬁcance than the ﬁrm variables in all four estimations.

There is also a positive effect coming from the rate of devaluation, and a negative
effect coming from its volatility. Net foreign reserves signiﬁcantly reduce dollar loans.
A one percentage point increase in net foreign reserves to GDP reduces credit
dollarization by 0.8 —- 1.1 percentage points with the POLS, FE, and RE estimations, and
by .5 percentage points with the FD estimator. This last result is surprising since we
expect uncertainty to decrease credit dollarization. Potential explanations link the use of
ﬁnancial instruments denominated in foreign currencies to costs associated with
uncertainty and lack of credibility in domestic currency/ domestic macroeconomic
policies. Jeanne (2001) argues that the uncertainty with respect to domestic monetary
policy increases the cost of borrowing/lending in domestic currency (the so called “peso
effect”), and therefore leads to more borrowing/lending in dollars. Similarly, Delgado et
al. (2002) argue that uncertainty about potentially high future depreciation associated
with a high real interest rate for domestic currency lending makes foreign currency
lending more attractive, regardless of the risks involved.

Overall, higher liberalization of trade and foreign exchange operations leads to
more dollar loans, but this effect is mostly insigniﬁcant. There does not seem to be any
signiﬁcant effect from the forward market indicator either.29

A second issue is the problem of endogeneity. Mainly, the measures of asset and

liability management for banks and liability management for ﬁrms might be endogenous.

 

29 Transition economies countries are a good choice for studying determinants of dollarization beyond
regulatory measures, since there does not seem to be too much variation in regulations: average level of the
index of liberalization of trade and foreign exchange operations is 3.33 (with a range from 1, highly
regulated, to 4.3, fully liberalized). Furthermore, the foreign exchange market is underdeveloped (if extant
at all) in most countries: the average is .28 for the forward market dummy (a value of 1 indicating that the
market exists and it is functional).

75

In the theoretical model we assume that deposit dollarization and bank net foreign assets
are exogenously determined. This might not be the case, if ﬁrms maintain correspondent
deposits with the banks that grant the loans. In addition, since we use low frequency
data, the newly issued loans might temporarily increase the volume of ﬁrms’ deposits.
To verify that our results are not plagued by these problems, instead of the total deposit
dollarization measure that includes both households and ﬁrms, we use only household
deposit dollarization. Arguably, household deposit dollarization can be assumed to be
supply-determined, and thus exogenous in this analysis.

Based on the simpliﬁed theoretical model, it is the net measure of foreign assets
rather than its components, foreign assets and foreign liabilities, which matters.30 But it
can be argued that while foreign assets are more likely endogenous, foreign liabilities of
domestic banks can be treated as exogenous, as they are mainly supply-determined.
Separating net foreign assets into foreign assets and foreign liabilities will shed some
light on these effects.

We believe that the estimated trade credit effect is not biased by endogeneity
problems. Underdeveloped ﬁnancial markets and few ﬁnancing sources prompt ﬁrms to
extensively employ trade credit as a cheaper alternative to bank credit. However, trade
credit is constrained only to cross-border transactions and we can assume that it is
predetermined in the loan demand equation.

For exports and the comovements, we use the same measures. If anything, the
coefﬁcient on exports might be biased upward, if ﬁrms depend on external ﬁnance (dollar

loans) to ﬁnance exports.

 

3° Since the external interest rate is the same for assets and liabilities, only the net position matters.

76

These new results with the alternative bank measures are summarized in Tables 2-
7 and 2-8. Notice ﬁrst that the sample size is smaller, and that some countries (Georgia,
the Kyrgyz Republic, Lithuania, and Macedonia) drop from the analysis. Nevertheless,
bank variables are still signiﬁcant with the FE, FD, and RE estimators: more household
dollar deposits and foreign liabilities, and less dollar assets, lead to more dollar loans.
Moreover, the effect of foreign liabilities is relatively larger than that of foreign assets.31
Overall, bank variables are jointly signiﬁcant in seven out of the eight estimations.

Firm measures, on the other hand, are only marginally signiﬁcant. Exports are
signiﬁcant in one estimation only, while trade credit, although negative, is signiﬁcant
only in the FE estimation. The co-movement measure of export prices and exchange rate
has now a negative but not always signiﬁcant effect. This is a (country invariant)
measure of primary commodity export prices (aggregated over all developing
economies). Since transition economies are not major primary commodity exporters
(except for the three big oil exporters), we think this variable may capture instead some
import price effect. This would justify the negative effect. Overall, ﬁrm variables are
jointly signiﬁcant in only two out of the eight estimations.

Similar results are obtained if one-period lagged ﬁrm and bank variables are used
instead of their contemporaneous values.

Other sensitivity tests performed involve further exploration of the uncertainty
measures.32 Results obtained so far seem to suggest that higher uncertainty of the future

exchange rate leads to more credit dollarization. However, when the net foreign reserve

 

3' We compute the standardized coefﬁcients for the FE and FD estimations. A one percent standard
deviation increase in foreign assets reduces credit dollarization by .193 (.127) percent standard deviation in
the FE (FD) estimation. A one percent standard deviation increase in foreign liabilities raises dollarization
by .479 (.402) percent standard deviation in the FE (FD) estimation.

’2 These results are not presented in the paper, but are available upon request from the authors.

77

and rate of devaluation measures are combined into one “fear of ﬂoating” index a la
Hausman et al. (2003) and Calvo and Reinhart (2003), the effect becomes insigniﬁcant.
Thus, while the level of reserves does seem to matter a lot, the volatility measure, taken
separately or combined with the exchange rate volatility, is insigniﬁcant.

The majority economists’ view (see, for example, Burnside et al. (2000), Mishkin
(1996), and Obstfeld (1998)) is that a ﬁxed exchange regime offers implicit guarantees
(implicit exchange rate insurance) and eliminates the need to hedge against the currency
risk, provided that it is credible. This implies that over time a more ﬂexible exchange
rate should reduce credit dollarization. It also suggests that across countries the levels of
dollarization should be higher in countries with ﬁxed exchange rate regimes, when
controlling for any credibility issues. The empirical evidence is mixed: Martinez and
Werner (2002) ﬁnd support for this hypothesis; Arteta (2003), on the other hand, shows
that credit dollarization does not vary across exchange rate regimes, and higher ﬂexibility
exacerbates rather than ameliorates banking mismatches. When we use dummies for the
“de facto” exchange rate regime (following Reinhart and Rogoff, 2004), we see a mixed
picture. Limited ﬂexibility exchange rate regimes will be less dollarized than the pegged
ones, but the freely ﬂoating exchange rate regimes experience more dollarization.33

Summing up, all the evidence presented suggests the same conclusion: currency
matching by banks of their assets and liabilities is the main driving force of credit
dollarization. We have also identiﬁed two effects coming from the ﬁrm side: (i) ﬁrms
hedge against exchange rate risk by denominating costs and revenues in the same

currency, and (ii) ﬁrms use dollar loans as an alternative to external trade credit.

 

3’ However, since only Albania has a freely ﬂoating exchange rate regime according to the classiﬁcation
we use, and it is also a highly dollarized economy, we are cautious about the implications of this result.

78

5. Conclusions and Future Extensions

Domestic banks in emerging markets and transition economies oﬁen grant to
residents dollar instead of domestic currency credit. Is this credit dollarization
phenomenon coming mostly from the ﬁnancial side (banks), or from the real side (ﬁrms)?
In the presence of ﬁnancial dollarization without real dollarization, there exist balance
sheet mismatches somewhere in the economy and the economy is highly exposed to
ﬁnancial and currency crises.

This paper analyzes the respective contributions of banks and ﬁrms to credit
dollarization. We group potential determinants into bank- and ﬁrm-speciﬁc (ﬁnancial
and mainly real) factors, and use a newly-assembled database for twenty-two transition
economies. We show that bank currency matching is the main driving force of credit
dollarization. We ﬁnd some evidence that real dollarization also causes ﬁnancial
dollarization, but this effect is relatively small. There are currency mismatches in these
economies, and they seem to be concentrated in the real sector. Banks, who are supposed
to do a better job at internalizing the exchange rate risk, just pass it over to ﬁrms. This
decreases bank exposure to currency risk, but it increases their exposure to default risk,
and ultimately the exposure of the economy to ﬁnancial crises.

Not surprisingly, our ﬁndings suggest that one way to reduce dollarization is to
reduce macroeconomic uncertainty and the lack of monetary policy credibility.
Otherwise, as long as domestic deposits are highly dollarized, and banks want to match
the currencies of their assets and liabilities, restrictions placed on dollar loans will most

likely lead to “exports” of deposits and overall reduction in domestic credit.

79

This paper is one of the ﬁrst attempts to empirically study credit dollarization.
There are several possible extensions that deserve further consideration. First, we would
like to extend the analysis to include other developing economies, especially Latin
American countries where dollarization reaches very high levels. That would provide not
only more data, but also suggest the generality of the conclusions. Transition economies
are a good starting point as they provide a lot of variation in data; this can also be a
downside, since it is sometimes hard to separate one effect from the others.

Second, using disaggregated data, such as individual loan data, would provide
better insights into the effect of ﬁrm and bank characteristics on ﬁnancial dollarization.

Third, more empirical work in general is necessary to better identify both the
determinants and the effects of ﬁnancial dollarization, and to evaluate policy implications

and the exchange rate exposure of various economic agents and the overall economy.

80

REFERENCES

Aghion, P., P. Bacchetta, and A. Banerjee. 2001. “Currency Crises and Monetary Policy
in a Credit-Constrained Economy.” European Economic Review, 45, pp. 1121-
1150.

Aguiar, M. 2002. “Devaluation, Foreign Currency Exposure, and Investment: The Case
of Mexico.” Mimeo. University of Chicago.

Allen, M., C. Rosenberg, C. Keller, B. Setser, and N. Roubini. 2002. “A Balance Sheet
Approach to Financial Crisis.” International Monetary Fund Working Paper
02/210.

Arteta, C. O. 2002. “Exchange Rate Regimes and Financial Dollarization: Does
Flexibility Reduce Bank Currency Mismatches?” Board of Governors of the
Federal Reserve System, International Finance Discussion Paper 738.

Barajas, A. and A. Morales. 2003. “Dollarization of Liabilities: Beyond the Usual
Suspects.” International Monetary Fund Working Paper 03/ 1 1.

Bonin, J. 2001. “Financial Intermediation in Southeast Europe: Banking on the Balkans.”
Mimeo. Vienna Institute for International Economic Studies.

Broda C. and E. Levy Yeyati. 2003. “Endogenous Deposit Dollarization.” Federal
Reserve Bank of New York Staff Report 160. .

Bubula, A. and I. Otker-Robe. 2002. “The Evolution of Exchange Rate Regimes since
1990: Evidence from De Facto Policies.” International Monetary Fund Working
Paper 02/155.

Burnside, C., M. Eichenbaum, and S. Rebelo. 2000. “Hedging and Financial Fragility in
Fixed Exchange Rate Regimes.” Mimeo. Northwestern University and National
Bureau of Economic Research.

Caballero, R. and A. Krishnamurthy. 2002. “Excessive Dollar Debt: Financial
Development and Underinsurance.” Journal of Finance, 58:2, pp. 867-894.

Calvo, G. A. 2001. “Capital Markets and the Exchange Rate: With Special Reference to
the Dollarization Debate in Latin America.” Journal of Money, Credit, and
Banking, 33:2, pp. 312-334.

Calvo, G. A. 2002. “On Dollarization.” Economics of Transition, 10:2, pp. 393-403.

Catao, L. and M. Terrones. 2000. “Determinants of Dollarization: The Banking Side.”
International Monetary Fund Working Paper 00/ 146.

81

Cespedes, L., R. Chang, A. Velasco. 2000. “Balance Sheets and Exchange Rate Policy.”
National Bureau of Economic Research Working Paper 7840.

Calvo, G. and C. Reinhart. 2002. “Fear Floating.” The Quarterly Journal of Economics,
117:2, pp. 379-408.

Choi, G. and D. Cook. 2002. “Liability Dollarization and the Bank Balance Sheet
Channel.” International Monetary Fund Working Paper 02/141.

Cotner, J. 1991. “Currency Risk in Long-term Borrowing.” Journal of Multinational
Financial Management, 1:2, pp. 49-64.

Delgado, F. L., D. S. Kanda, G. M. Casselle, and R. A. Morales. 2002. “Domestic
Lending in Foreign Currency.” In Building Strong Banks through Surveillance
and Resolution. C. Enoch, D. Marston, and M. Taylor eds. International Monetary
Fund: Washington DC, pp. 40-61.

Demirgﬁc-Kunt, A. and H. Huizinga. 2003. “ Market Discipline and Deposit Insurance.”
Journal of Monetary Economics, forthcoming.

EBRD. 1998-2002. “Transition Report.” European Bank for Reconstruction and
Development: London.

Gertler, M., S. Gilchrist, and F. Natalucci. 2003. “ External constraints on monetary
policy and the ﬁnancial accelerator.” BIS Working Papers No. 139

Giraldi, C. and R. Hamaui. 1991. “Foreign Exchange Market Liberalization and the
International Diversiﬁcation of Borrowing by Italian Firms.” Journal of
Economics and Business, 43:4, pp. 309-327.

Hausmann, R., U. Panizza, and E. Stein. 2001. “Why Do Countries Float the Way They
Float?” Journal of Development Economics, 66:2, pp. 387-414.

Honohan, P. and A. Shi. 2002. “Deposit Dollarization and the Financial Sector in
Emerging Economies.” World Bank Working Paper 2748.

IMF. 1990-2002. “Annual Report on Exchange Arrangements and Exchange
Restrictions.” International Monetary Fund: Washington, DC.

Ize, A. and E. Levy Yeyati. 2003. “Financial Dollarization.” Journal of International
Economics, 59:2, pp. 323-347.

Ize, A. and E. Parrado. 2002. “Dollarization, Monetary Policy, and the Pass-through.”
International Monetary Fund Working Paper 02/188.

Jeanne, O. 1999. “Foreign Currency Debt, Moral Hazard, and the Global Financial
Architecture.” Mimeo. International Monetary Fund: Washington, DC.

82

Jeanne, O. 2001. “Why Do Emerging Markets Borrow in Foreign Currency?” Mimeo.
International Monetary Fund: Washington, DC.

Kawai, M. “The Behavior of an Open-Economy Firm under Flexible Exchange Rates.”
Economica, 48: 189, pp. 45-60.

Kedia, S. and A. Mozumdar. 2002. "Foreign Currency Denominated Debt: An Empirical
Examination." The Journal of Business: forthcoming.

Keloharju, M. and M. Niskanen. 2001. "Why Do Firms Raise Foreign Currency
Denominated Debt? Evidence from Finland." European Financial Management,
7:4, pp. 481-496.

Krugman, P. 1999. "Balance Sheet, the Transfer Problem, and Financial Crises."
International Tax and Public Finance, 6:4, pp. 459-472.

Martinez, L. and A. Werner. 2002. "The Exchange Rate Regime and the Currency
Composition of Foreign Debt: The Mexican Experience." Journal of Development
Economics, 69:2, pp. 315-334.

Levy-Yeyati, E. 2003. "Financial Dedollarization: A Carrot and Stick Approach." CIF
Working Paper 8/2003.

Mishkin, F. 1996. "Understanding Financial Crises: A Developing Country Perspective."
in Annual World Bank Conference on Development Economics 1996. M. Bruno
and B. Pleskovich eds. Washington DC, pp. 29-62.

Obstfeld, M. 1998. "The Global Capital Market: Benefactor or Menace?" Journal of
Economic Perspectives, 12:4, pp. 9-30.

Reinhart, C. M. and K. S. Rogoff. 2002. "The Modern History of Exchange Rate
Arrangements: A Reinterpretation." National Bureau of Economic Research
Working Paper 8963.

83

APPENDIX

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2-1 Data Deﬁnitions and Sources
VARIABLE DEFINITION SOURCE
Credit dollarization Foreign currency denominated credit to total CB
credit, from domestic banks to domestic
enterprises, calculated as a percent. Data for a
few countries include credit to individuals,
ﬁnancial enterprises and/or government.
Deposit Foreign currency denominated deposits to total CB
dollarization deposits held by residents at domestic banks,
calculated as a percent.
Household deposit Household foreign currency denominated deposits CB
dollarization to total household deposits with the domestic
banks, calculated as a percent.
Bank net foreign Bank foreign assets minus bank foreign liabilities IFS, CB
assets (line 21...ZF... - line 26C...ZF), as a percent of
total bank deposits.
Exports Total exports of goods and services as a percent WDI
of GDP.
Index of export The index of primary commodity export prices for IMF_CP
prices all developing countries (1995=100).
Correlation of Annual correlation of monthly changes in the IMF_CP,
export prices and index of primary commodity export prices and the IFS
exchange rate monthly rate of devaluation of the end-of-period
dollar exchange rate
Correlation of Annual correlation of monthly domestic inﬂation IFS
domestic prices and and the monthly rate of devaluation of the end-of-
exchange rate eriod dollar exchange rate
Exchange rate Annual standard deviation of the monthly rate of IFS
volatility devaluation of the exchargg rate
Foreign interest Libor interest rate on the three-month USD Economagic
rate denominated dgrosits
Rate of devaluation Percentage change of the annual end-of-period IFS
of the exchange exchange rate (expressed as domestic currency
rate per USD, line RF..ZF. . .).
Net foreign Net reserves of the central bank as a percent of IFS
reserves GDP (line 1L..DZF ...*100/GDP).

 

84

 

Table 2-1 (Continued)

 

VARIABLE

DEFINITION

SOURCE

 

Foreign exchange
and trade
liberalization

Index of trade and foreign exchange rate system
liberalization, =1 if widespread import and/or
export controls or very limited legitimate access
to foreign exchange, =2 if some liberalization of
import and/or export controls, almost full current
account convertibility in principle but with a
foreign exchange regime that is not fully
transparent (possibly with multiple exchange
rates), =3 if removal of almost all quantitative and
administrative import and export restrictions,
almost full current account convertibility, =4 if
removal of all quantitative and administrative
import and export restrictions (apart form
agriculture) and all signiﬁcant export tariffs,
insigniﬁcant direct involvement in exports and
imports by ministries and state-owned trading
companies, no major non-uniformity of customs
duties for non-agricultural goods and services, full
current account convertibility, =4.3 if standards
and performance norms of advanced industrial
economies, removal of most tariff barriers, WTO
membership.

EBRD

 

Forward market
liberalization

Dummy variable=l if a forward market for
foreign exchange exists and is functional, =0 if it
does not exist or is underdeveloped.

AREAER

 

Peg

Limited ﬂexibility
Managed ﬂoat
Free ﬂoat

Free fall

 

 

Dummy variables differentiating between pegged,
limited ﬂexibility, managed ﬂoat, freely ﬂoating,
and freely falling exchange rate regimes,
according to Reinhart and Rogoff (2004)
classiﬁcation. These are all measures of the “de
facto” regimes, classiﬁes using information on
parallel exchange markets.

 

Reinhart
and Rogoff
(2004)

 

Abbreviations: AREAER: IMF Annual Report an Exchange Arrangements and
Exchange Restrictions (various issues). CB: Central Banks publications (various issues).
IFS: International Monetary Fund International Financial Statistics.

Commodity Prices.

WDI: World Bank World Development Indicators.

IMF_CP: IMF

EBRD:

European Bank for Reconstruction and Development Transition Reports (various issues).
BIS: Bank of International Settlements International Banking Statistics.

85

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2-2 Country and Dollarization Data Coverage
Country Data Availability
Albania I998 — 2001
Armenia 2000 — 2001
Azerbaijan 2000 — 2001
Bulgaria 1990 — 2001
Croatia 1993 - 2001
Czech Republic 1993 — 2001
Estonia 1993 - 2001
Georgia 1997 - 2001
Hungary 1992 — 2001
Kazakhstan 1996 — 2001
Kyrgyz Republic 1999 — 2001
Latvia 1994 - 2001
Lithuania 1993 — 2001
Macedonia 1995 — 2001
Moldova 1999 — 2001
Poland 1997 — 2001
Romania I990 — 2001
Russia 1996 — 2001
Slovak Republic 1992 — 2001
Slovenia 1991 - 2001
Ukraine 1992 — 2001

 

86

 

 

 

 

 

 

 

 

 

 

 

 

 

 

mg cﬂ ES ovm .3: AZ
30.3 _ _ GE . _ a 555::
So. T 2w: mNNH ...—:55:
Rad mum. com. EYE Son ....m
2%.— _w_ . o _ ct Kwdm 5:52
mmmd _w_. wmo. mm_._v 53:
5.0.3 8.: ...vo can 3.: ...88 can 3.35m max—mas»
0.3.5. 83.:— .Ee_. ...—:0 82.:— dxu ...—:0

$— 52 mm _ NE of .3: 62
museum 21am: vwméa www.mm 2.54% 5.5532
c new.» o mama «ewe.— 55552

$9.? 833 wamém va.w_ amvdm Son .3
Kmdm momdm omoém 8qu www.mm 5:82
30.2” S #3 3.3m $0.0m NS .3 :82

82.53. $.28 33.8 ewes. 5:32.52. 5355.8 eeeﬁtéee

....am 5.5.: .9... 2238: 52.5 5.20 mam—55c!
832% inseam 2 2.5.

87

ﬁne—.22.:— =_ muss»-.— .Ea Set—3.53.. ..e LEE—.2

 

 
 
 

3:. .3: 88. .8: $2. .Nw: 8mm. .3: $3.58 9:..8: in .S: as. .m: 2.:
e8; an woo. woo. m3- 42.- 8o. So. 532.98
A8”. .8: Ga. .8: 8mm. .8: :8. .3: 82:3: am”. .ﬁ: 3%. on: 32...:
98.- 8a.- 89. at; So. .8; c3- ears...
as. .8: 88:8: 83. .8: 88:6: 32. .3: so». .9: bars?
at. :9- 3:. :2..- 25.- «8.- a: ...aﬁ
:8. .8: S3. .8: 38. .em: ASN.3: as. it: 3.: .55
mm _. so; m _ NF woo. me. E; 82....
.Ec—u ..EOU
:8. .8: 33%: GB. .3: 5N5: 3.: ...aa
42.- EC. e8. 8o. E: 83:.
.93 FED
£2. .5: 88. .ﬁ: :2. .8: are:
a..- e:.. :7
5m. .3: Eve. .4“: 33.8 ewes.
8c. N5- .2. ion
88. .3: 5.35:3.
8m. :8an
3a.- ..E awn—...; 3a.— A—uxo 3a.. .auxu 3L0 x”.— muOmm-w EOENNTSE—v Homaanta=cv
away—em 8a.. .55.. was 83.5 ...—a moot.— =w_o..e._ amen—on «:55
.Eo—v 5.50 .93 LLOU no: v—uam—

ﬁbaz uouﬂoteu TN 03:.

 

88

 

 

 

 

 

 

 

 

 

 

Table 2-5 Basic Model Estimation Results
POLS FE FD RE
Deposit dollarization .839*** .753*** .440*** 570*"
(150) (.268) (.147) (.105)
Bank net foreign assets -.218*** -.228*** -.149*** -.150***
(.065) (.072) (.030) (.035)
Exports .027 .800*** .335" .206*
(.259) (.294) (.143 (.1 12)
Correlation of export prices and -.257 .654 1.279 .728
exchange rate (6.183) (4.698) (2.078) (2.820)
Correlation of domestic prices 2.377 -2.213 .189 .932
and exchange rate (3.065) (3.881) (2.031) (1.924L
Volatility of depreciation -.209 -. 142 .024 -.009
(.193) (.143) (.087) (.077)
Foreign interest rate -2.197 -1.162 1.681M -3.770*
(3.346) (2.099) (.757) (2.264)
Rate of depreciation of the .016 .018 .018* .015"
exchange rate (.017) (.012) (.009) (.006)
Obs. 141 141 120 141
No. countries 21 21 21 21
R-squared .591 .777 .418 .383
Joint signiﬁcance of bank 15.95 6.41 17.27 45.00
variables (.000) (.002) (.000) (.000)
Joint signiﬁcance of ﬁrm .23 2.69 3.50 4.30
variables (.875) (.050) (.034) (.231)

 

Robust standard errors in parentheses (adjusted for heteroskedasticity and serial
correlation) for POLS, FE, and FD. Adjusted R-squared for POLS and FE, overall R-

squared for RE.
For the joint signiﬁcance tests, p-values are reported in parentheses.
Also included in the regressions are year dummies and a constant term.
*** Signiﬁcant at 1%
** Signiﬁcant at 5%
* Signiﬁcant at 10%

89

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2-6 Extended Regressions
POLS FE FD RE
Deposit dollarization .762*** .693*** .479*** 579*"
(.206) (.185) (.158) (.120)
Bank net foreign assets -.211*** -.180*** -.118*** -.141***
(.044) (.042) (036) (.038)
Exports .082 .576" .371 * .242“
(.255) (.279) (.187) (1 15)
Correlation of export prices and -8.533 -.402 1.521 .969
exchange rate (7.922) (3.848) 43.337) (3.258)
Correlation of domestic prices 1.334 .346 .365 1.537
and exchange rate (4.295) (3.316) (2.428) (2.275)
Trade credit -.245 -2.499** -2.474* -l.621**
(.697) (1.1 13) (1.366) (.650)
Volatility of depreciation -.279* -.213* -.055 -.O69
(.156) (.113) (.051) (.086)
Foreign interest rate -.352 -.069 1.101*** -.018
(.61 1) (.400) (.352) (.450)
Rate of depreciation of the .027* .023* .013 .015“
exchage rate (.015) (012) (.009) (.007)
Net foreign reserves -1.119** -1.134*** -.519"‘ -.829***
(.429) (.404) (.288) (.259)
Foreign exchange and trade 7.257 * 2.377 -.723 -.083
liberalization (3.630) (2.982) (1.597) (2.547)
Forward market liberalization 1.838 -8.848 -3.261
(5.788) (9.705) (5.582)
Obs. 119 119 84 119
No. countries 21 21 21 21
R-squared .657 .872 .461 .414
Joint signiﬁcance of bank 13.32 11.95 14.38 36.21
variables (.000) (.000) (.000) (.000)
Joint signiﬁcance of ﬁrm .53 1.85 3.61 11.48
variables (.718) (.128) (.023) (.022)

 

Robust standard errors in parentheses (adjusted for heteroskedasticity and serial
Adjusted R-squared for POLS and FE, overall R-

correlation) for POLS, FE, and FD.

squared for RE.

For the joint signiﬁcance tests, p-values are reported in parentheses.
Also included in the regressions are year dummies and a constant term.

*** Signiﬁcant at 1%
** Signiﬁcant at 5%
* Signiﬁcant at 10%

9O

 

 

 

 

 

 

 

 

 

 

 

Table 2-7 Basic Model with Household Deposit Dollarization
POLS FE FD RE
Household deposit dollarization .267 .459*** .278*** 275*"
(.166) (.163) (.091) (.101)
Bank foreign assets -.025 -.251 -.100* -.083
(.173) (.161) (.047) (.059)
Bank foreign liabilities .253 .387* .269" .251***
(.188) (.222) (.106) (.069)
Exports -.188 .666* .277 .090
(.305) (.339) (.164) L136)
Correlation of export prices and -2.240 -5.957 1.178 .107
exchange—rate (8.150) (6.863) (2.446) (3 .266)
Correlation of domestic prices -1.536 -5.976 -2.114 -1.788
and exchange rate (3.369) (4.820) (2.086) (2.056)
Volatility of depreciation .455* .607 .270 .294*
(.222) (.379) (.256) Q 59)
Foreign interest rate -9.498*** -7.031*** 3375*" -5.396*"‘
(3.267) (2.621) @005) (2.294)
Rate of depreciation of the .048" .012 .017 .022"
exchalge rate (.021) (.025) (.017) (.011)
Obs. 117 117 100 117
No. countries 17 17 17 17
R-squared .530 .768 .470 .325
Joint signiﬁcance of bank 6.23 2.74 5.35 21.53
variables (.005) (.049) (.010) (.000)
Joint signiﬁcance of ﬁrm .19 2.19 1.26 1.01
variables (.900) (.095) (.320) (.799)

 

Robust standard errors in parentheses (adjusted for heteroskedasticity and serial
correlation) for POLS, FE, and FD. Adjusted R-squared for POLS and FE, overall R-
squared for RE.
For the joint signiﬁcance tests, p-values are reported in parentheses.
Also included in the regressions are year dummies and a constant term.
*** Signiﬁcant at 1%

** Signiﬁcant at 5%

* Signiﬁcant at 10%

91

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2-8 Extended Regressions with Household Deposit Dollarization
POLS FE FD RE
Household deposit dollarization .221 .263" .204 .182
(.196) (.124) (.213) (.119)
Bank foreign assets -.027 -.196* -.058 -.048
(.145) (.112) (.044) (.069)
Bank foreign liabilities .209 .146 .143* .165"
(.165) (.095) (.081) (.083)
Exports -.116 .289 .370 .107
(.327) (.283) (.330) (.149)
Correlation of export prices and -16.676** -10.291* -1.l79 -3.749
exchange rate (6.851) (5.556) (3.624) (3.885)
Correlation of domestic prices -3.827 -1.436 -.625 -.907
and exchagge rate (3.880) (3.320) (2.122) (2.548)
Trade credit -.765 -3.057* -2.130 -1.687**
(889) (1.605) (1.658) (.792)
Volatility of depreciation .293 .283 .356 .368
(.389) (.362) (.324) (.280)
Foreign interest rate -6.922** -.792 .297 -.825
(2.567) (.522) (.423) (.516)
Rate of depreciation of the .052*** .050*** .024 .027
exchange rate (.017) (.015) (.020) (.017)
Net foreign reserves -1.057 -.767* -.186 -.497*
(.654) (.443) (.347) (.296)
Foreign exchange and trade .318 4.132 -.840 -1.625.
liberalization (6.671) (4.596) (1.955) (3.726)
Forward market liberalization -2.441 -16.087* -7.353
(8.192) (9.543) (6.065)
Obs. 98 98 69 98
No. countries 17 17 17 17
R-squared .612 .865 .512 .367
Joint signiﬁcance of bank 4.23 2.05 5.84 6.99
variables (.022) (.1 17) (.007) (.072)
Joint signiﬁcance of ﬁrm 2.26 2.50 .66 6.05
variables (.108) (.052) (.625) (.195)

 

Robust standard errors in parentheses (adjusted for heteroskedasticity and serial
correlation) for POLS, FE, and FD. Adjusted R-squared for POLS and FE, overall R-
squared for RE.

For the joint signiﬁcance tests, p-values are reported in parentheses.

Also included in the regressions are year dummies and a constant term.

*** Signiﬁcant at 1%, ** Signiﬁcant at 5%, * Signiﬁcant at 10%

92

PART THREE. CURRENCY MISMATCHES AND BANKING CRISES

1. Introduction

Many economies in Eastern Europe and Central Asia experienced banking
instability in the 1990534. Prolonged banking crises extended for most of the decade and
in many cases they were result of the inability of the central bank to maintain credibility
in the domestic currency. As incidents of free falling exchange rate peaked and were
followed by a peak in the occurrence of banking crisis, it is likely that large exchange rate
depreciation may have resulted in banking crises in transition economies, or made them
more severe. Given that these countries had underdeveloped ﬁnancial markets and had
not yet established strict banking regulations for monitoring exchange rate risk, balance
sheet currency mismatches35 were a possible channel for transmission of exchange rate
shocks to the banking system.

This paper explores the link between currency mismatches and banking crises in
the countries of Eastern Europe and Central Asia. Balance sheet currency mismatches
have been blamed for ﬁnancial crises in Asian and Latin American countries during the
19903. Mishkin (1997) suggested that differences in the currency structure of ﬁrms’ and
banks’ revenues and debt reinforced the Mexican banking crisis. Krugman (1999) stated
that currency mismatches in ﬁrms’ balance sheets deepened the Asian banking crises
with the onset of currency crises. Considering the experience of Asian and Latin
American countries, it is likely that currency mismatches have made exchange rate

instability more harmful for the banking systems in transition, as well.

 

3‘ See Appendix 1 and 2.
35 A currency mismatch exists when assets and liabilities have different currency structures.

93

Central bank weakness and lack of credibility in the early 19905 forced many
savers to prefer foreign over domestic currency deposits. Domestic capital was scarce
and foreign capital entered the countries as bank deposits in foreign currency. With high
deposit dollarization36 and limited options to diversify their portfolios, banks granted a
larger proportion of their loans in foreign currency. Reduction in trade barriers and the
prospects of selling in the world markets also fostered the increase in foreign currency
loans.

However, ﬁnancial markets in transition economies lacked a scope of services and
instruments, and economic agents had limited ability to insure themselves against
different types of risks. Thus, unhedged open currency positions were common for both
banks and ﬁrms. Even when restrictions on open currency positions were introduced to
limit banks’ direct exchange rate risk, they brought about an additional source of loan
default when banks transferred the unhedged exchange rate risk to their borrowers.

With no markets to spread the risk, banks and ﬁrms were likely to exhibit risk
aversion and only mismatch to the extent that they have covered the expected exchange
rate risk. Thus, if ﬁrms had stronger risk aversion than banks, the level of credit
dollarization would resemble ﬁrm’s revenue pattern and vise versa. Luca and Petrova
(2004) ﬁnd that the two most important factors of credit dollarization in transition
economies were banks’ foreign currency deposits and net foreign assets. There is very
weak evidence that trade might have had an impact on credit dollarization. These facts

suggest that banks tended to be more risk averse or under stricter obligation to match

 

3° Financial dollarization occurs when a signiﬁcant share of residents’ domestic ﬁnancial contracts
(including deposits and loans) are denominated in foreign currencies, not necessarily the US dollar. Dollar
deposits and loans denote all foreign currency deposits and loans. Deposit dollarization and credit
dollarization refer to the phenomenon of denominating deposits and loans in foreign currencies.

94

their assets and liabilities. In any case, they had stronger incentives to avoid direct
exchange rate risk. However, the transfer of this risk did not eliminate it, it transformed
it into default risk. Since most borrowers were state-owned enterprises, banks may have
been further induced to transform risk because of governments’ implicit guarantees on
claims to these enterprises, while direct exchange rate losses were much less tolerated.
Thus, banks were willing to take on higher default risk while avoiding exchange rate risk
altogether.

This paper builds on the results from Luca and Petrova (2004) that identify the
sector bearing the concentration of exchange rate risk. It explores whether the
concentration of exchange rate risk through mismatched balance sheets has increased the
probability of banking crises and whether the transfer of risk away from the banking
sector has reduced the threat to the banks. If banks in transition economies have avoided
direct exchange rate risk, were they ultimately able to prevent the loss of net worth due to
exchange rate ﬂuctuations?

The paper models the case of small open economies which are unable to borrow
externally in their own currency and have poor domestic currency credibility. Depositors
determine the currency composition of banks’ deposits and ﬁrms receive revenues in
foreign currency that depend on world prices and external demand. The equilibrium level
of credit dollarization depends on ﬁrms’ and banks’ expected revenues in foreign
currency and their ability to absorb exchange rate risk. Thus, in a partial equilibrium
analysis, the paper combines two strands of the currency (mis)matching literature. The
ﬁrst focuses on ﬁrms’ incentives to reduce exposure to exchange rate risk (Kawai (1981),

Giraldi and Hamaui (1991), Cotner (1991)). Matching ﬁrms’ revenues and debt currency

95

structures occurs if ﬁrms have no access to hedging opportunities through the forward
market (Kawai (1981)). The second strand of the literature describes banks’ currency
matching incentives. Banks match their deposit and loan currency structures as long as
they are not protected from exchange rate risk in some way, e.g. a ﬁxed exchange rate
(Ize and Levy-Yeyati (2003)).

If banks’ risk aversion dominates, banks avoid exchange rate risk. However, by
doing so, banks may create a currency mismatch in ﬁrms’ balance sheets and face higher
default risk on their portfolios. On the other hand, if ﬁrms’ matching dominates, banks
accommodate the currency structure of borrowers’ revenues and debt payments and
absorb the increased exchange rate risk”. A Diamond-Dybvig (1983) - style model
describes the link between banks’ and ﬁrms’ optimal currency mismatches to poor bank
net worth and bank runs.

A common approach used in the theoretical literature to connect currency
mismatches to banking crises is to assume that ﬁrms’ and banks’ hedging incentives are
distorted by regulations or macroeconomic policy. Mismatches (and ﬁnancial crises,
respectively) occur because government intervention (guarantees to foreign investors)
fosters overborrowing in foreign currency by the non-tradable sector (Burnside et. al.
(2000), Schneider and Tomell (2000), MacKinnon and Pill (1998)). Chang and Velasco
(1998) show that banking crises (more speciﬁcally, panics) are more likely in a country

with limited foreign currency reserves if the exchange rate regime is ﬁxed rather than

 

’7 Levy-Yeyati (2003) and De Nicolo et al. (2003) discuss some additional balance sheet effects related to
currency matching in the banking sector. Even if banks match the currency structure of their assets and
liabilities, there is a risk inﬂicted by foreign currency denominated government debt when the government
receives revenues in domestic currency only. If banks are primary holders of this type of debt, they face
default risk if the currency depreciates. Delgado et al. (2002) point out that an increased default risk may
prompt additional currency risk if banks set aside loan-loss provisions on foreign currency loans, thus
turning a covered position into an open currency one.

96

ﬂoating.

In this paper the degree of risk aversion summarizes distortionary regulation.
Thus, the emphasis is not to look for the culprit of banking crises, but to identify whether
currency mismatches have served as a transmission channel of exchange rate shocks to
the banking system. However, the implications raise questions whether the government
should implement policies that (while protecting investors) enforce ﬁrms’ and banks’ risk
taking behavior.

The paper contributes to the ﬁnancial dollarization literature, which has recently
turned its focus on the implications of dollarization for ﬁnancial stability. Arteta (2003)
and De Nicolo et al. (2003) look for a direct causality from dollarization to bank losses
and crises. Arteta (2003) does not ﬁnd evidence that dollarization (deposit or credit) has
any effects on banking stability or that banking and currency crises are more costly if
banks face higher dollarization ratios. De Nicolo et al. (2003) ﬁnd a direct relationship
between dollarization and banking stability but they only consider the effects of deposit
dollarization. Instead, it is argued here that dollarization is not harmful unless it creates a
mismatch in either ﬁrms’ or banks’ balance sheets and that mismatches prompt banking
crises only in interaction with large unexpected changes in the exchange rate”.

Secondly, this paper contributes to the less explored area of the determinants of
banking crises in transition economies. Although it does not claim that currency

mismatches have been the main reason for the occurrence of banking crises in these

 

’8 While not looking for an effect of currency mismatches on the probability of banking crises, Demirguc-
Kunt and Detragiache (2002) and Beck at al. (2003) attempt to control for banks’ exchange rate risk. Both
studies include exchange rate depreciation and the ratio of M2 to foreign exchange reserves in their
estimations and ﬁnd a positive relation between depreciation and banking crises. However it is signiﬁcant
only in Demirgilc-Kunt and Detragiache (2002).

97

countries”, it certainly shows that currency mismatches have prolonged the crises.

This study uses a panel data set for 21 transition economies. The benchmark
econometric model used in the paper is probit but the results are compared to a linear
probability model. The main ﬁndings of the econometric analysis are:

0 There is strong evidence that banks in transition economies are exposed to
default risk related to currency mismatches in ﬁrms’ balance sheets. This
result is obtained when currency mismatches are measured with credit,
deposit and real dollarization levels.

0 With an overall measure of banks’ foreign currency positions that includes
foreign assets and liabilities, the results show that in addition to default
risk, banks also face direct exchange rate risk, even though their open
positions are generally very small.

0 There is evidence that foreign exchange and trade liberalization has
increased, while institutional development of transition economies
banking systems has reduced, the likelihood of banking crises.

The paper is organized as follows. Section 2 provides a theoretical framework
relating currency mismatches to the possibility of banking crises and bank runs and
discusses its empirical implications. Section 3 deﬁnes the econometric model and
estimation methods. Section 4 describes the data. Estimation results are presented in

Section 5. Section 6 discusses some policy implications and concludes the analysis.

 

’9 Inarguably, institutional factors are far more important.

98

2. A Simple Model

If foreign currency deposits are predetermined or exogenously determined by
capital inﬂow and exports depend on external demand, the possibility of an economy-
wide currency mismatch raises the question whether banks can avoid losses resulting
from domestic currency depreciation. This Diamond and Dybvig (1983)- style model of
bank runs shows that a mismatch in either banks’ or ﬁrms’ balance sheets can channel
exchange rate shocks to the banking system and increase the probability of a banking
crisis.

There is an inﬁnite number of depositors, each of whom owns a unit of savings.

A proportion (Jr of the depositors holds foreign currency, while a proportion d=1 -df holds

local currency.

Depositors live for two periods. With probability ¢, they will consume only in the
ﬁrst period, while with probability I-¢, they will consume and need liquidity only in the
second period. At time 0 depositors — and any other market participant — know only the
probability 4;, but not whether an individual depositor is an early or late consumer.

Depositors have access to a safe storage technology that yields 1 in periods 1 and
r in period 2 on both domestic and foreign currency savings. Depositors face a
prohibitive cost of monitoring producers and, in the absence of ﬁnancial intermediation,
will prefer to put their savings in the storage technology.

There is an inﬁnite number of ﬁrms distributed between 0 and 1, each of whom
owns a unit of capital and has access to a Leontief technology. Firms use a second input

in ﬁxed proportion to capital to produce output Y in period 2. They ﬁnance the purchase

of the input with loans in foreign, If, and domestic, l, currency. Firms’ revenues are

99

realized in period 2 in either domestic or foreign currency, in period 1 ﬁrms earn 0. A

proportion nf of the output (determined by foreign demand) is exported and yields
a:
revenue if p (1 +51) (1 +52) Y. The rest sells in the domestic market and brings return equal
4! t
to an. It is assumed that purchasing power parity holds at period 0, p =p, and p Y>r.

£1 and £2 denote the depreciation rate of the real exchange rate of local currency with

respect to the US. dollar between periods 0 and l and between periods 1 and 2,

respectively. It is assumed that 81 and 62 are independently distributed and

E 0( 81) =5 0( 82) =0, Var0(81) = V ar0( 82) = 032-
Firms’ second period proﬁt function is
nf = nfp’(1+a,)(1+£2)Y+an—rL1—r[(1+gl)(1+az)1f
They maximize expected utility of proﬁts
E[nf ]— % yVar[r1f J
and their demand for foreign and domestic currency loans is

[f = ‘[rLf(l+£e)-rL]+y20'3nfp*YrLf : -[rLf(l+£e)—rL] +nfp.Y

2 2
720'ng yZO'ErLf r Lf

 

 

121—If
f

where rL and rL denote the interest rates on domestic and foreign currency loans,

and yis ﬁrms’ coefﬁcient of risk aversion.
A single bank accepts deposits and extends loans in both domestic and foreign

currency and faces no monitoring cost. Its second period proﬁt function is

100

11” = rLl+rLf(l+81X1+£2)lf —de—rg(1+gl)(1+az)df
where rD and er denote the interest rates on domestic and foreign currency

deposits.

The bank is risk averse and maximizes
E[I'Ib]- %6Var[l'lb]

where 8 denotes banks’ degree of risk aversion.
The supply of foreign currency loans is
,f = [r[(1+.~e)—2rL] H11
520§rLf
In equilibrium interest rates on loans and deposits in the same currency are equal,

but there is a premium on foreign currency loan and deposit interest rates.

rLf=rf=r
"L =rD =r[1—2r60'§(lf —df)]

The equilibrium loan dollarization level in this economy is

If: 7 "fp.Y+ 6 df
7+6 r 7+6

Thus, the more risk averse sector has a larger impact on the degree of credit
dollarization, while the other bears the exchange rate risk.

Depositors will use bank services if at time 0 they expect that the bank will be
liquid in period 1 and solvent in period 2. To ensure that it has enough liquidity in the
ﬁrst period, the bank invests a proportion ¢ of the domestic and foreign currency deposits

in the storage technology. Therefore, depositors will invest in the bank if

101

(1— ¢){r(lf— df)+r(1— d)[1— magi-(1f— CHI}: (1-¢){2aa§r2(-1f dff}>o

which is always satisﬁed.

Depositors learn in period 1 whether they are early or late consumers. However,
this information is not revealed to the bank and late consumers may run on the bank if
they expect that it may become insolvent in period 2. The bank is expected to be solvent

in period 2 if its expected net worth is positive

E1[W2]= (l- ¢)min{p Y(nf(l+£1)E1[l+£2]+n)
r(1f (1 + 31 )E1 [1 + gz]+1[1 - 2raa§(1f - df )D}
—¢)r[df (1 + £1)El[1+52]+d[1—2r603(1f —df)1|2 0
The ﬁrst term in this expression is the gross repayment by the ﬁrms. It depends
on the proportion of exported output, credit dollarization and the realized and expected
depreciation. If the gross interest payment is larger than ﬁrms’ revenues, it is assumed
that the bank can appropriate ﬁrms’ revenues without a cost“). The second term is the

repayment to late consumers in domestic and foreign currency.

2.1. Demand Driven Credit Dollarization (6=0)

Firms would prefer to borrow in the currency in which they receive their revenues
in order to avoid the exchange rate risk. Consequently, if the bank can hedge against the
exchange rate risk, it may consider matching the currency of borrowers’ revenues and

expenditures to minimize loan defaults. Thus, the bank absorbs the currency risk and

 

4° The maximum amount that the bank is able to collect from a borrower in case of default is the
borrower’s gross return.

102

offers foreign currency contracts in proportion to exports. Dollarization of bank’s assets
is entirely driven by the openness of the real economy and the demand for foreign

currency loans.

[lef—pl/
r

Thus, ﬁrms’ and banks’ currency mismatches are

nf _,f =,,f[ -111]

r

[f_df:nfp_Y_df
r

and the banks’ expected net worth becomes
E1[W2] = (l —¢)min{nfp*Y—Ifrl(l + £1)E1[1 + £2]— l]+ p*Y— 230}
+ (1 —¢)(1f -df)v[(1 +51)E1[1+ 62]— 1]

= (l—¢{nf£:-):—df]r[(l +31)E1[1+£2]—1]

r

When the domestic currency depreciates and is expected to depreciate in the
second period, the expected net worth of the bank increases when it has relatively more
dollarized loans than deposits. To avoid banking crisis with certainty, the depreciation

rate and the bank currency mismatch need to move in the same direction. However, if the

*
bank has a negative foreign currency position, such that r/ p Y/r-df <0, any increase in the

depreciation reduces bank’s net worth.

103

2.2. Bank Currency Asset and Liability Matching

The bank matches its assets and liabilities if ﬁrms are risk averse or if there are
restrictions on banks’ open foreign currency positions“. However, should a bank
maintain the same proportion of foreign currency assets and liabilities, it passes the
exchange rate risk onto its borrowers. In this case, the bank becomes exposed to default
risk when producers are not able to meet their obligations due to large exchange rate
ﬂuctuations. Therefore, the banks may have to cope with reduced net worth and the
possibility that depositors will run even though it is not exposed to exchange rate risk.

The banks’ and ﬁrms’ currency mismatches equal

If—df=o

"f _1f = nf _ df

while the expected net worth change is

E1[W2] = (1 —¢)min{(nfp’r -1f r10 +31 )151 [1 + 52]- 1]+ p’Y 4,0}

+(1-¢)(1f -df)-[(l+al)E1[l+£2]-l]

= (1 —¢)min{(nf p’Y—df r10 +31 )1?1 [1 + £2]- 1]+ [fr—no}

In this case, the critical expected depreciation rate still moves in the direction of
the mismatch. If deposits are more dollarized than ﬁrms’ revenues, the critical
depreciation rate becomes

ptY—r
nfp*Y—dfr

 

{(l+d1)El(l+d2)—1}= _

This equality shows that the higher capital productivity relative to the real interest

rate and the higher the exports, the higher the depreciation that the bank will be able to

 

" By 1999 most transition economies have adopted restrictions on open currency positions.

104

overcome.

This model has some important empirical implications. First, deposit or loan
dollarization, per se, does not affect the stability of the banking system. There must be a
mismatch in the currency structure of the banks’ balance sheets or in the real sector for
dollarization to have an effect on bank solvency. Second, even if a mismatch exists, it
would not pose a threat for the banks unless the exchange rate ﬂuctuates. Therefore, any
effect of the open position in foreign currency or of the deposit or credit dollarization on
the probability of a banking crisis should be jointly examined with the effect of the

changes in the exchange rate.

3. Econometric Model and Estimation Methods

This theoretical framework is used to develop an econometric model to test
whether currency mismatches in the real and the banking sector affect the stability of the
banking system. The probability of a banking crisis is deﬁned as

Prob(bcris is I cd, dd, rd, deprate, other controls )

= g(cd, dd, rd, deprate, other conrols)

where bcrisis is 1 if there is a banking crisis, and zero otherwise. Variable cd is
the ratio of dollar credit to total ﬁrm credit, dd is the ratio of dollar deposits to total
deposits in the banking system, and rd measures real dollarization or the ratio of the
exports to GDP. G(-) is the conditional probability function of a banking crisis. Other
controls include variables that shape the banking system structure, and the regulatory and
macroeconomic environment.

The primary econometric model used here is probit. Speciﬁcally, the sign of the

105

latent variable, net worth, indicates whether there is a banking crisis or not. Following
the theoretical model

Prob(bcrisis=1 I net worth<0)

where negative net worth is deﬁned as

— (net_worth ) = ,6] + ﬂzdeprate + ﬂ3m1'smatch + ,B4mismatch * deprate

+ 51' other _ controls + u]

,84 is expected to take negative values as predicted by the theoretical model”.

The sign of ,8; will depend on the level of mismatch at which the effect of depreciation is

estimated. For positive values of the mismatch, the depreciation rate should have a

negative coefﬁcient, while for negative values of the mismatch, the effect of real
depreciation should be positive. Similarly, the sign of ,83 depends on the rate of

depreciation at which the effect of the mismatch is estimated.

Due to the use of panel data, there might be signiﬁcant country effects. A second
issue that arises from the data set and the selection of countries is persistence in the
dependent variable. Some transition economies (Czech Republic, Slovakia, Romania)
have struggled with problems in the banking system throughout the 1990’s. To resolve
this issue, onset of banking crisis can be used as a dependent variable”. However,

multiple triggers of a banking crisis may exist and, even if not causing a crisis itself, they

 

’2 The econometric model actually deﬁnes the negative net worth as a latent variable, and so all
coefﬁcients should have an opposite sign from the one in the theoretical model.

’3 Click and Hutchison (2000), Domac and Martinez-Peria(2001), Arteta (2003) and others have adopted
this approach. Demirgiic-Kunt and Detragiache (1998, 2002) throw out all observations after the initial
onset of the banking crisis. In a study of the effect of concentration on banking crises, Beck et al. (2003)
use the second approach as a benchmark. They compare the benchmark results to the results obtained
when all observations aﬁer the initial onset date are considered to be crisis periods, and when all
observations after the initial date are considered to be non-crisis periods. In their study, the method does
not have an effect on the main conclusions.

106

may contribute to its persistence after the onset. Moreover, using only the onset of
banking crises will not provide any interesting insights on the incidence of such events in
transition economies. For this reason it is important to maintain the entire period of a
banking crisis in the sample data.

Ideally, a full set of country dummies should be included in the probit model to
control for the country effects. However, the data set does not support the use of probit
with country dummies, since some countries were in a banking crisis during the whole
sample period“, while others experienced no banking crisis at all“. A linear probability
model (LPM) with ﬁll] set of dummies is used in parallel with the probit model46 to
control for the individual country effects.

Since a banking crisis may occur at any time during the year, while all stock
variables and variables controlling for regulations are reported at the end of the year, all

independent variables are taken with a lag.

4. Data
This paper conducts empirical analysis using an unbalanced panel dataset. The
data used in the panel are annual for the period 1990-2001. Tables 3-1 and 3-2 in the
Appendix present data description and sources while Figure 3-2 shows data summary

statistics.

 

’4 Azerbaijan, Czech Republic, Latvia, Romania, and Slovak Republic.

’5 Armenia, Georgia, Kazakhstan, Kyrgyz Republic, Macedonia, and Moldova.

’6 One drawback of using LPM is that the errors are undoubtedly heteroskedastic and the standard errors
must be corrected. A second issue with using LPM is the functional form. Equation (3) in the theoretical
model shows that the net worth is linear in the variables for currency mismatch and in the interactions
between them and the depreciation rate. The probability of a banking crisis is certainly not linear in these
variables. However, the goal is not to predict banking crises but rather to ﬁnd whether there is evidence
that currency mismatches contribute to the incidence of banking crises.

107

Information about the incidence of systemic banking crises is obtained from
World Bank studies. The most comprehensive source is Caprio and Klingebiel (2003).
A banking crisis is deﬁned as systemic if most or all of the banking system capital is
eroded (Caprio and Klingebiel, 1999).47 Compiled this way, the sample contains 121
observations with 54 periods of existing banking crises and 6 episodes of onset of

banking crises.

4.1. Variables Speciﬁed in the Theoretical Model

Real dollarization is measured as the ratio of exports to GDP in percentage terms.
Deposit dollarization is the ratio of foreign currency denominated deposits to total
deposits. Credit dollarization is the ratio of non-bank sector foreign currency credit to
non-bank sector total credit. Banks’ currency mismatch is measured by the difference in
credit and deposit dollarization, while ﬁrms’ currency mismatch is the difference between
real and credit dollarization. Measured this way, the currency mismatch in banks’
balance sheets represents the difference in the currency structure of bank loans and
deposits. The ﬁrms’ mismatch measures the difference in the currency structure between
ﬁrms’ revenues and ﬁrms’ bank loans.

The depreciation rate measures the change in the end of period exchange rate to
the US dollar. The depreciation rate of the real exchange rate is calculated from the
nominal rate by correcting for domestic inﬂation. Currency crises dates are calculated

from monthly end of period exchange rates. A currency crisis is any occasion when the

 

’7 Crisis dates from Tang et al. (2000) are used where data are missing from Caprio and Klingebiel
(2003 ). Data for Moldova are from annual reports of the central bank. Although fairly underdeveloped as
rated by the EBRD, the banking system of Moldova seems to have avoided major distress including during
the Russian crisis of 1998.

108

local currency depreciates in real terms per month by more than two standard deviations
from the average monthly depreciation in any month of the year, or the monthly
depreciation is more than 15 percent and 10 percentage points higher than the previous
month. With these requirements it is insured that the currency crises are truly unusual

and signiﬁcant events“.

4.2. Other Determinants of Banking Crises

The theoretical model presented in this paper treats only issues related to the
currency and structure of the banks’ balance sheets, and shows how the mismatch may
trigger instability. However, there are some institutional and macroeconomic factors that
must be controlled for. In transition economies — where the institutional framework is
still being built and consistent macroeconomic policies are still being established - the
structure of the banking sector, regulations, and policy measures may have much larger
effects on the stability of the banking system than banks’ portfolio choices. In fact, it is
likely that the institutional framework dictates banks’ balance sheet management
choices").

In the early transition years, most banks were state owned: spawned from a few
highly specialized banks or from a single bank responsible for government ﬁnancing and

the distribution of credit resources to state owned enterprises. The newly established

 

’8 I also compared the dates with the rule given by Frankel and Rose (1996) — 25 % annual depreciation
of the average exchange rate and more than 10% change in the depreciation rate. Since a currency crisis
may be a very short-term event, the dates determined on a monthly basis is the preferred method here.
However, these dates do not differ signiﬁcantly from the ones determined using Frankel and Rose rule.

’9 A survey of bank regulations and supervisions compiled by Barth et al. (2001) gave rise to a signiﬁcant
amount of literature studying the effects of institutions on bank performance. Unfortunately, the survey
data cover only 1998 and/or 1999 and are not appropriate for use with the panel data in this paper.

109

banks inherited their asset portfolios from the previous regime. They were often coerced
by the government to lend to speciﬁc sectors or enterprises and in most cases, bank
management was unwilling or unable to deal with the severe asymmetric information
problems they faced. These institutional problems translated into weak balance sheets
and the subsequent depletion of banks’ net worth in most transition economies”. Only
after governments divested from the state-owned banks, and rules on capital adequacy,
currency positions, and limits on credit to single borrowers and related parties were
imposed and enforced, did the incidence of banking crises decrease.

To control for institutional and regulatory development of the banking system, an
index of banking reform is included in the regressions. An index of foreign exchange and
trade liberalization controls for liberalization of banking services in the capital account.

Concentration in the banking sector plays a role in systemic stability. High
concentration allows for large proﬁt margins and having a few large banks provides
banks with a buffer against adverse shocks. Beck et al. (2003) ﬁnd strong empirical
evidence for this idea“. On the other hand, large banks tend to be central banks’
favorites when it comes to last resort lending. Also, governments in transition economies
often substitute large commercial banks’ bad loan portfolios for safe government bonds.
This government appropriation of risk only exacerbates the adverse risk taking behavior
of bank managers and makes it more likely for banks to fail. Thus, the sign of the
concentration variable cannot be predicted in advance. A measure of the three largest

banks’ assets as a percent of total bank assets controls for concentration.

 

5° Tang (2000), Koford and Tschoegl (1997)
5 However, as a common practice in current empirical literature, the study does not include transition
economies.

110

Similarly to concentration, explicit deposit insurance schemes create incentives
for banks’ risk taking behavior, and thus may increase banking fragility. Demirgiic-Kunt
and Detragiache (2002) ﬁnd strong evidence that deposit insurance increases the
probability of banking crises. A binary variable indicating the existence of formal

deposit insurance is included in the estimations.

Macroeconomic and Development Indicators

Macroeconomic policies contribute to banks’ systemic health. Generally,
consistent attempts to boost economic growth and improve productivity bring high pay-
offs for the bankssz. A government that pursues non—inﬂationary policies also creates a
safe environment for banks’ operations because it protects the real value of debt, and thus
the real net worth of banks.

Gross domestic product per capita is used to control for overall economic
development. Macroeconomic uncertainty is measured with the change in the consumer
price index. The grth rate of GDP controls for the possibility of lending booms.
Domestic credit to enterprises to GDP controls for the level of ﬁnancial development

since more developed ﬁnancial systems tend to be more stable.

Trend, Time, and Regional Effects
Trend and year dummies are included in the analysis to control for any

development in the banking system apart from the adoption of new regulations.

 

’2 Demirgﬁc-Kunt and Detragiache (2002) and Beck et al. (2003) ﬁnd that real GDP growth is always
signiﬁcant and has negative effect on the probability of banking crises.

Ill

Dummies separating the countries into four main regions —Central Europe”, the Baltic
countries”, South Eastern Europe (SEE)55 and the Commonwealth of Independent States

(CIS)56 are used as an attempt to capture country effects.

5. Estimation Results

The paper investigates the effect of banks’ and ﬁrms’ currency mismatches ﬁrst
by looking at the overall effect of currency mismatches. Then, currency mismatches are
analyzed only when there is a large depreciation of the real exchange rate. Finally, the
effect of depreciation in the real exchange rate is examined for a range of plausible bank
and real sector currency mismatch values.

In all cases the interaction term between currency mismatches and the
depreciation rate uses the deviation of the depreciation rate from its sample mean. Thus,
the currency mismatch coefﬁcient can be used to obtain the marginal effect of
mismatches at the average level of depreciation (rather than zero depreciation).

Table 3-3 presents the results from the probit and linear probability models not
taking currency crises into consideration. When currency mismatches in both the real
and the banking sector and their interactions with real depreciation are included, these
variables have no obvious effects on the probability of banking crises. To avoid possible

multicollinearity between the variables, the effect of mismatches in the banking and the

 

’3 Includes Czech Republic, Hungary, Poland, Slovakia, and Slovenia.

5’ Includes Estonia, Latvia, and Lithuania.

5’ Includes Albania, Bulgaria, Croatia, Macedonia, Romania, and Yugoslavia.

’6 Includes Armenia, Azerbaijan, Georgia, Kazakhstan, Kyrgyz Republic, Moldova, Russia, and Ukraine.

112

real sector are included in separate regressions”. Only the interaction term between bank
currency mismatch and real depreciation seems to have a marginally signiﬁcant positive
effect. The positive sign is opposite from the expected one if banks are facing exchange
rate risk, which suggests that the captured effect may be default risk rather than direct
exchange rate risk. The probit model including only mismatches in the real sector gives
some indication that an increase in the real economy’s dollar revenues during an
exchange rate depreciation will reduce the probability of banking crises.

In the linear probability model the interaction terms have the correct signs when
both banks’ and ﬁrms’ mismatches are included. When only banks’ mismatches are used
in the regression, the sign is positive. This indicates again that the coefﬁcient may be
capturing the default risk.

Table 3-4 presents the results of interactions between a dummy for the incidence
of currency crises and real depreciation, currency mismatches, and their joint effects.
The probit model estimation shows that an increase of 1 percentage point in banks’
currency mismatch raises the probability of a banking crisis by 3.4 percent at the average
depreciation rate in a currency crisis. Additional increase in this probability will occur if
the currency mismatch increases simultaneously with real depreciation. Once again, it
appears that the marginal effect of bank currency mismatch captures default rather than
direct exchange rate risk. In the estimation that includes only real sector mismatches, the
interaction term is signiﬁcant and provides evidence that a simultaneous increase in

dollar net revenues in the real sector and real depreciation reduce the likelihood of

 

’7 The correlation coefficient between the real depreciation and the interaction term with banks’ mismatch
is 0.74, while with ﬁrms’ mismatch is -O.56. The correlation coefﬁcient between the two interaction terms
is «0.8.

113

banking crises.

There is little evidence that banks’ or ﬁrms’ currency mismatches affect the
probability of banking crises when currency crises are not taken into consideration.
However, when the real exchange rate becomes extremely high, increases in those
mismatches provide evidence that banks face increased default risk.

The results shown in Tables 3—3 and 3-4 also provide evidence that banking crises
have institutional origins and depend on the level of banking system development. The
index measuring existence and enforcement of regulations in the banking system has a
strong negative impact on banking crises in the probit model. In the linear probability
model, this effect is still negative but insigniﬁcant.

Concentration and foreign exchange and trade liberalization appear to increase the
probability of a banking crisis. They are marginally signiﬁcant in the linear probability
model. Bank concentration in transition economies is highly correlated with banks’ state
ownership and thus it reﬂects the degree of bank deregulation and the inefﬁciency of
state owned banks”. The index of foreign exchange and trade liberalization simply
presents the fact that an increase in the scope of the ﬁnancial system, besides bringing
new proﬁtable opportunities, carries inherent risks for banking stability”.

The depreciation of the real exchange rate may change signiﬁcance depending on

 

’8 Beck et al. (2003) ﬁnd that concentration reduces the probability of banking crises in a sample that
does not include transition economies. However, they also conclude that tighter entry restrictions increase
banking fragility. Bonin et al. (2003) ﬁnd that large banks in transition economies are less efﬁcient and
explain this with the fact that large banks are more likely to be state-owned. State ownership is not
included as a control variable in the main regressions since it signiﬁcantly reduces the sample size.

’9 Demirgﬁc-Kunt and Detragiache (1998) provide evidence that ﬁnancial liberalization increases
instability of the banking system, less so in banking systems with sound institutions. However, as in
previous studies, they associate ﬁnancial liberalization strictly with the liberalization of real interest rates.
Here the index of foreign exchange and trade liberalization is more likely to indicate increases in the risk of
banks’ non-interest income. Financial liberalization in this sense is discussed by Edwards and Mishkin
(1995) and Chami et al. (2003).

114

the level of mismatch in the banking or real sector. Figures 3-2 and 3-3 present the
marginal effects of real depreciation on the probability of banking crises when the level
of mismatch is varied within the interval of existing mismatches“). The marginal effect
of real depreciation is signiﬁcant for positive values of the banks’ currency mismatch —
between 40 and 60 percentage points in the probit model6| and between 10 and 60 in the
linear probability model. At 60 percentage points mismatch in banks’ balance sheets, a
percentage point increase in real depreciation increases the probability of a banking crisis
by 2.7 percentage points. The result is scaled down in the linear probability model to 0.4
percentage points, which indicates the importance of the country effects included in this
model.

When the currency mismatch in the real sector is varied instead, the marginal
effect of real exchange rate depreciation is largest for the negative values of the
mismatch. In the linear probability model the marginal effect is signiﬁcant in the (-60, 0)
percentage points range of the mismatch. These results provide additional support to the
evidence that banks face default risk as a result of real exchange rate depreciation. At -60
percentage points of the mismatch, a percentage point increase in real depreciation

increases the probability of a banking crisis by .6 percentage points in both models.

5.1. Alternative Speciﬁcations

To check the sensitivity of the results, some important alterations in the

speciﬁcations are made. First, nominal depreciation rather than real exchange rate

 

6° In these estimations, the interaction term includes the deviation of the currency mismatch from a
selected level, at which the effect of real depreciation is evaluated.
6' However, the underlying coefﬁcient is still signiﬁcant at 50 and 60 percentage points.

115

depreciation is used. Second, the overall bank currency mismatch scaled to total banks’
assets is used instead of the difference between credit and deposit dollarization. Third,
instead of an index of banking system reform, more institutional variables are included in

the estimations.

Nominal Depreciation

Although, it is most appropriate to use the real exchange rate in calculating
currency crises dates, periods of rapid nominal exchange rate depreciation may also
affect the banking system. The magnitude of this effect depends on the way banks
account for inﬂation expectations when setting deposit and loan interest rates. Also state-
owned borrowers, who received the largest share of loans to enterprises in transition
economies, tended to have controlled prices and thus their loan repayment ability may
have been hurt during large nominal depreciation of the currency. Therefore, nominal
instead of real depreciation is used.

Table 3-5 presents the estimation results and Figures 3-2 and 3-3 show the
marginal effects of nominal depreciation. The probit model yields results that are very
similar to the ones obtained when using real depreciation. The currency mismatch in the
banking sector and its interaction with the depreciation rate increase the probability of
banking crises, while the mismatch in the real sector and its interaction with the
depreciation rate decrease it. However, the signs change and the signiﬁcance disappears
once country effects are taken into account. In both probit and linear probability models,
the marginal effects of nominal depreciations suggest that banks face signiﬁcant default

risk.

116

Overall Bank Currency Mismatch

The exchange rate risk that banks face may be better controlled with an overall
measure of currency mismatch rather than the difference between credit and deposit
dollarization. This overall currency mismatch is calculated as the difference between
foreign currency assets (foreign assets and foreign currency loans) and foreign currency
liabilities (foreign liabilities and foreign currency deposits) and it is scaled to the level of
total bank assets. Results are presented in Table 3-6. Although the overall currency
mismatch seems to increase the probability of banking crises at the average depreciation
rate, its joint effect with the depreciation rate has a negative sign. In fact, once country
effects are accounted for, only the joint effect is marginally signiﬁcant. These results
disappear in the probit and the linear probability models when the real depreciation rate,
the currency mismatch, and their joint effect are taken only in periods of currency crises.

Figure 3-4 shows the marginal effect of depreciation at different levels of overall
currency mismatch. Clearly, banks face exchange rate risk even at low levels of currency
mismatch. According to the probit estimations, the marginal effect of the depreciation
rate is signiﬁcant and positive when the mismatch is in the (-20, 10) range. In the linear
probability model, the depreciation rate is signiﬁcant for the whole negative range of the
mismatch.

Therefore, if only deposit and credit dollarization are taken into account, default
risk seems to have a dominant effect. However, it becomes evident that banks face

exchange rate risk when their overall currency mismatch is used in the estimations.

117

Institutional Factors

To separate the effects of different institutional factors, speciﬁc regulations can be
used as controls instead of the index of banking system development. Three new
variables are used to measure regulations and structure of the banking system. Entrance
of foreign banks may have either beneﬁts or drawbacks depending on whether foreigners
succeed in bringing new expertise in banking technology and thus increase banking
efﬁciency, or their proﬁts suffer because of informational disadvantages. Also because
foreign banks may drain good customers from the existing local banks, their presence
may contribute to banking system fragility”. Foreign bank entrance is measured with the
ratio of foreign banks to total number of banks”. Binary variables for the existence of
capital adequacy requirements and restrictions on open currency positions control for
regulation of banks’ risk taking behavior and should have negative effects on banking
crises.

Two variables control for regulations in the corporate sector“. A dummy for the
existence of bankruptcy law allows an assessment of how easily banks are able to recover
unpaid loans. Another binary variable for the adoption of International Accounting
Standards measures banks’ ability to evaluate borrowers’ credit worthiness and truthful
presentation of ﬁnancial results. Lack of accounting standards subjects banks to both
adverse selection and moral hazard problems.

The de facto exchange rate regime provided by the IMF is also included in the

 

’2 Claessens (1998) and Bonin (2003) empirically support the idea that foreign banks improve efficiency.
However, Beck et al. (2003) do not ﬁnd a connection between foreign ownership and banking crises.

’3 Claessens et al. (1998) use this variable to measure foreign participation in the banking system. Barth at
al. (2001) and Beck et al. (2003) use foreign ownership in asset terms. However, there is no single and
complete source for foreign ownership of banks in transition economies.

6" These include both banks and non-bank companies.

118

regressions to control for the possible implicit guarantee of the ﬁxed exchange rate
regimes and compare the results with ﬂexible regimes.

Tables 3-7 and 3-8 provide results for these extended regressions. Banks’
currency mismatch and its interaction with real depreciation become signiﬁcant in the
probit model even when a currency crisis is not accounted for. However, the interaction
term still shows a positive sign. Once the currency crisis is accounted for, the
signiﬁcance of the joint effects becomes even more pronounced.

The extended regressions provide some insights into the institutional
characteristics of transition economies that affect their banking stability. The existence
of capital adequacy requirements is signiﬁcant in the probit and linear probability models
and indicates a reduction in the probability of a banking crisis. The existence of a deposit
safety net enters with a positive sign after country effects are accounted for. This result is
consistent with the hypothesis that deposit insurance creates incentives for banks to take
on more risk. The restrictions on open foreign currency positions reduce the probability
of a banking crisis but this variable is only marginally signiﬁcant.

The percentage of foreign banks to the total number of banks reduces the
probability of banking crises in both probit and linear probability models. The result is
consistent with previous studies that ﬁnd a positive relationship between foreign
participation and bank efﬁciency“.

Among regulations directed at both banks and ﬁrms, international accounting

standards have a very signiﬁcant and negative effect, which is maintained in all

 

’5 Claessens et al. (1998) ﬁnd that foreign bank share (number of foreign banks to total number of banks)
reduces both bank proﬁtability and banks’ overhead costs. They conclude that foreign bank entrance
increases competition but it also raises efﬁciency. Bonin et al. (2003) ﬁnd that foreign banks, especially
the ones owned by institutional investors, have increased proﬁts and efﬁciency.

119

estimations. The adoption of international accounting standards allows banks to better
evaluate their customers creditworthiness. Also, use of these standards forces banks to
realistically assess their own ﬁnancial state. As a result banks’ ﬁnancial health becomes
more apparent and easily comparable to other banking institutions, thus providing more

information to shareholders and depositors.

5.2. Endogeneity

One serious criticism of these results comes from the fact that banking crises in
transition economies have been very persistent and the explanatory variables may be
affected by a feedback from the endogenous variable. A way to reduce the problem is by
using only the onset of banking crises. However, with the available data set, only a few
onset dates enter the estimation sample“.

Another method is using instrumental variables in the linear probability model to
account for the unobserved country effects and correct for persistence. Possible
instrumental variables for the real depreciation, the bank and real sector currency
mismatches are real growth rate of advanced or European Union countries, LIBOR,
membership in WTO, and international price indices (price index for metals, energy, or
oil products). However, these instrumental variables are very weak and do not produce
any signiﬁcant results even though the instrumental variable estimation maintains the

sign of the variables.

 

66 There are only six onset dates that enter in the sample data set used in the estimations. Alternative
dependent variable can be composed of the dates when banking crises are recognized by depositors,
markets, or the government. This variable may be a closer measure of bank runs as determined in the
theoretical framework. However, there are still only 10 observations with such dates that can be used in the
estimations.

120

6. Policy Implications and Conclusions

This paper investigates whether direct and indirect sources of exchange rate risk
have been a reason for the occurrence and persistence of banking crises in transition
economies. The paper presents a simple theoretical model to derive testable empirical
implications about the link between currency mismatches in banks’ and ﬁrms’ balance
sheets and banking crises. Generally, if the real exchange rate and the open currency
positions move in the same direction, there should not be a risk to the banking system.

The empirical investigation takes two approaches. First, the effects of currency
mismatches are explored in the context of a currency crisis. Second, the effect of the real
exchange rate is tested at different levels of banks’ and ﬁrms’ currency mismatch.
Estimations with the level of ﬁrms’ currency mismatch show that the depreciation rate
increases the probability of banking crises when credits are more dollarized than ﬁrm’s
revenues. This is evidence of the indirect exchange rate risk (default risk) that banks face
as a result of real depreciation.

When the bank currency mismatch is measured by the difference between credit
and deposit dollarization, an increase in the mismatch paired with an increase in the
depreciation rate, increases the probability of banking crises. This outcome hints that the
mismatch indicator overwhelmingly measures the default rather than the direct exchange
rate risk for the banks. When the measure of mismatch is substituted with the banks’
total open currency positions, which also includes net foreign assets and is scaled to
banks’ assets, the results present evidence of existing exchange rate risk. Speciﬁcally,

real depreciation is more likely to increase the probability of banking crises when banks

121

have a negative open position.

There are several aspects in which the analysis needs improvement. First,
although there is no strong evidence of endogeneity in depreciation and mismatches, a
better assessment of the effects can be made if the dependent variable is also measured
with the onset of banking crises rather than the existence of banking crises. Without
better data for mismatches for the ﬁrst period of the 1990’s such analysis cannot be
carried out. However, this paper proposes an approach that can be used to test the effect
of currency mismatches in a larger set of countries.

Second, the paper’s ﬁndings have important policy implications with respect to
countries’ choice of exchange rate regime, or alternatively about their choice of
mechanisms to protect banks from currency risk. The results provide support for the
“fear of ﬂoating” idea — the fact that governments are unwilling to let the exchange rate
move independently if the economy is heavily indebted in foreign currency. Although
this paper does not ﬁnd evidence that the exchange rate regimes have an effect on the
probability of banking crises, a further investigation is necessary. The analysis should
take into consideration the joint effect of mismatches and exchange rate regimes. Domac
and Martinez-Peria (2000) study this link but their measure of mismatch (the ratio of
foreign liabilities and assets) is incomplete and covers only the banking sector.

Finally, since the effects of mismatches and real depreciation depend on
countries’ institutional characteristics, the analysis should include a measure of the
approaches transition economies have adopted to resolve the problems in the banking

system.

122

REFERENCES

Arteta, C. O. 2003. “Are Financially Dollarized Countries More Prone to Costly Crises?”
Board of Governors of the Federal Reserve System, International Finance
Discussion Paper 763.

Barth, J., G. Caprio, and R. Levine. 2001. “Bank Regulation and Supervision: What
Works Best.” World Bank Working Paper 2725.

Barth, J., G. Caprio, and R. Levine. 2001. “The Regulation and Supervision of Banks
Around the World: A New Database.” World Bank Working Paper 2588.

Beck, T., A. Demirgiic-Kunt, and R. Levine. 1999. “A New Database on Financial
Development and Structure.” World Bank Working Paper 2146.

Beck, T., A. Demirgiic-Kunt, and R. Levine. 2003. “Bank Concentration and Crises.”
NBER Working Paper 9921.

Beim, D. O. 2001. “What Triggers a Systemic Banking Crisis.” Columbia University,
www.gsb.columbia.edu/journals/ﬁles/ chazen/Banking_Crises.pdf.

Bonin, J. P., I. Hasan, and P. Wachtel. 2003. “Bank Performance, Efﬁciency and
Ownership in Transition Countries.” 9” Dubrovnik Economic Conference, June
2003: Dubrovnik, Croatia.

Burnside, C., M. Eichenbaum, and S. Rebelo. 2001. “Hedging and Financial Fragility in
Fixed Exchange Rate Regimes.” European Economic Review, 45:7, pp. 1151-93.

Caprio, G. and D. Klingebiel. 2003. “Episodes of Systemic and Borderline Financial
Crises.” http://econ.worldbank.org/view.php?type=18&id=23456.

Chami, R., M. S. Khan, and S. Sharma. 2003. “Emerging Issues in Banking Regulation.”
IMF Working Paper 03/101.

Chang, R. and A. Velasco. 1998. “Financial Fragility and the Exchange Rate Regime.”
Journal of Economic Theory, 92:1, pp. 1-34.

Claessens, S., A. Demirgiic-Kunt, and H. Huizinga. 1998. “How Does Foreign Bank
Entry Affect the Domestic Banking Market?” World Bank Working Paper 1918.

Cotner, J. 1991. “Currency Risk in Long-Term Borrowing.” Journal of Multinational
Financial Management, 1:2, pp. 49-64.

De Nicolo, G., P. Honohan, and A. Ize. 2003. “Dollarization of the Banking System:
Good or Bad?” World Bank Policy Research Working Paper 3116.

123

Delgado, F. L., D. S. Kanda, G. M. Casselle, and A. R. Morales. 2002. “Domestic
Lending in Foreign Currency,” in Building Strong Banks: Through Surveillance
and Resolution. Charles Enoch, David Marston and Michael Taylor eds.
Washington, DC: International Monetary Fund.

Demirgiic-Kunt, A. and E. Detragiache. 1998. “Financial Liberalization and Financial
Fragility.” IMF Working Paper 98/83.

Demirgiic-Kunt, A. and E. Detragiache. 1998. “The Determinants of Banking Crises in
Developing and Developed Countries.” IMF Staff Papers, Vol. 45, No.1 (March
1998)

Demirgiic-Kunt, A. and E. Detragiache. 2002. “Does Deposit Insurance Increase Banking
System Stability? An Empirical Investigation.” Journal of Monetary Economics,
49:3, pp. 1373-406.

Diamond, D. W. and P. H. Dybvyg. 1983. “Bank Runs, Deposit Insurance, and
Liquidity.” Journal of Political Economy, 91 :3 , pp. 401-19.

Domac, I. and M. S. Martinez-Peria. 2000. “Banking Crises and Exchange Rate Regimes:
Is There a Link?” World Bank Working Paper 2489.

EBRD. 1998-2002. “Transition Report.” European Bank for Reconstruction and
Development: London.

Edwards, F. R. and F. S. Mishkin. 1995. “The Decline of Traditional Banking:
Implications for Financial Stability and Regulatory Policy.” Federal Reserve Bank
of New York, Economic Policy Review, pages 27-45.

Frankel, J. A. and A. K. Rose. 1996. “Currency Crashes in Emerging Markets: An
Empirical Treatment.” Board of Governors of the Federal Reserve System,
International Finance Discussion Paper # 534.

Giraldi, C. and R. Hamaui. 1991. “Foreign Exchange Market Liberalization and the
International Diversiﬁcation of Borrowing by Italian Firms.” Journal of
Economics and Business, 43, pp. 309-27.

Glick, R. and M. Hutchison. 1999. “Banking and Currency Crises: How Common Are
Twins?” Federal Reserve Bank of San Francisco, Paciﬁc Basin Working Paper
Series 99-07.

Ize, A. and E. Levy Yeyati. 2003. “Financial Dollarization.” Journal of International
Economics, 59:2, pp. 323-347.

Kaminsky, G. L. and M. S. Khan. 1999. “Currency and Banking Crises: The Early
Warnings of Distress.” IMF Working Paper 99/178.

124

Kawai, M. “The Behavior of an Open-Economy Firm under Flexible Exchange Rates.”
Economica, 48: 189, pp. 45-60

Koford, K. and A. E. Tschoegl. 1997. “Problems of Bank Lending in Bulgaria:
Information Asymmetry and Institutional Learning.” Wharton School Center for
Financial Institutions, University of Pennsylvania, Working Paper 97-41.

Krugman, P. 1999. “Balance Sheet, the Transfer Problem, and Financial Crises.”
International Tax and Public Finance, 6, pp. 459-72.

Levy Yeyati, E. 2003. “Financial Dedollarization: A Carrot and Stick Approach.” CIF
Working Paper 8/2003.

Luca, A. , R. Pecchenino, I. Petrova, and P. Pollard. 2003. “Over Exposure: Foreign
Currency Open Positions in an Uncertain World.” Michigan State University,
Department of Economics Working Paper.

Luca, A. and I. Petrova. 2004. “Credit Dollarization in Transition Economies: Is it F irms’
or Banks’ Fault?” Michigan State University, Department of Economics Working
Paper.

McKinnon, R. and H. Pill. 1998. “International Overborrowing: A Decomposition of
Credit and Currency Risks.” Stanford University, Department of Economics,
Working Paper 98004.

Mishkin, F. S. 1997. “Understanding Financial Crises: A Developing Country
Perspective.” NBER Working Paper 5600.

Reinhart, C. and K. Rogoff. 2003. “The Modern History of Exchange Rate
Arrangements: A Reinterpretation.” NBER Working Paper 8963, 2002,
forthcoming in Quarterly Journal of Economics, 2004.

Rivers, D. and Q. H. Vuong. 1988. “Limited Information Estimators and Exogeneity
Tests For Simultaneous Probit Models.” Journal of Econometrics, 39 (1988), pp.
347-66.

Schneider, M. and A. Tomell. 2000. “Balance Sheet Effects, Bailout Guarantees and
Financial Crises.” NBER Working Paper 8060.

Talley, S., M. M. Guigale, and R. Polastri. 1998. “Capital Inﬂow Reversals, Banking
Stability, and Prudential Regulation in Central and Eastern Europe.” World Bank
Policy Research Working Paper 2023.

Tang, H., E. Zoli, and I. Klytchnikova. 2000. “Banking Crises in Transition Countries:
Fiscal Costs and Related Issues.” World Bank Working Paper 2484.

125

APPENDIX

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 3-1 Data Deﬁnitions and Sources
VARIABLE DEFINITION SOURCE
Banking Crisis Binary response variable =1 in the existence of a C&K ,
systemic banking crisis, 0 otherwise Tang et al.,
Beim, CB
Currency Crisis Dummy variable = 1 if the monthly depreciation IFS, CB
rate of the real exchange rate is higher than two
standard deviations above the monthly average in
any month of the year, or if it is more than 15 %
and 10 percentage points above the previous
month, =0 otherwise. Calculated from the end of
period exchange rate
Exports/GDP Total exports of goods and services, as a percent WDI, IFS
of GDP
Deposit Foreign currency denominated deposits to total CB
Dollarization deposits held by residents at domestic banks,
calculated as a percent
Credit Foreign currency denominated credit to total CB
Dollarization credit from domestic banks to domestic
enterprises, calculated as a percent. Data for a few
countries include credit to individuals, ﬁnancial
enterprises and/or government.
Net foreign assets Banks’ foreign assets minus banks’ foreign IFS, CB
liabilities (line 21...ZF... - line 26C...ZF), as a
percent of total bank deposits (total bank credit, or
total bank assets).
Depreciation Rate Percentage change of the average exchange rate IFS
(expressed as domestic currency per USD, line
RF..ZF...)
Bank Assets of the three largest banks as percent of the WB,
Concentration total assets of banks reported by Banscope. original data
from
BankScope
Number of Foreign Number of foreign banks as percent of total EBRD
Banks number of banks.
Bankruptcy Law Dummy variable, =1 if Bankruptcy law is EBRD
enforced, =0 otherwise
Capital Adequacy Dummy variable, =1 if capital adequacy EBRD
requirement is enforced, =0 otherwise
Deposit Insurance Dummy variable, =1 if there is an explicit deposit WB, Local
insurance scheme, =0 otherwise Authorities

 

 

 

126

 

Table 3-1 (Continued)

 

 

 

 

 

VARIABLE DEFINITION SOURCE
International Dummy variable, =1 if IAS are adopted, = 0 Tang et al.,
Accounting otherwise IASC, WB
Standards
Open Position Dummy variable for open foreign exchange AREAER

position limits =1 if there are limits, = 0 otherwise
Index of banking Index of banking reform and interest rate EBRD

reform (IXBR)

 

liberalization, =1 if little progress beyond
establishment of a two tier system, =2 if
signiﬁcant liberalization of interest rates and
credit allocation, =3 if substantial progress in
establishment of bank solvency and of a
framework for prudential supervision and
regulation, full interest rate liberalization with
little preferential access to cheap reﬁnancing,
signiﬁcant lending to private enterprises and
signiﬁcant presence of private banks, =4 if
signiﬁcant movement of banking laws and
regulations towards BIS standards, well
functioning banking competition and effective
prudential supervision, signiﬁcant term lending to
private enterprises, substantial ﬁnancial
deepening, =4.3 if standards and performance
norms of advanced industrial economies, full
convergence of banking laws and regulations with
BIS standards, provision of full set of competitive
banking services.

 

 

127

 

Table 3-1 (Continued)

 

VARIABLE

DEFINITION

SOURCE

 

Index of foreign
exchange and trade
liberalization
(IXFTL)

Index of trade and foreign exchange rate system
liberalization, =1 if widespread import and/or
export controls or very limited legitimate access
to foreign exchange, =2 if some liberalization of
import and/or export controls, almost full current
account convertibility in principle but with a
foreign exchange regime that is not fully
transparent (possibly with multiple exchange
rates), =3 if removal of almost all quantitative and
administrative import and export restrictions,
almost full current account convertibility, =4 if
removal of all quantitative and administrative
import and export restrictions (apart form
agriculture) and all signiﬁcant export tariffs,
insigniﬁcant direct involvement in exports and
imports by ministries and state-owned trading
companies, no major non-uniformity of customs
duties for non-agricultural goods and services, full
current account convertibility, =4.3 if standards
and performance norms of advanced industrial
economies, removal of most tariff barriers, WTO
membership.

EBRD

 

GDP per capita

GDP at market prices (current USD)per capita

WDI

 

Exchange rate

regime

Discrete variable for exchange rate regime, =0 if
ﬁxed, =1 if intermediate, =2 if ﬂoating

AREAER

 

Freely
Exchange
Regime

Falling
Rate

A dummy variable =1 if the twelve month rate of
inﬂation exceeds 40 %, or the six months after a
currency crisis (measured in a Frankel and Rose
(1996) variant) marking the transition from a
ﬁxed to ﬂoating exchange rate regime, =0
otherwise.

Reinhart
and Rogoff
(2003)

 

Inﬂation Rate

Percentage change in the consumer price index

EBRD

 

Domestic credit

 

 

Domestic credit to enterprises as a percent of
GDP.

 

CB

 

Notation:

AREAER — IMF Annual Report on Exchange Arrangements and Exchange Restrictions
C&K — Caprio and Klingebiel (2003)

CB — Central banks

EBRD — European Bank for Reconstruction and Development Transition Reports
IASC — International Accounting Standards Commission
IFS — International Financial Statistics, IMF

WDI — World Development Indicators, World Bank

128

 

Table 3-2 Banking and Currency Crises in Transition Economies

 

Country Banking Crisis High Real Free Falling
Exchange Rate Exchange Rate
Dgireciation“

 

Central Europe

 

Czech Republic 1989-2001 1998-1999 -

Hungary 1991-1995 1994, 1997 -

Poland 1990-1998 1989, 1991-1992 1989, 1991-1992
Slovak Republic 1991-2001 1993, 1999 -

Slovenia 1992-1994 1992-1993, 1995 1991-1992

 

Average Length of 8 years
a Banking Crisis

 

South Eastern Europe

 

Albania 1992-2001 1992, 1997 1991-1992, 1997
Bulgaria 1995-1997 1990, 1994, 1996- 1990-1996
Croatia 1996 19957, 1997 1993-1994
Macedonia 1993-1994 1997 1993-1994
Romania 1990-2001 1991-1992, 1997 1990-2000
Yugoslavia 1989-2001

 

Average Length of 7 years
a Banking Crisis

 

Baltic Countries

 

Estonia 1992-1995 - 1991-1992
Latvia 1995-2001 1992 1991-1993
Lithuania 1995-1996 1992 1991-1994

 

Average Length of 4 years
a Banking Crisis

 

 

67 High real exchange rate depreciation is deﬁned as monthly depreciation in excess of two standard
deviations from the mean depreciation rate for the country, or depreciation higher than 15 % and more than
10 points higher than the previous month.

129

Table 3-2 (Continued)

 

Country Banking Crisis High Real Free Falling
Exchange Rate Exchange Rate
Depreciation68

 

Commonwealth of Independent States

 

Armenia 1994-1996 1993-1994 1992-1995
Azerbaijan 1995-2001 1998-1999 1993-1995
Georgia 1991-1997 1998-1999 1991-1996
Kazakhstan 1994-1996 1994, 1999 1991-1995
Kyrgyz Republic 1991-1999 1996, 1998 1991-1999
Moldova - 1998 1998-1999
Russia 1995, 1998-1999 1992, 1998 1992-1996, 1999
Ukraine 1997-1998 1993-1994, 1998 1991-1996

 

Average Length of 4 years
a Banking Crisis

 

Sources:

Banking Crises

Caprio and Klingebiel (2003), Tang et al. ( 2001) for Georgia and Kazakhstan, Central
Banks for Moldova and Yugoslavia.

Real Exchange Rate Depreciation

IFS, Central Banks

Free Falling Exchange Rate
Reinhart and Rogoff (2003)

 

6’ High real exchange rate depreciation is deﬁned as monthly depreciation in excess of two standard
deviations from the mean depreciation rate for the country, or depreciation higher than 15 % and more than
10 points higher than the previous month.

130

 

l

 

Incidence of Banking Crises

 

Incidence of Free Falling Exchange Rate

Figure 3-1: Annual and Sample Variable Means

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a“. i.i #‘ir
. '
_oo~ “ _oo~“
_ z :
_ v _
ooa_ _ u _ aao_
w . _
L _ m
hoo_ ._ _ V eaa_ ‘
: . _ m
: H T "
$82 gm _ :32 .w
. _ _ .
f “.a _ w _ ..w
r s
ﬁmoa_ _ “ ._maa_ MM
I h * U :
. :
_oo_"~ u r ~aa~_w
V _ f
: W ._
mwo_ __ :14111rrowa_ ”V
EH . E1

 

 

131

Figure 3-1 (Continued)

Total Banks' Open Currency Positions

M
O\
O\

 

 

 

 

 

 

E

_
_:
h
.1,
_~
n .._
.m __.
t h.
a :
U :
n .3
e __
c ,—
n m.
0 h.
C ._
_:
L

A
z
:
M.

 

Deposit Insurance

 

 

 

132

Figure 3-1 (Continued)

1 Real GDP Growth
(%)

 

 

Domestic Credit

 

 

 

 

i (% ofGDP)

I

l

; 40 - ---—_ — -—~ - --—+——f “-.— w “m,

‘ O ‘f ‘“T-—T==‘r"~r—’—r— 1 T T 1—* T a ”1 J1
O5 .— m In r\ O\ .—
00 Ch 0\ O\ 0\ ON C

J O\ O\ O\ O\ O5 O\ a

 

 

 

2300 I

 

Inﬂation Rate 1
(%) *

 

 

Note:
Graphs present annual averages across countries. The straight line shows the sample
mean.

133

Table 3-3

Overall Effect of Real Depreciation and Currency Mismatches

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 1 2 3
Probit Probit Probit
Depreciation 0.1 166 0.1446 0.3917
(0.4289) (0.4212) (0.3759)
Banks' Currency Mismatch -0.0008 0.0004
(0.0052) (0.0050)
(bank mismatch)*(depreciation) 0.0456 ** 0.0398 "'
(0.0219) (0.0210)
Firms' Currency Mismatch -0.0007 -0.0005
(0.0036) (0.0036)
(ﬁrm mismatch)*(depreciation) 0.0066 -0.0030
(0.0117) (0.0078)
Concentration 0.1049 0.0723 0.1222
(0.4634) (0.4559) (0.4512)
Deposit Insurance -0.1555 -0.l418 -0.l708
(0.1623) (0.1448) (0.1620)
IXBR -0.6700 *** -O.6564 *** -0.6090 *"
(0.1807) (0.1902) (0.161 1)
IXFI‘L 0.1159 0.1359 0.1163
(0.2024) (0.1773) (0.2095)
Domestic Credit 0.0042 0.0040 0.0038
(0.0067) (0.0064) (0.0073)
GDP/Capita -0.0001 "* -0.0001 *** -0.0001 *"
(0.0001) (0.0001) (0.0001)
Inﬂation Rate -0.0023 -0.0023 -0.0019
(0.0017) (0.0016) (0.0019)
Real GDP Growth -0.0173 -0.0143 -0.0141
(0.0163) (0.0157) (0.0176)
Baltic Countries -0.3322 -0.3183 -0.3481
(0.1894) (0.1798) (0.1872)
CIS -0.6975 "* -0.6616 "* -0.7014 "*
(0.1947) (0.1737) (0.1879)
SEE -0.5757 *** -0.5323 *** -0.5869 "*
(0.1872) (0.1589) (0.1742)
Trend -0.0187 -0.0186 -0.0145
(0.0367) (0.0353) (0.0327)
Number of Observations 123 123 123
Number of Countries 21 21 21

 

 

134

 

Table 3-3 (Continued)

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 1 2 3
LPM LPM LPM
Depreciation 0.2280 ** 0.2089 0.1706 *
(0.1016) (0.1373) (0.0989)
Banks' Currency Mismatch -0.0027 -0.0028
(0.0053) (0.0031)
(bank mismatch)*(depreciation) -0.0057 0.0036
(0.0065) (0.0052)
Firms' Currency Mismatch -0.0014 0.0008
(0.0044) (0.0026)
(ﬁrm mismatch)*(depreciation) -0.0101 * -0.0076 *
(0.0053) (0.0042)
Concentration 0.6993 ** 0.6352 ** 0.6886 **
(0.2958) (0.2933) (0.3219)
Deposit Insurance 0.0439 0.0556 0.0626
(0.1282) (0.1201) (0.1232)
IXBR -0.2186 -0. 1796 -0.2202
(0.1440) (0.1477) (0.1296)
IXFTL 0.3029 ** 0.2569 *" 0.2862 ”*
(0.1085) (0.1148) (0.1012)
Domestic Credit -0.0040 -0.0045 -0.0051
(0.0071) (0.0066) (0.0063)
GDP/Capita -0.0001 -0.0001 -0.0001
(0.0001) (0.0001) (0.0001)
Inﬂation Rate -0.0001 -0.0001 -0.0001
(0.0001) (0.0001) (0.0001)
Real GDP Growth 0.0033 -0.0005 0.0036
(0.0072) (0.0076) (0.0071)
Baltic Countries
CIS
SEE
Trend -0.021 1 -0.0299 -0.0265
(0.0175) (0.0204) (0.0163)
Number of Observations 123 123 123
Number of Countries 21 21 21

 

 

135

 

Table 3-3 (Continued)

Notes: Standard errors in parentheses are robust for serial correlation

*** - signiﬁcant at the 1% level, ** - signiﬁcant at the 5 % level, * - signiﬁcant at the 10
% level

Marginal effects and their standard errors are presented for the probit model.
Signiﬁcance is for the underlying coefﬁcient.

Depreciation is calculated around the mean value of 0.93 in the interaction term with
mismatch.

All explanatory variables are lagged one period.

LPM includes a full set of country dummy variables. In all regressions, they are jointly
signiﬁcant at the 1 % level.

136

Table 3-4 Effect of Real Depreciation and Currency Mismatches in the

Existence of a Currency Crisis

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 4 5 6
Probit Probit Probit
Depreciation -2.0678 -1.2431 *** -0.2081
(1.3069) (0.6033) (0.2633)
Banks' Currency Mismatch 0.0329 ** 0.0173 *
(0.0187) (0.0125)
(bank mismatch)*(depreciation) 0.1450 * 0.0709 *"‘*
(0.0775) (0.0287)
Firms' Currency Mismatch 0.0085 *** 0.0033
(0.0028) (0.0036)
(ﬁrm mismatch)*(depreciation) 0.0261 -0.0206 *“
(0.0161) (0.0073)
Concentration 0.0797 0.1219 0.1 147
(0.3428) (0.3881) (0.4341)
Deposit Insurance -0.0770 -0.0920 -0.1110
(0.1 152) (0.1276) (0.1364)
IXBR -0.4378 *** -0.4852 "* -0.5190 "*
(0.1477) (0.1510) (0.1569)
IXFI‘L 0.0751 0.0806 0.1090
(0.1554) (0.1723) (0.1755)
Domestic Credit 0.0015 0.0021 0.0026
(0.0052) (0.0058) (0.0062)
GDP/Capita -0.0001 “* -0.0001 *"‘ -0.0001 ***
(0.0000) (0.0000) (0.0000)
Inﬂation Rate -0.0021 -0.0023 -0.0021
(0.0010) (0.0012) (0.0014)
Real GDP Growth -0.0073 -0.0100 -0.0129
(0.0135) (0.0149) (0.0163)
Baltic Countries -0.2237 -0.2684 -0.3067
(0.1360) (0.1465) (0.1680)
CIS -0.5017 *** -0.5631 "* -0.6231 ***
(0.1545) (0.1525) (0.1598)
SEE -0.3935 *** -0.4549 *** -0.5068 ***
(0.1489) (0.1480) (0.1534)
Trend -0.0156 -0.0125 -0.0065
(0.0234) (0.0254) (0.0270)
Number of Observations 122 122 122
Number of Countries 21 21 21

 

 

137

 

 

Table 3-4 (Continued)

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 4 5 6
LPM LPM LPM
Depreciation -0.2544 -0.2359 0.0092
(0.2251) (0.2126) (0.1480)
Banks' Currency Mismatch -0.0014 0.0001
(0.0048) (0.0043)
(bank mismatch)*(depreciation) 0.0187 0.0246 *
(0.0121) (0.0124)
Firms' Currency Mismatch -0.0018 -0.0012
(0.0025) (0.0018)
(ﬁrm mismatch)*(depreciation) -0.0073 -0.0130 "
(0.0052) (0.0062)
Concentration 0.6397 "' 0.6321 "' 0.6781 **
(0.3235) (0.3299) (0.3007)
Deposit Insurance 0.0657 0.0677 0.0644
(0.1120) (0.1102) (0.1144)
IXBR -0.1673 -0. 1540 -0. 1929
(0.1484) (0.1423) (0.1430)
IXFI‘L 0.2716 "' 0.2490 " 0.2799 "
(0.1357) (0.1221) (0.1326)
Domestic Credit -0.0050 -0.0053 -0.0052
(0.0072) (0.0068) (0.0070)
GDP/Capita -0.0001 " -0.0001 * -0.0001 "'
(0.0001) (0.0001) (0.0001)
Inﬂation Rate -0.0002 ** -0.0002 *" -0.0002 *
(0.0001) (0.0001) (0.0001)
Real GDP Growth 0.0002 0.0002 0.0010
(0.0078) (0.0075) (0.0073)
Baltic Countries
CIS
SEE
Trend -0.0290 -0.0309 -0.0283
(0.0227) (0.0224) (0.0206)
Number of Observations 122 122 122
Number of Countries 21 21 21

 

 

138

 

Table 3-4 (Continued)

Notes: Standard errors in parentheses are robust for serial correlation

*** - signiﬁcant at the 1% level, ** - signiﬁcant at the 5 % level, * - signiﬁcant at the 10
% level

Marginal effects and their standard errors are presented for the probit model.
Signiﬁcance is for the underlying coefﬁcient.

Depreciation is calculated around the mean value of 1.09 during currency crises in the
interaction term with mismatch.

All explanatory variables are lagged one period.

LPM includes a full set of country dummy variables. In all regressions, they are jointly
signiﬁcant at the 1 % level.

139

 

 

 

 

 

 

 

Table 3-5 Effect of Nominal Depreciation and Currency Mismatches
Dependent Variable: Banking Crisis 7 8 9
Probit Probit Probit
Depreciation 0.6052 ** 0.6426 * 0.5287 **
(0.3260) (0.3696) (0.2376)
Banks' Currency Mismatch 0.0166 0.0223 *
(0.0146) (0.0126)
(bank mismatch)*(depreciation) 0.0211 0.0287 **
(0.0159) (0.0143)
Firms' Currency Mismatch -0.0077 -0.0144 *
(0.01 17) (0.0085)
(ﬁrm mismatch)*(depreciation) -0.0103 -0.0180 "
(0.0120) @0084)
Concentration 0.13 80 0.1587 0.1287
(0.4564) (0.4741) (0.4381)
Deposit Insurance -0.2534 -0.2521 -0.2360
(0.1899) (0.1947) (0.1803)
IXBR -0.7867 *** -0.7778 “* -0.7090 “*
(0.1852) (0.1985) (0.1632)
IXF’I‘L 0.1944 0.2051 0.1462
(0.2173) (0.1892) (0.1947)
Domestic Credit 0.0036 0.0048 0.0032
(0.0081) (0.0082) (0.0078)
GDP/Capita -0.0002 *** -0.0002 *** -0.0002 ”*
(0.0001) (0.0001) (0.0001)
Inﬂation Rate -0.0040 *** -0.0038 *" -0.0040 "*
(0.0013) (0.0013) (0.0013)
Real GDP Growth -0.0108 -0.0121 -0.0121
(0.0196) (0.0192) (0.0170)
Baltic Countries -0.3309 -0.3315 -0.3304
(0.1990) (0.2216) (0.1770)
CIS -0.7571 *** -0.7835 *** -0.7212 “*
(0.1477) (0.1270) (0.1368)
SEE -0.6695 *** -0.7032 *** -0.6229 ***
(0.1484) (0.1285) (0.1293)
Trend -0.0028 -0.0043 -0.0054
(0.0375) (0.0372) (0.0324)
Number of Observations 123 123 123
Number of Countries 21 21 21

 

 

140

 

Table 3-5 (Continued)

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 7 8 9
LPM LPM LPM
Depreciation 0.0053 0.0191 0.0225
(0.0612) (0.0646) (0.0832)
Banks' Currency Mismatch -0.0041 -0.0007
(0.0056) (0.0032)
(bank mismatch)*(depreciation) 0.0045 0.0033 "'
(0.0030) (0.0017)
Firms' Currency Mismatch -0.0043 -0.0006
(0.0061) (0.0041)
(ﬁrm mismatch)*(depreciation) 0.0006 -0.0017
(0.0043) (0.0033)
Concentration 0.7489 ** 0.8033 ** 0.7995 **
(0.3183) (0.3266) (0.3558)
Deposit Insurance 0.0376 0.0774 0.0894
(0.1253) (0.1086) (0.1 106)
IXBR -0. 1580 -0.1723 -0.1926
(0.1550) (0.1437) (0.1359)
IXFI‘L 0.1646 0.1970 0.231 1
(0.1584) (0.1523) (0.1597)
Domestic Credit -0.0036 -0.0047 -0.0064
(0.0081) (0.0074) (0.0075)
GDP/Capita -0.0001 -0.0001 -0.0001
(0.0001) (0.0001) (0.0001)
Inﬂation Rate 0.0000 0.0000 -0.0001
(0.0002) (0.0002) (0.0003)
Real GDP Growth 0.0015 0.0021 0.0019
(0.0082) (0.0085) (0.0084)
Baltic Countries
CIS
SEE
Trend -0.0194 -0.0224 -0.0263
(0.0202) (0.0192) (0.0179)
Number of Observations 123 123 123
Number of Countries 21 21 21

 

 

141

 

Table 3-5 (Continued)

Notes: Standard errors in parentheses are robust for serial correlation

*** - signiﬁcant at the 1% level, ** - signiﬁcant at the 5 % level, * - signiﬁcant at the 10
% level

Marginal effects and their standard errors are presented for the probit model.
Signiﬁcance is for the underlying coefﬁcient.

Depreciation is calculated around the mean value of 0.94 in the interaction term with
mismatch.

All explanatory variables are lagged one period.

LPM includes a full set of country dummy variables. In all regressions, they are jointly
signiﬁcant at the 1 % level.

142

 

 

 

 

 

 

 

 

Table 3-6 Effect of Total Bank Currency Mismatch
Overall Effect of Real Effect of Real Depreciation and
Dependent Variable: Banking Depreciation and Currency Currency Mismatches in the
Crisis Mismatches Existence of a Currency Crisis
10 10 11 11
Probit LPM Probit LPM
Depreciation 1.3684 *" 0.3172 "* -0.0497 0.2522
(0.4484) (0.0389) (0.2585) (0.1462)
Banks' Currency Mismatch 0.0393 "" 0.0023 0.0000 -0.0148 "
(0.0083) (0.0066) (0.0129) (0.0072)
(bank mismatch)*(depreciation) -0.0899 *"'* -0.0359 * -0.0907 -0.0341
(0.031 1) (0.0169) (0.0809) (0.0568)
Concentration -0.1477 0.7206 0.3617 0.7887 "
(0.8385) (0.4885) (0.6819) (0.3886)
Deposit Insurance -0.0843 -0.0868 0.0251 -0.0417
(0.2105) (0.0939) (0.2155) (0.0882)
IXBR -0.9271 “"‘ -0.1152 -0.5595 *" -0.0919
(0.3239) (0.1127) (0.1864) (0.1235)
IXFTL 0.6022 ** 0.0727 0.5851 " 0.0676
(0.2417) (0.1525) (0.2269) (0.1766)
Domestic Credit 0.0328 "* 0.0000 0.0112 -0.0065
(0.0081) (0.0083) (0.0077) (0.0073)
GDP/Capita -0.0001 -0.0002 “"’ -0.0001 " -0.0002 *"
(0.0001) (0.0001) (0.0001) (0.0001)
Inﬂation Rate -0.0006 *"' -0.0004 "' -0.0014 "' -0.0005 "
(0.0003) (0.0002) (0.0008) (0.0002)
Real GDP Growth -0.0058 -0.0039 -0.0151 -0.0100
(0.0184) (0.0101) (0.0165) (0.01 12)
Baltic Countries 0.4128 -0.0312
(0.3004) (0.3560)
SEE 0.0032 -0.3820
(0.4425) (0.2951)
Trend -0.1162 " 0.0031 -0.0713 " 0.0069
(0.0592) (0.0191) (0.0401) (0.0227)
Number of Observations 89 89 89 89
Number of Countries 15 15 15 15

 

 

 

 

 

Notes: Standard errors in parentheses are robust for serial correlation
*** - signiﬁcant at the 1% level, ** - signiﬁcant at the 5 % level, * - signiﬁcant at the 10

% level

Marginal effects and their standard errors are presented for the probit model.
Signiﬁcance is for the underlying coefﬁcient.

Depreciation is calculated around the mean value of 0.93 and 1.09 during currency crises
in the interaction term with mismatch.

All explanatory variables are lagged one period.

LPM includes a full set of country dummy variables. In all regressions, they are jointly

signiﬁcant at the l % level.

143

Table 3-7
Alternative Speciﬁcation

Overall Effect of Real Depreciation and Currency Mismatches——

 

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 1 2 3
Probit Probit Probit
Real Depreciation 0.1650 0.6598 0.7393 "
(0.5649) (0.5676) (0.3560)
Banks' Currency Mismatch -0.0133 " -0.0058
(0.0064) (0.0056)
(bank mismatch)*(depreciation) 0.0964 "“ 0.0586 "'
(0.0346) (0.0322)
Firms' Currency Mismatch -0.0121 *" -0.0071
(0.0043) (0.0051)
(ﬁrm mismatch)*(depreciation) 0.0160 -0.0016
(0.0155) (0.01 16)
Concentration -0.83 18 -0.864 1 -0.5891
(0.5586) (0.5880) (0.5637)
Foreign Banks -0.0061 -0.0027 -0.0029
(0.0082) (0.0067) (0.0072)
Capital Adequacy -0.498l ”* -0.4809 "”" -0.546l ""
(0.1 145) (0.1608) (0.0970)
Deposit Insurance -0.48ll * -0.2646 -0.3281
(0.2360) (0.2480) (0.2484)
Open Positions -0.3720 -0.5334 “ -0.l694
(0.1947) (0.1379) (0.2359)
Bankruptcy Law 0.3458 0.0461 0.0728
(0.2553) (0.3000) (0.2885)
IAS -0.8143 ""‘ -0.7296 “* -0.819l "W
(0.1000) (0.1141) (0.1012)
IXFTL 0.3625 0.4277 "' 0.3766
(0.2435) @2447) (0.2507)
Domestic Credit 0.0069 0.0027 0.0000
(0.0086) (0.0074) (0.0085)
Exchange Rate Regime -0.0092 0.0274 0.0290
(0.1299) (0.1226) (0.1097)
GDP/Capita -0.0002 "‘ -0.0002 ”" -0.0001 "
(0.0001) (0.0001) (0.0000)
Inﬂation Rate -0.0013 -0.0010 -0.0007
(0.0022) (0.0018) (0.0010)
Real GDP Growth 0.0016 0.0089 0.0235
(0.0235) (0.0255) (0.0227)
Baltic Countries -0.0804 0.0686 -0.0081
(0.4437) (0.3846) (0.3752)
CIS -0.8626 "* -0.6598 "" -0.7409 ""
(0.0836) (0.0843) (0.1329)
SEE -0.4664 -0.2682 -0.3027
(0.2855) (0.2835) (0.2389)
Trend 0.0899 " 0.0560 0.0513
(0.0524) (0.0440) (0.0429)

 

144

 

Table 3-7 (Continued)

 

 

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 1 2 3
LPM LPM LPM
Real Depreciation -0.0354 0.0334 -0.0430
(0.2014) (0.1836) (0.1 128)
Banks' Currency Mismatch -0.0029 0.0041
(0.0034) (0.0025)
(bank mismatch)*(depreciation) -0.0009 0.0060
(0.01 14) (0.0078)
Firms' Currency Mismatch -0.0081 " -0.0063 ”
(0.0038) (0.0023)
(ﬁrm mismatch)*(depreciation) -0.0107 -0.0099 "
(0.0068) (0.0048)
Concentration 0.5482 0.5068 0.5200
(0.3380) (0.4104) (0.3561)
Foreign Banks -0.0135 "‘ -0.0118 “" -0.0134 "‘
(0.0028) (0.0025) (0.0026)
Capital Adequacy -0.2457 -0.2375 -0.2508
(0.1579) (0.1558) (0.1594)
Deposit Insurance 0.0953 " 0.1192 " 0.0942
(0.0550) (0.0635) (0.0568)
Open Positions -0.0872 -0.0735 -0.0866
(0. 1726) (0.2044) (0. 1767)
Bankruptcy Law 0.0247 0.0411 0.0362
(0.1530) (0.1608) (0.1478)
IAS -0.3649 ” -0.3353 ” -0.3694 "
(0.1560) (0.1588) (0.1514)
IXFI‘L 0.3717 ‘“ 0.3042 ”‘ 0.3647 ”"
(0.1 109) (0.0698) (0.091 1)
Domestic Credit -0.0138 " -0.0162 ‘" -0.0152 ”
(0.0059) (0.0056) (0.0054)
Exchange Rate Regime 0.0325 0.0124 0.0251
(0.0683) (0.0578) (0.0622)
GDP/Capita -0.0002 "' -0.0002 ’ -0.0002 "
(0.0001) (0.0001) (0.0001)
Inﬂation Rate -0.0008 “‘ -0.0008 ‘" -0.0008 ‘“
(0.0002) (0.0003) (0.0002)
Real GDP Growth 0.0107 0.0042 0.0106
(0.0102) (0.0094) (0.0097)
Baltic Countries
CIS
SEE
Trend 0.0524 "" 0.0266 0.0477 "
(0.0183) (0.0206) (0.0175)
Number of Observations 109 109 109
Number of Countries 21 21 21

 

145

 

Table 3-7 (Continued)

Notes: Standard errors in parentheses are robust for serial correlation

*** - signiﬁcant at the 1% level, ** - signiﬁcant at the 5 % level, * - signiﬁcant at the 10
% level

Marginal effects and their standard errors are presented for the probit model.
Signiﬁcance is for the underlying coefﬁcient.

Depreciation is calculated around the mean value of 0.93 in the interaction term with
mismatch.

All explanatory variables are lagged one period.

LPM includes a full set of country dummy variables. In all regressions, they are jointly
signiﬁcant at the 1 % level.

146

Table 3-8

Effect of Real Depreciation and Currency Mismatches in the
Existence of a Currency Crisis — Alternative Speciﬁcation

 

 

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 4 5 6
Probit Probit Probit
Real Depreciation -2.2883 "" -2.1558 "“ -0.1967
(0.8450) (0.6164) (0.2974)
Banks' Currency Mismatch 0.0174 0.0108
(0.01 13) (0.0106)
(bank mismatch)*(depreciation) 0.1670 ”* 0.1366 ”"
(0.0499) (0.0289)
Firms' Currency Mismatch 0.0042 0.0081 "
(0.0030) (0.0047)
(ﬁrm mismatch)*(depreciation) 0.0166 -0.0330 ‘"
(0.0104) (0.01 1 1)
Concentration -0.5704 -0.5716 -0.5577
(0.5391) (0.5497) (0.5513)
Foreign Banks -0.0018 -0.0017 -0.0024
(0.0056) (0.0057) (0.0063)
Capital Adequacy -0.6109 "" -0.6123 "" -0.4907 ”"
(0.1482) (0.1523) (0.1518)
Deposit Insurance -0. l 791 -0. l 921 -0.2039
(0.2426) (0.2476) (0.2206)
Open Positions -0.4423 ‘ -0.4538 " -0.3557
(0.2449) (0.2331) (0.2444)
Bankruptcy Law 0.1304 0.1308 0.0907
(0.2054) (0.2091) (0.2418)
IAS -0.7369 ‘" -0.7500 ”'" -0.7212 "‘
(0.1393) (0.1326) (0.1314)
IXFI'L 0.5498 "’ 0.5576 "" 0.4786 "
(0.2043) (0.2077) (0.1991)
Domestic Credit -0.0021 -0.0021 -0.0012
(0.0071) (0.0072) (0.0075)
Exchange Rate Regime 0.0802 0.0830 0.0491
(0.1355) (0.1371) (0.1150)
GDP/Capita -0.0002 *” -0.0002 “" -0.0002 ‘"
(0.0001) (0.0001) (0.0001)
Inﬂation Rate -0.0020 ” -0.0021 “ -0.0016 "
(0.0009) (0.0009) (0.0009)
Real GDP Growth 0.0136 0.0133 0.0149
(0.0201) (0.0205) (0.0197)
Baltic Countries 0.0522 0.0512 0.0032
(0.3499) (0.3541) (0.3322)
CIS -0.5002 *” -0.5157 ”“ -0.5791 ""
(0.1055) (0.0976) (0.0997)
SEE -0.2270 -0.2394 -0.2194
(0.2198) (0.2223) (0.2142)
Trend 0.0558 0.0610 0.0567
(0.0438) (0.0442) (0.0389
Number of Observations 109 109 109
Number of Countries 21 21 21

 

147

 

Table 3-8 (Continued)

 

 

 

 

 

 

 

 

 

Dependent Variable: Banking Crisis 4 5 6
LPM LPM LPM
Real Depreciation -0.6561 " -0.5191 ‘ -0.2579
(0.3495) (0.2766) (0.1802)
Banks' Currency Mismatch -0.0006 0.0006
(0.0034) (0.0034)
(bank mismatch)*(depreciation) 0.0227 0.0312 "
(0.0188) (0.0153)
Firms' Currency Mismatch -0.0023 -0.0017
(0.0029) (0.0026)
(firm mismatch)*(depreciation) -0.0144 -0.0188 "
(0.0097) 41.0083)
Concentration 0.6922 ‘ 0.6492 0.7028 ‘
(0.3841) (0.3874) (0.3771)
Foreign Banks -0.0139 "“ -0.012l ”“ -0.0143 ""
(0.0037) (0.0028) (0.0037)
Capital Adequacy -0.2449 -0.21 70 -0.24 l 8
(0.1557) (0.1482) (0.1534)
Deposit Insurance 0.1532 " 0.1411 “ 0.1695 “
(0.0710) (0.0651) (0.0677)
Open Positions -0.0297 -0.0642 -0.0014
(0.1855) (0.2022) (0.1927)
Bankruptcy Law 0.0040 0.0009 -0.0014
(0.1710) (0.1726) (0.1625)
IAS -0.2690 ‘ -0.2888 " -0.2628
(0.1410) (0.1270) (0.1613)
IXFTL 0.3859 ”‘ 0.3222 ”" 0.4033 “‘
(0.1235) (0.0823) (0.1 186)
Domestic Credit -0.0140 “ -0.0140 " -0.0134 ”
(0.0054) (0.0052) (0.0053)
Exchange Rate Regime 0.0404 0.0232 0.0408
(0.0664) (0.0630) (0.0574)
GDP/Capita -0.0002 -0.0002 -0.0002 '
(0.0001) (0.0001) (0.0001)
Inﬂation Rate -0.0012 "" -0.0010 “‘ -0.0011 ”“
(0.0003) (0.0003) (0.0003)
Real GDP Growth 0.0029 0.0018 0.0037
(0.0109) (0.01 10) (0.0095)
Baltic Countries
CIS
SEE
Trend 0.0399 0.0355 0.0396 ‘
(0.0233) (0.0261) (0.0210)
Number of Observations 109 109 109
Number of Countries 21 21 21

 

148

 

Table 3-8 (Continued)

Notes: Standard errors in parentheses are robust for serial correlation

*** - signiﬁcant at the 1% level, ** - signiﬁcant at the 5 % level, * - signiﬁcant at the 10
% level

Marginal effects and their standard errors are presented for the probit model.
Signiﬁcance is for the underlying coefﬁcient.

Depreciation is calculated around the mean value of 1.09 during currency crises in the
interaction term with mismatch.

All explanatory variables are lagged one period.

LPM includes a full set of country dummy variables. In all regressions, they are jointly
signiﬁcant at the 1 % level.

149

 

 

mow-no
annual aaa aa a

on em ow an on o_ o 3. cm. cm- ow- an- oo-
N:-
illlal‘l‘l‘llI‘I‘l

- P p r b p p p p b P o

 

 

 

 

II N c
To

 

 

 

3::
an: u 3.355 .o .9553;
on» no uoﬁa—uouaon:£ﬁoz uo «ovum

cc cm av on an o— c on- om-cm- c..- an. ce-

gov-no

 

 

 

 

 

 

 

 

 

 

 

 

 

Agra-C
auto 2.3.25 2 banana: 2..
no :oﬁauuoumon_au& Go «on: amass-2

 

 

 

co 0m ow On an A: c on- GNU-em- 31cm- oo-

 

 

 

 

 

 

 

 

00 cm o.. cm on O“ o om- on- on- ow- cm- oo-

uu-vo

 

 

 

 

 

 

 

 

 

 

”I
. - T - N-
. . . . .ll.\111.l1....ll - - o i . Ii.lll.l.l..|1111 . . . . o
ll ll 1 ~ llll N
m w
333 .3 9:28
v.35 mam-3:. banana... .35 wax-am: 9.32:5 ......
o... no Iota—ovuaonzﬂaoz no 383— no 2.283.279 :0“ he 38:— zine-2
..Soom memo—gum 95

3 £82.52 5:955 he £95..— .3359 .a 83.5 “—5.55 ..e 3:35:95 05 .5 55.59509 .3 802m «Am 953%

 

150

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.52: “5.?
O OOOGQQDDQOO. I90...
00 cm 3. on ON on o 07 o~- cm- ov- cm- oo- oo cm ow on on A: o o"- om- cm. 0... cm- oo-
~.OI . QC-
fink/p h o b n n / P p h 0
hi no 11!, .....o
m o ”o
3:: 2.3.:
aauo was—...;— 3 baa—So: 2: .3 £95 3.225 .9 5:333.— 2:
accumuuumun Elie?“ he tutu Ends-:7- no mac-392.9: 23% be «no—um 259:2
F.-—Z no-2
II I. 9.. .0. ... ID. i.
I {I Id. QC. 0.. ID. IO. 0.. G
cc cm on em on o— o 3. on- cm- 3.. cm. oo- oo 0m ow cm on o“ o ou- om- om- 0... cm- oo-
7 o
F W p b P n _ p - n r p N.°
if 0 11111 to
- ill-Ill _ 0.0
N ad
938 my 539::
«3.0 wax-am .0 953:3: 2: .3 auto wan-am do b22235 a...
nota—ooumon—zaﬁez uo Seam—«Ewan: no aomuﬂooumonnaom Go to: u—«Swuaz

 

 

 

 

..38m .3“ 2:
E 5.55:2 5:955 he £964 2.9.0.25 «a mambo uni—:5 he 3:539:— 05 .5 5539950: .3 got-m min 95qu

 

151

 

Figure 34 Effect of Real Depreciation on the Probability of Banking Crises at
Different Levels of Total Currency Mismatch in the Banking Sector

 

i Marginal Eﬂ‘ect of Real Depreciation on the Probability

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

J ofBankingCr's'B
(Probit)
I 6 A r “* * r ” ’“M‘ ‘
I
: 0 ‘3 h”??? r“7—T 1"“T*rr"1—T"‘7"W*T 1 l "7 . r r , ﬁY a I
z [ i.
‘3 "T‘ _ “”7 " ﬂ __ PP l_l “W“
ormomomovnOWOtnOi/uo
Y'w‘P‘Y‘Q‘TT' ”"NNM
***************************ﬁt** * I”!
i 1|! *************** I!
} total bank currency m'srmtch
, 7,, ~ MM” wg—u “-gﬁ.g-r~__g__—_i 7
} Marginal Eﬂ‘ect of Real Deprec'ntion on the Probability '
5 ofBankingCrisis
1 (LPM)
' l
l 2 :
I 3
1 f
I o.
l
y -1 1 “ 7“ T "*' ‘ r] ’ “d
l o,momomOmow‘o.m'cmc
{ 'T.’?‘?"‘."7"." "I”‘NNM'
f Hut "in *umntuueutitu I ’ l i .
4 total bank currency misrmtch
Notes to tables 3-2 to 3-4

In the probit model: First row presents signiﬁcance of the underlying coefﬁcient.
Second row presents signiﬁcance of the marginal effect.

*** - signiﬁcant at the 1% level, ** - signiﬁcant at the 5 % level, * - signiﬁcant at the
10 % level

152

   

uWiring"MIMWM