Are Foreign Banks a 'Safe Haven'? Evidence from Past Banking Crises

Contributor Notes

Author’s E-Mail Address: gadler@imf.org; ecerutti@imf.org.

The presence of foreign banks in emerging markets has increased markedly over the last two decades, raising questions about their potentially stabilizing or destabilizing role during times of financial distress. Most studies on this subject have focused on banks’ asset side (i.e., their lending behavior). This paper focuses on their liability side, studying the behavior of depositors vis-à-vis foreign banks. We rely on data from the banking crises in Argentina and Uruguay over the period 1994-2002 to conduct the study. The paper focuses on three questions; (i) are foreign banks perceived as a safe haven during bank runs?; (ii) does their legal structure (branch versus subsidiary) matter?; (iii) do perceptions depend on the nature of the crisis? Contrary to the commonly held view that foreign banks play a stabilizing role during domestic banking crises, we do not find robust evidence in this regard. Only in one (large) bank run episode, out of five studied, there is evidence of safe haven perceptions towards foreign branches.

Abstract

The presence of foreign banks in emerging markets has increased markedly over the last two decades, raising questions about their potentially stabilizing or destabilizing role during times of financial distress. Most studies on this subject have focused on banks’ asset side (i.e., their lending behavior). This paper focuses on their liability side, studying the behavior of depositors vis-à-vis foreign banks. We rely on data from the banking crises in Argentina and Uruguay over the period 1994-2002 to conduct the study. The paper focuses on three questions; (i) are foreign banks perceived as a safe haven during bank runs?; (ii) does their legal structure (branch versus subsidiary) matter?; (iii) do perceptions depend on the nature of the crisis? Contrary to the commonly held view that foreign banks play a stabilizing role during domestic banking crises, we do not find robust evidence in this regard. Only in one (large) bank run episode, out of five studied, there is evidence of safe haven perceptions towards foreign branches.

I. Introduction

Over the last two decades, foreign banks have markedly increased their participation in emerging markets, raising questions about their potentially stabilizing or destabilizing role during times of financial distress.1 Most empirical studies in this area have focused on the foreign banks’ asset side (i.e., their lending behavior)2 but little attention has been paid to their liability side. From this perspective, foreign banks could play a stabilizing role in host countries during domestic crises if depositors perceived them as a “safe haven” (i.e., for “flying to quality” within the country’s banking system). Various reasons are behind such usual perceptions: foreign banks may be backstopped by their parent banks or suffer less from liquidity shortages as they may have (more stable) internationally diversified funding bases. Foreign banks, however, could also play a destabilizing role as they could provide avenues for capital flight through facilitating access to the foreign bank’s international network, or could be more predisposed to pull out from emerging markets in times of distress. Foreign banks’ legal structure (branch versus subsidiary) along with the nature of the banking crisis (systemic versus non-systemic) could also determine their stabilizing or destabilizing role, as these factors could affect the degree of parent banks’ support as well the depositors desire to fly out of the banking system.

A few papers have analyzed depositors’ behavior towards foreign banks, with overall inconclusive results. Barajas et al (2007) find that, after controlling for bank fundamentals and macroeconomic variables, foreign banks actually lost proportionally more deposits than domestic banks during the 2001 bank runs in Argentina. The study, however, focuses only on the 2001 crisis, disregarding other previous banking crisis episodes, and overlooks the legal structure (branch, subsidiary) of foreign banks. Similarly McCandless et al (2003) study the determinants of bank runs during the 2001 Argentine crisis, but pay little or no attention to the role of foreign banks, and their legal form. Goday et al (2005) study depositors’ role in exerting market discipline on Uruguayan banks during the 2002 crisis, and find little evidence to support the safe haven hypothesis.3 For other countries, Kraft and Galac (2006) find that foreign banks were perceived by Croatian depositors as safe havens during the 1998-99 banking crisis. In a more comprehensive cross-country study, Arena et al (2007) indirectly conclude that foreign banks have an advantage over domestic banks in attracting deposits, as their deposit and lending rates tend to be smoother during crisis periods. However, by focusing on annual data, the study overlooks the dynamics of deposits and is not suitable for capturing short-lived bank runs.

This paper sheds light on some of these gaps in the literature, focusing on three relatively unexplored questions related the role of foreign banks from the perspective of the liability side of their balance sheet:

  1. Are foreign banks perceived as safe havens in host countries during banking crises? That is, do they outperform domestic banks in attracting/retaining deposits, above and beyond what can be explained by differences in bank fundamentals and compensation for risk (i.e., interest paid)?;

  2. Does the legal form (branch versus subsidiary) matter? In theory, foreign branches offer more protection to depositors than foreign-owned subsidiaries since a parent bank is under no legal obligation to honor subsidiary liabilities in excess of the capital invested.

  3. Does the nature of the crisis matter? The attitude of depositors towards foreign banks could be different during systemic and non-systemic banking crises, since the former are more likely to happen together with macroeconomic and political crises, increasing the parent bank’s cost of rescuing affiliates, and triggering ring fencing provisions.4

We address these questions by studying the behavior of depositors vis-à-vis foreign banks during episodes of financial distress in Argentina over the period 1994-2002, and Uruguay over the period 1999-2002. We have chosen these two countries and periods because of their large foreign bank participation, the local/regional characteristic of the banking crises, and the availability of public bank-level monthly balance sheet data. Another important advantage for our analysis is the fact that parent banks of the foreign banks under study were not significantly affected during the crises under study, with the exception of some foreign regional banks, which we identify and separate from non-regional foreign banks. Moreover, the focus on the Argentinean banking sector during much of the 1990s allows us to study multiple episodes of both systemic and non-systemic bank runs, as well as to compare the performance of domestic- versus dollar-denominated deposits in the context of a currency board.5 The case of Uruguay is of additional interest, as it allows us to distinguish between the behavior of resident and non-resident depositors (e.g. non-resident deposits represented 40 percent of total deposits at the time of the crisis).

Our results indicate that, although foreign banks may sometimes be perceived as safe havens during domestic bank runs, this is not a general pattern, including after controlling for bank fundamentals and interest rate responses. In fact, we find that only in one out of the five cases studied (Argentina, 1995), there is evidence of safe haven perceptions, and only towards foreign branches. At the same time, there is one other episode (Argentina, 2001), when foreign branches actually faced larger deposit withdrawals, even controlling for fundamentals and interest rate responses, possibly indicating different expectations regarding the possible triggering of ring-fencing provisions. Foreign subsidiaries, in turn, do not appear to have been perceived differently in any of the cases.

The structure of the paper is as follows: Section II briefly describes the structure and evolution of the Argentinean and Uruguayan banking systems, the behavior of depositors, and the main macroeconomic developments, during the periods under consideration. Section III presents the econometric methodology and results. Section IV discusses the key conclusions.

II. Background

A. Argentina’s Banking System

During the period 1994-2001, Argentina’s banking sector underwent numerous transformations, consolidating the number of institutions and making considerable improvements in the regulatory framework. Along with these changes, there was a marked increase in the market share of foreign banks. Most of the reforms took place following the 1995 banking crisis and meant that, by late 1998, the Argentine banking industry had been catapulted to a rank of second (after Singapore, and tied with Hong Kong) in terms of the quality of its regulatory environment, according to the World Bank.6

During this seven year period, the total number of institutions decreased by about half, from 168 to 83, mostly as a result of numerous merger and acquisitions of cooperative banks and the privatization of several small provincial banks (see Table 1).7 Simultaneously, the number of foreign-owned institutions increased from 31 banks in November 1994 to 36 banks in November 2001, driven by the rise in the number of foreign-owned subsidiaries. As a result, the increase in market share of foreign banks was substantial, controlling approximately half of the assets of the system towards the last years of the sample.8

Table 1.

Argentina: Structure of the Banking System, November 1994 and November 2001.

article image
Source: Central bank of Argentina and IMF staff estimates.

Based on non-financial private sector deposits.

Percentage of deposits within the same bank group.

Included both branches and subsidiaries of regional public and private banks.

Argentina’s banking system faced four distinct bank run episodes during this period: December 1994-April 1995 (associated with the Mexican crisis); October-November 1997 (Asian crisis); August-September 1998 (Russian crisis); and October 2000-November 2001 (Argentina’s own crisis).9

These events had a noticeably different impact on foreign banks compared to domestic banks (see aggregate and bank-level figures in Figure 1 and Annex 1’s Figure A, respectively).

Figure 1.
Figure 1.

Argentina: Deposit Evolution During Bank Runsa/

(Private resident non-financial sector’s deposits; percent change during selected periods)

Citation: IMF Working Papers 2015, 043; 10.5089/9781616357047.001.A001

Note: a/ Figures are corrected from merger and adquisitions among banks belonging to different ownership groups during each bank run episode.

The December 1994-April 1995 bank run episode was triggered by agents’ fears that Argentina would exit the currency board after Mexico’s currency devaluation on December 20, 1994.10 The run started on peso denominated deposits but spread to dollar deposits soon after. Within four months, resident deposits had fallen by about 16 percent, with almost 90 percent of the banks losing deposits at the peak of the crisis. Interestingly, most foreign branches, especially the largest ones, were able to increase both peso and dollar denominated deposits during this time. Not even a higher increase in interest rates was able to stop the deposit losses of small domestic private banks.

The October-November 1997 and August-September 1998 bank runs followed the Asian crisis—more specifically the first attack on Hong Kong’s currency board—and the Russian default, respectively. In the aggregate, both episodes were essentially a run on peso denominated deposits without systemic characteristics, in the sense that the bank system as a whole did not experience a significant change in the level of total deposits. There was only a 1 percent deposit loss during the 1998 bank run, and even a small 2 percent increase in 1997. However, there was again an important redistribution of deposits across banks, with most foreign branches and subsidiaries gaining deposits, at the expense of domestic institutions.

Finally, the more pronounced bank run occurred during October 2000-November 2001. This was a long episode in relative terms, mostly driven by concerns about the sustainability of the currency board. The run was triggered by the resignation of the Argentinean vice-president on October 6, 2000, and ended with the deposit freeze on December 1, 2001.11 During this period, resident deposits fell by about 20 percent, with more than 80 percent of the banks facing deposit loses. Unlike previous bank runs, foreign banks, especially branches, experienced proportionally higher deposit withdrawals than domestic banks, reaching close to 25 percent of their deposits. Finally, foreign regional banks lost more deposits than any other group, but their market share was very small (less than 1 percent).

B. Uruguay’s Banking System

During much of the 1990s and up to the 2002 financial crisis, Uruguay’s banking system was perceived as a safe financial hub in Latin America, partly reflecting an implicit and unrestricted government guarantee, a major presence of foreign banks in the system and a strong institutional framework (Uruguay was one of a few Latin American countries with investment grade status). Despite being composed of only 22 banks and 6 cooperatives, the domestic financial system was relatively large, with assets representing about 110 percent of GDP by end-2001.12 Both foreign banks and public domestic banks had a strong presence in the system (Table 2), in the first case reaching about 58 percent market share, including a 10 percent participation of regional banks; and in the second case reaching a 39 percent market share. Unlike Argentina, Uruguay’s banking system had a large participation of non-resident depositors—mainly Argentines—accounting for about 40 percent of total deposits at end-2001.13 Notably, deposits from non-residents were largely concentrated in foreign banks. Indeed, non-residents using the Uruguayan banking system, invested about 77 percent of their funds in foreign banks, 10 percent in local private banks and 13 percent in local public banks; compared to 21 percent, 18 percent and 57 percent respectively for resident depositors. As a result, non-resident deposits accounted for about 60 percent of the depositor base of foreign banks, while accounting for only 5 percent of the depositor base of domestic banks and cooperatives. Another remarkable aspect of the Uruguayan banking system at the time of the crisis was the high degree of dollarization, evidenced in the 91 percent share of foreign currency deposits.

Table 2.

Uruguay: Structure of the Banking System 1/

article image
Source: Central bank of Uruguay and IMF staff estimates.

Excluding offshore and non-bank financial institutions.

Includes both branches and subsidiaries of regional private and public banks.

Based on non-financial private sector deposits.

Percentage of deposits within the same bank group.

Starting in 1999, a number of adverse shocks—including the devaluation of Brazil’s real, the recession in Argentina and an outbreak of foot-and-mouth disease—brought the economy to a recession. Still, up until 2001, confidence in the Uruguayan banking system remained stable. Furthermore, as the crisis in Argentina developed, Uruguay was able to attract a large volume of deposits from Argentines. In the first quarter of 2002, however, public confidence in the Uruguayan banking system began to erode at the time that cash-strapped Argentine depositors, unable to access their accounts in Argentine banks (following a deposit freeze in December 2001 and pesification in early 2002), started withdrawing their funds from Uruguay. Soon after, the crisis spread to Uruguayan residents.14

The 2002 financial crisis that followed had an enormous impact on the Uruguayan financial system, leading to the bankruptcy of many institutions and an unprecedented bank run, which resulted in a loss of about 46 percent of total deposits of the system in a period of only seven months (December 2001-July 2002).

However, while all banks were affected in one way or another, the extent of the deposit run was not homogenous across them (see aggregate and bank-level figures in Figure 2 and Annex 1’s Figure B, respectively). A first glance at the data shows that private domestic banks faced deposit losses of about 27 (22) percent of foreign (local) currency deposits, while foreign banks experienced substantially larger deposit losses, averaging 54 (28) percent. Further, in the aggregate (including resident and non-residents deposits) non-regional foreign affiliates were most affected, with subsidiaries and branches experiencing foreign-currency deposit losses of 65 and 57 percent, respectively. Regional foreign banks, public domestic banks and cooperatives were also largely affected, although deposit losses were significantly lower.

Figure 2.
Figure 2.

Deposit Evolution During 2002 Bank Run in Uruguay

(Private non-financial sector deposits; percent change)

Citation: IMF Working Papers 2015, 043; 10.5089/9781616357047.001.A001

Note: 1/ In Uruguayan pesos; 2/ In US dollar terms.

However, these statistics mask the effect of the deposit run by non-residents, which were mainly concentrated in foreign banks.15 When excluding non-resident deposits, the picture changes significantly. Most noticeable is the fact that deposit losses of foreign branches were very much in line with those of domestic (public and private) banks—at around 30 percent—and quite lower than those of subsidiaries (51 percent), possibly suggesting that depositors were able to discriminate among foreign banks according to their legal structure and their different embedded risks. Finally, regional institutions were the most affected, with losses averaging 52 percent of deposits, possibly reflecting their exposure to Argentine risk (many of the owned by Argentinean banks) and some idiosyncratic solvency issues.16

The patterns of deposit withdrawals across the different episodes in Argentina and Uruguay suggest that foreign banks may have played a stabilizing role in some episodes of local financial stress, but such role may have varied depending on the nature of the crisis as well as the legal structure of those banks. However, these patterns could reflect, at least in part, differences in the strength of bank balance sheets across groups. Disentangling whether foreign banks (and their different legal forms) played a stabilizing role because of safe haven perceptions or because of differences in fundamentals requires controlling for such differences. This is performed in the next section.

III. Econometric Exercise

Our goal is to assess whether the patterns of deposit flows vis-a-vis foreign banks described above reflected safe haven perceptions. This requires controlling for bank fundamentals (as institutions with stronger balance sheets would naturally retain more deposits, independently of their ownership type) as well as interest rate responses (as different institutions may have reacted differently to prevent deposit losses).

A. Methodology

Following the literature on market discipline, the behavior of depositors vis-à-vis different types of banks during bank run episodes is explored through bank-level panel data estimations of the following equation:

yi,t=yi,t1α+ii,tβ+fi,tγ+mtδ+dioθ+εi,t   (1)

where i is the bank and t is the month; yi,t is the monthly change in total resident deposits; ii,t is the (implicit) average interest paid by each bank, and it captures the bank’s endogenous response to retaining or attracting deposits; fi,t is a vector of bank level fundamentals that capture banks’ asset quality, liquidity, profitability, capitalization levels, size as well as banks’ exposure to exchange rate and sovereign risk (see Tables A and B in Appendix 1 for a description and the definition of each variables in Argentina and Uruguay, respectively);17 mt is a vector of macroeconomic variables, including measures of devaluation expectations and country risk, meant to capture systemic risks. Finally, dio is a vector of bank-ownership dummy for public banks, foreign branches, foreign subsidiaries, or foreign regional banks. These dummies—which capture differences in bank-ownership and legal structure—are our variables of main interest, as they are a proxy for the unobserved characteristics of foreign banks, such as safe haven perceptions. See Tables C and D in Appendix 1 for summary statistics.

Equation (1) is estimated using the system linear generalized method of moments (GMM), also known as Arellano-Bover/Blundell-Bond GMM estimator. This estimation method is well suited for our objectives, since it is designed to efficiently estimate the effect of time-invariant variables (e.g. ownership characteristics), while allowing for right-hand side variables that are not strictly exogenous, such as the banks’ interest rates.18

The analysis of depositor behavior during bank runs is carried over the period December 1994-November 2001 in Argentina, and December 2001-July 2002 in Uruguay. Bank run episodes are defined as the periods when there are two or more months of deposit losses in the banking system, or when more than one-half of the banks were losing deposits. Based on these rules we study the previously mentioned 4 bank run episodes in Argentina and 1 in Uruguay. The criteria of two or more months with a decrease in the banking system deposits identifies 3 bank runs periods in Argentina (Dec 94–Apr 95, Oct 98–Nov 98, and Sep 01–Nov 01) and one in Uruguay (Dec 01–July 02). Based on the criterion that more than one-half of the banks were losing deposits, we select the same 4 previous episodes and the period Oct 97–Sep 97 in Argentina. This latter period was characterized by an important redistribution of deposits in the system.

The distinction between systemic and non-systemic deposit bank runs is established through the size of the bank run; specifically a drop in deposits of more than one standard deviation during the entire bank run period was classified as a systemic bank run. The December 94–April 95 and September 00–November 01 episodes in Argentina, and the December 01–July 02 episode in Uruguay are identified as systemic bank runs.

B. Regression Results

Argentina

Table 3 reports the results of the system GMM estimations for the different episodes in Argentina. The different between the two columns for each episode is based on the inclusion or not of the interest rate response. Robust standard errors are reported. The results can be summarized as follows:

Table 3.

Argentina: Resident Depositors’ Reaction During Bank Runs 1/

article image

This table reports Arellano-Bover regressions with robust standard errors of the change of resident deposits on bank fundamentals, deposit interest rates, exchange rate risk, and bank dummy variables. A constant and time dummies are estimated but not reported.

Due to data limitations a one month lag is used instead of three-months lag.

First, there is evidence that, even after controlling for bank characteristics, foreign branches and foreign regional banks were less affected than other banks during the Tequila bank run. This suggests that depositors perceived these foreign institutions as safe havens. However, these perceptions do not seem to have been significant during the two non-systemic bank runs, and even had the opposite effect during the 2001 systemic episode. In particular, while estimations indicate that the foreign branch dummy is positive both for the Tequila crisis and for the non-systemic bank runs, they are only statistically significant in the first case. The results also highlight that the gain in deposits by foreign subsidiaries and branches during most bank runs, as explored in the previous descriptive section, can be explained by bank characteristics in all cases for foreign subsidiaries and in many cases for foreign branches.19 However, during the 2001 systemic episode, depositors’ attitudes towards foreign banks, especially foreign branches, seem to indicate that they withdrew deposits proportionally more from foreign branches than from other banks, after controlling for bank fundamentals. This latter phenomenon could be explained in two potential ways. Branches might have facilitated the capital flight more easily than other banks, in the face of a systemic crisis. Alternatively, this result may reflect depositors anticipating the possible triggering of foreign branches’ ring fencing provisions. Unfortunately, we do not have data to explore if any of these two potential explanations are relevant.

Second, devaluation expectations played a key role in depositors’ withdrawal decisions over most analyzed bank runs.20 This evidence concurs with the literature findings in the analysis of the 2001 bank run by Levi Yeyati et al (2004) and Barajas et al (2007), and extends these findings to previous bank runs periods.

Third, bank fundamentals and bank-level variables generally seem to have the expected signs, but they are most significant in the case of the 2001 systemic bank run. It is worthwhile to highlight that public banks seem to have benefited from significant positive depositors’ perceptions during the 2001 bank run. Similarly, larger banks seem to have been perceived as too big to fail in the 1995 bank run.

Finally, the lag coefficient of the dependent variable being negative and statistically significant during the non-systemic 1998 bank run seems to indicate some level of overshooting on the adjustment on bank deposits even during the bank run period (e.g. capturing differences in interest rates beyond the average control rates used). Instead, the lag coefficient of the dependent variable is positive and statistically significant during systemic 2001 bank run. This could reflect the longer nature of the episode, and the fact that concerns about the sustainability of the currency board grew over time.

Uruguay

Table 4 presents the results for Uruguay. Most of the coefficients related to bank fundamentals have the expected sign (or are statistically insignificant). Aggregate risks related to exchange rate risk appear to have been mayor common factors driving deposit withdrawals. As before, the lag coefficient of the dependent variable indicates some level of overshooting on the adjustment on bank deposits even after controlling for interest rates.

Table 4.

Uruguay: Resident Depositors’ Reaction During 2002 Bank Run 1/

article image

This table reports Arellano-Bover regressions with robust standard errors of the percentage change of resident deposits on bank fundamentals, deposit interest rates, exchange rate risk, and bank dummy variables. A constant is estimated but not reported. Standard errors are in brackets. +, *, **, *** mean significance at 15, 10, 5 and 1% respectively. For more description of the data see Appendix Table 1.D.

The results suggest that, after controlling for banks’ fundamentals, there is no safe haven perception vis-a-vis foreign banks. The estimated coefficients for these banks are statistically insignificant, both for branches and subsidiaries.21 Coupled with the unconditional evidence presented in Figure 2, these results suggest that the larger withdrawals faced by foreign subsidiaries (relative to domestic private banks) can be explained by weaker fundamentals. Foreign branches, on the other hand, show similar performance relative to domestic private banks, both unconditionally and controlling for fundamentals.

IV. Conclusions

Our analysis indicates that, while there is a commonly held view that depositors often ‘fly to (foreign bank) quality’ in times of domestic financial distress, this is not the case in all domestic crises, including after controlling for bank fundamentals and interest rate responses. In fact, we find that only in one out of the five cases studied (Argentina, 1995), there is evidence of safe haven perceptions, and only towards foreign branches, while in one other episode (Argentina, 2001) foreign branches actually faced larger deposit withdrawals. Foreign subsidiaries do not appear to have been perceived differently in any of the cases either. Overall, the results suggest that favoring the entry of foreign branches over foreign subsidiaries may not be warranted from a safe haven perspective only.

Are Foreign Banks a 'Safe Haven'? Evidence from Past Banking Crises
Author: Gustavo Adler and Mr. Eugenio M Cerutti
  • View in gallery

    Argentina: Deposit Evolution During Bank Runsa/

    (Private resident non-financial sector’s deposits; percent change during selected periods)

  • View in gallery

    Deposit Evolution During 2002 Bank Run in Uruguay

    (Private non-financial sector deposits; percent change)