Spillovers Through Banking Centers
A Panel Data Analysis
  • 1 https://isni.org/isni/0000000404811396, International Monetary Fund

Contributor Notes

To test the role of bank lending in transmitting currency crisis we examine a panel of BIS data on bank flows to 30 emerging markets disaggregated by 11 banking centers. We find that bank exposures to a crisis country help predict bank flows in third countries after the Mexican and Asian crisis, but not after the Russian crisis. In the latter, there is evidence of a generalized outflow from emerging markets, rather than outflows linked to prior exposure to Russia.

Abstract

To test the role of bank lending in transmitting currency crisis we examine a panel of BIS data on bank flows to 30 emerging markets disaggregated by 11 banking centers. We find that bank exposures to a crisis country help predict bank flows in third countries after the Mexican and Asian crisis, but not after the Russian crisis. In the latter, there is evidence of a generalized outflow from emerging markets, rather than outflows linked to prior exposure to Russia.

I. Introduction

International banks are a major source of financing for emerging economies and also one of the most volatile ones. In Asia, for example, banks were the single largest group of creditors before the crisis and bank lending was the most variable component of capital flows during the crisis. In 1996, net flows from banks into 29 emerging markets accounted for USD120 billion, or about a third of total private inflows. In 1997 banks had reduced their share to about 9 percent of private inflows and by 1998 net inflows had turned into net outflows of about USD30 billion. 2 Understanding banks’ responses to crisis appears to be an important link in explaining the international transmission of currency crisis. 3 In the wake of the recent currency crises, observers have pointed to a number of reasons why banking centers may add to financial contagion which can be classified into two types: the first has been referred to as a “common lender” effect and the second as a “wake up call” effect.

A “common lender” effect exists if countries that share the same bank creditor become vulnerable to spillovers through this financial linkage. The idea is that banks’ responses to unexpected losses are fairly mechanistic: banks’ needs to rebalance their portfolios (due to capital adequacy and/or margin sales and/or their “Value-at Risk” models or similar models) lead to an automatic reduction of bank lending to other countries in which they hold positions. Schinasi and Smith (1999), for example, show how portfolio management rules such as “Value at Risk” tend to produce contagion when the investor is leveraged in the face of events which reduce capital. They also show that alternative rules of portfolio diversification generate the same response patterns. Contagion across assets whose returns are positively correlated can also occur in response to “volatility events”, which rather than involving actual losses involve an increase in the variance of an asset’s future return, as long as asset managers operate under loss constraint rules (such as VaR) and the risk tolerance of the portfolio manager is sufficiently high.

By contrast “wake up calls” refer to a sudden shift in perceptions for an entire asset class following an initial crisis due to reinterpretation of information and revisions of expected returns in this asset class, or due to a generalized increase in risk-aversion. This kind of response also leads to outflows from emerging markets. However, all countries are vulnerable irrespective of whether they share a common bank creditor with the primary crisis country. In this view banks react to a crisis with a generalized reduction of credit to other emerging markets. Such behavior leads to “pure contagion,” using the terminology of Masson (1998), or contagion that is not caused by mechanistic spillovers.

From a policy standpoint it is important to understand which kind of financial contagion is more relevant. Large spillovers through common bank lenders imply that emerging markets are mainly vulnerable through this channel and that they should carefully monitor the composition of their creditors. Countries might reduce contagion risk by diversifying the sources of their financing and by avoiding borrowing from creditors who are important to potential crisis countries. The policy implications are different, if, on the other hand, bank responses can be characterized as generalized wake up calls, that is as changes in flows that are unrelated to their previous exposures and to potential losses. In this case, countries’ only protection against contagion may be to lengthen the maturity structure of their debt and to rely more on foreign direct investment rather than debt financing. This latter conclusion has already been drawn in policy discussions. However, the role of the composition of lenders has so far not been stressed in the policy discourse, possibly because there was only little empirical evidence regarding the importance of this effect.

This paper attempts to explain the pattern of international bank lending during three recent crisis episodes, the Mexican, the Asian, and the Russian crisis in order to determine the role of spillovers through common bank lenders. According to the common bank creditor hypothesis, the spread of a currency crisis is caused by banks’ response to potential or actual losses in a first crisis country. The testable hypothesis therefore is whether bank flows can by explained by exposures in a first crisis country (using exposures as a proxy for potential losses).

To test this hypothesis we propose to look at disaggregated flows, by creditor and emerging market country. 4 Specifically, we examine the link between flows and exposure to the “ground zero country,” controlling for other determinants of bank flows. We calculate exposure on the eve of the Mexican, Thai, and Russian currency crises, and flows in the subsequent 6-12 month period based on the BIS semi-annual consolidated banking statistics. The emphasis on disaggregated flows is new to the literature. While the existence of a common bank lender channel in emerging market crises has been examined by a number of authors (Kaminsky and Reinhart (1998), Caramazza, Ricci, and Salgado (1999), and Van Rijckeghem and Weder (1999)), this has been done in an aggregated way, examining the effect of a proxy for competition for funds on exchange market pressure or other measures of contagion. 5 This evidence is suggestive of the existence of a common lender effect in the Asian crisis, as well as the 1982 debt crisis (Kaminsky and Reinhart, 1998). Evidence on the existence of a common lender effect based on disaggregated flows would greatly add to the confidence to be placed on these earlier findings.6

At the outset it should be said that the role of banks goes beyond what can be captured with our data for a number of reasons. First, reduced supply of bank credit could manifest itself as higher yields with unchanged flows. Thus, in theory, if prices rather than quantities adjust, we could find an insignificant effect on flows, even in the presence of a common lender effect. In practice, it is likely that there will be at least some adjustment in flows. Work by Eichengreen and Mody (1998) suggests that this is the case in the bond market, where issuance is postponed when the climate for issuance deteriorates (in their study, higher US interest rates). If a similar mechanism is in place for bank loans (so that demand for bank loans has some elasticity with respect to interest rates), a reduced supply of bank credit will have at least some impact on flows. A second issue is that banks can have indirect exposures to crisis countries, through hedge funds for example, and similarly that available data on exposures do not capture off-balance sheet positions. Under those circumstances the link between exposure and flows in the data may appear weaker than it really is, tending to reduce the significance of the results.

The paper is organized as follows. Section II describes the regional flows of international bank lending. Section III makes the case that bank losses during the three crisis episodes were sizable, and so could potentially give rise to the mechanical responses to losses described above. Section IV presents the empirical strategy and section V the results. Section VI concludes.

II. International Bank Lending Flows

This section takes a first look at the pattern of bank lending flows during crisis periods. We begin by looking at the distribution of international bank lending by regions and by major banking centers for the period covering the Asian and the Russian crises. Table 2 shows the distribution of banks’ international claims from mid-1997 to end-1998. Note that changes in these positions incorporate valuation changes (exchange rate changes, marking to market of securities, and write-downs of non-performing loans) and may differ somewhat from the true lending flows. However, as explained below, in practice these differences should not be large since only a small part of the bank portfolio should be affected. In the estimates we make an adjustment to flow data to adjust for valuation changes.

Table 1.

Net Private Capital Flows to 29 Emerging Market Economies

article image
Source: IIF (1999).
Table 2.

Distribution of International Bank Claims

article image
Source: BIS, The BIS Consolidated International Banking Statistics, Tables 1 and 2, various issues.

Figures 1-3 are extracted from Table 2 to provide an impression of the relative importance of different banking centers by region. These figures illustrate the dominating position of European banks in international lending. European banks are clearly the largest creditors in all regions. It follows that the behavior of European banks may be key in the understanding of spillovers through banking centers. The figures also show that banks tend to lend in “their” region. The majority of North American banks’ loans tends to go to Latin America and of Japanese banks’ loans to Asia. European lending is more balanced.

Figure 1.
Figure 1.

International Bank Claims on Asian Countries Outside the Reporting Area

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

Figure 2.
Figure 2.

International Bank Claims on Latin American Countries

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

Figure 3.
Figure 3.

International Bank Claims on Eastern European Countries

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

Source: Table 2.

Figures 4-6 are extracted from Table 2 and illustrate the shifts in portfolios of European, North American, and Japanese banks during the Asian and the Russian crises. Japanese banks consistently withdrew from Asia reducing their lending from USD 124 billion in mid 1997 to USD86 billion by end 1998. North American banks mainly shifted their lending among emerging markets during the Asian crisis (from Asia to Latin America and Europe) while they reduced their positions in all three regions during the Russian crisis. European banks initially, that is, after the Thai crisis, continued to build up their lending to all three regions (including Asia) 7 and only during the first half of 1998 did they reduce their holdings in Asia, while increasing them in Latin America and Eastern Europe. Finally, as was the case for US banks, European banks reduced their holdings in all three regions during the Russian crisis.

Figure 4.
Figure 4.

Change in Bank Claims for Japanese Banks During the Asian and Russian Crises, by Destination

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

Figure 5.
Figure 5.

Change in Bank Claims for North American Banks During the Asian and Russian Crises, by Destination

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

Figure 6.
Figure 6.

Change in Bank Claims for European Banks During the Asian and Russian Crises, by Destination

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

While providing an interesting overview of financial flows, it is clear that this data is too aggregated to answer the question of whether banks tended to pull out where they were most exposed to losses. We turn to country-by-country data to answer this question in section IV, after first making a case that bank exposures and losses during the recent crisis episodes were sufficiently large that banks might have wanted to rebalance their portfolios in response.

III. Bank Exposures and Losses During Recent Financial Crises

By a number of accounts international banks lost a sizable amount of money in the Asian and Russian crises. (This was true to a lesser extent in the Mexican crisis at end-1994.) In the 4 Asian crisis countries (Korea, Indonesia, Malaysia, and Thailand), exposures ranged from 20-30 percent of capital for banks from the United States, France, Germany and the United Kingdom, and 70 percent of capital in Japan. 8 Exposures to Thailand ranged from 3-5 percent of capital for the European banks and 29 percent of capital in Japan. The aggregate non-performing loan rate for the four crisis countries was expected to be about 25-30 percent.

In Russia, exposures were smaller, but expected losses greater—about 90 cents on the dollar. For European banks, the exposure of 9 selected banks was estimated at $8 billion in Russia, compared to $48 billion in the 4 Asian crisis countries. Provisions as of October 1998 were $2.3 billion in Russia and $7.1 billion in the 4 Asian crisis countries respectively. Based on market views of ultimate losses of 90 percent of exposure in Russia and 30 percent in Asia, this means losses were expected to be about half as large in Russia as in the four Asian crisis countries. German (both commercial and Landesbanken), Swiss, Austrian, French, and US banks had the largest exposures.9

Rating actions confirm that notwithstanding their large capital, major investment banks active in emerging markets were affected in the Asian and Russian crises. By way of illustration, Figures 7-9 show the rating actions (rating changes and watches) 10 Moody’s took over the period 1995-1999 for banks with a large presence in emerging markets. Box 1 describes the reasons for the downgrades and negative watches. Table 3 provides information on the total capital of the banks to illustrate the size of the capital of banks involved in emerging markets.

Figure 7.
Figure 7.

Numerical Ratings - European Banks

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

Figure 8.
Figure 8.

Numerical Ratings - U.S. and Canadian Banks

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

Figure 9.
Figure 9.

Numerical Ratings - Japanese Banks

Citation: IMF Working Papers 2000, 088; 10.5089/9781451851151.001.A001

Table 3.

Total Assets (in billion USD)

article image
Sources: For European, US, and Canadian banks, Moody’s Research Online; for Japanese banks, the figures are based on the banks’ consolidated balance sheets provided by BankStat of Thomson Financial Bankwatch, Inc.; for Chase Manhattan Bank the figure is based on the bank’s unconsolidated balance sheets provided by BankStat of Thomson Financial Bankwatch, Inc.

3 months ending March 31.

10 months ending October 31.

6 months ending June 25.

9 months ending September 30.

The picture that emerges is the following. During the Mexican crisis episode, the downgrades and watches do not appear to have been related to the Mexican crisis. In the Asian crisis episode, Moody’s put a large number of banks on watch or downgraded them. For the Japanese banks, the majority of banks were put on watch or downgraded, for domestic reasons but often also with a mention of the impact of the Asian crisis (Box 1). Among European and North American banks, Commerzbank, Societe Generale, Credit Lyonnais, Standard Chartered, JP Morgan, and Royal Bank of Canada were put on negative watch during the period July 1997-July 1998 for reasons related to the impact of the Asian crisis (Box 1). Finally, in the Russian crisis episode, a number of European and US banks were put on negative watch or downgraded, for reasons related to the Russian crisis (CSFB, Deutsche, JP Morgan, Bankers’ Trust, and Republic New York). Downgrades for Japanese banks were not related to developments in Russia or emerging markets.

Note that a common bank lender effect could be present even if banks do not immediately suffer losses in the primary crisis country since loss of capital (a “capital event”) is only one reason why banks may choose to rebalance portfolios, another reason being a “volatility event” (involving an increase in the variance of an asset’s future return rather than actual losses). As noted in the introduction, asset managers who operate under loss constraint rules (such as VaR) will under certain circumstances sell an asset whose return is positively correlated with their assets in a primary crisis country hit by a volatility event, thereby contributing to contagion across emerging markets. Thus, exposure to an initial crisis country could still give rise to a common bank lender effect even when not manifesting itself in actual losses.

Overall the above review of information on exposures, losses, and rating actions suggests that the Asian and Russian crisis episodes were “capital events”, giving one reason to believe that bank flows would have been affected in these episodes. No equivalent loss of capital appears to have been present in the Mexican crisis, but if this crisis is interpreted as a volatility event, an effect on bank flows might also be expected. We turn to an empirical investigation of whether this was in fact the case next.

IV. Empirical Strategy

Our aim is to explain the pattern of outflows (and inflows) of bank lending during a crises period. Since we are interested in financial contagion, we omit the first crisis country (Mexico, Thailand, Russia) and only study the reaction of bank flows in the other countries. We estimate the following equation for each crisis episode:

ΔExposureci/Exposurec=α+β(Exposurec0,/Exposurec)+γ(Exposureci/Exposurec)+φMacro-Controlsi+δTradei+ε

where 0 stands for the ground zero country, 11 c stands for the common creditor (11 banking centers), and i indexes the receiving country. Exposure ci represents bank flows from a creditor country c to an emerging market i, Exposurec is the total exposure of a bank creditor c to developing countries as a whole (including Eastern Europe), and Exposure c0 is exposure of a bank creditor c to the ground zero country. ΔExposure ci is the flow of bank lending during the crises period. 12

For example (Exposure c0,/Exposurec) could refer to German banks’ lending to Thailand as a share of total lending of German banks to developing countries. This is used as a proxy for the exposure to loss that German banks face in the event of a crisis in Thailand. A significant β, the coefficient on Exposures c0/Exposure c, is evidence in favor of a common lender effect.

A significant γ points to the presence of generalized inflows or outflows proportional to initial exposure, as one would expect to find when there is a general shift in investor’s attitudes towards investing in emerging markets.

Macro-controls i are a set of macroeconomic variables that have been identified in the crisis literature13 and should in principle determine bank flows to the extent that banks use these criteria in their lending decision (current account/GDP, budget deficit/GDP, M2/Reserves, growth of credit to the private sector, and real exchange rate appreciation14). 15 Trade linkages are captured in two ways—as direct trade (calculated as the percent of total exports destined for the ground zero country) 16 and as trade competition in third markets (the trade share index of Glick and Rose (1999)). 17

We also examine a number of variants. In a first variant, we use changes in flows as the dependent variable. In a second variant, we adjust the data to control for the fact that flows could be a statistical artifact, reflecting valuation changes. In a third variant, we include the reserves to short-term debt ratio among the macro-controls.18 In a fourth variant, we introduce liquidity as an additional control variable. We use the JP Morgan liquidity measure in the month of the crisis in the ground zero country. The effect of liquidity could go either way. On the one hand, banks could try to sell those securities with low bid-ask ratio as this would minimize losses from “firesales”; on the other hand, in periods of tight liquidity, banks might prefer to exit markets with low liquidity, in order to remain liquid themselves. When JP Morgan provides a liquidity rating for more than 1 Brady or Eurobond, we use the highest (most liquid) rating.

Bank exposures refer to the positions of banks on the eve of the respective crisis episodes (December 1994 for Mexico, June 1997 for Thailand, and June 1998 for Russia). For flows we use the 6-month flows subsequent to these dates, with the exception of the Thai crisis where we use flows for the entire subsequent year (i.e. June 1997-June 1998). We use the BIS’ semi-annual consolidated data covering banking systems in 11 industrialized countries (the “reporting area”).19

Market participants have been skeptical of the usefulness of the BIS data, pointing out that it captures only on-balance sheet positions, whereas banks typically hedge their positions with off-balance sheet positions. Maintaining such hedges is nevertheless expensive, and hence tends to be done more when a crisis is widely anticipated, as was the case in Brazil. For the Mexican, Asian, and Russian crises, which were generally not anticipated, the data is more likely to capture overall positions closely. 20 A second caveat is that indirect exposures are not covered by the data. To the extent that commercial and investment banks maintain sizeable exposures to other commercial entities which invested heavily in ground zero countries, this means that the data misses indirect exposures of banks to ground zero countries. In practice, this appears to have been important only during the Russian crisis, on account of high exposures of hedge funds to Russia. 21 Third, exposure data do not capture the effect of any off-setting guarantees. Because this is known to be important for the largest common creditor in the Russian crisis—Germany—this country is omitted from the Russian crisis regressions. 22 Finally, bank claims reflect exchange rate changes and write-downs and marking to market of securities. Hence the BIS data will point to changes in exposures without physical flows having occurred. With a depreciation in a borrowing country, for example, inflows will appear smaller than physical inflows; similarly, outflows will appear larger. Below, we attempt to adjust the data for this bias using an exchange market pressure index. Because of gaps in the availability of the data, the data is organized into an unbalanced panel. It covers the 30 main emerging markets.

V. Regression Results

Table 4 provides the results based on a panel of data on bank flows to each emerging market disaggregated by 11 creditors,23 for a subset of 30 emerging markets.24 Note that it is the availability of data by creditor which yields the panel dimension, not time, as the regressions are run for a point in time (i.e. separate regressions are run for the Mexican, Asian, and Russian crises). The flow from a given creditor to a given emerging market (the dependent variable) is scaled by the creditor’s total claims on emerging markets. The common lender effect is tested by including creditor country exposure to the ground zero country (scaled by the creditor’s total claims on emerging markets) as an independent variable. The creditor’s claims on an emerging market (again scaled by its total claims on emerging markets) is introduced as an independent variable, to test whether inflows and outflows are proportional to exposure (generalized inflows and outflows). Two types of regressions are run. In the first, trade competition and macro-controls, which vary across but not within countries, are included. In the second, which corresponds to fixed effects, country-dummies (for recipient countries) replace these control variables. 25

Table 4.

Disaggregated Contagion Indicators Coefficients and T-Statistics of OLS estimates Dependent Variable: Flows by Emerging Market (i) by Creditor (c) 1/

article image
Bold entries have the expected sign and are significant at the 10 percent level or better.

As a percent of creditor j total exposure in emerging markets.

2/ Trade based on direct trade in Mexico and Russia; based on shares in Thailand.

The results point to the existence of a common lender effect in the Mexican 1994 and Asian 1997 crises, but not the Russian crisis. For the Mexican crisis, the results point to a small common lender effect which is significant at the 5 or 10 percent level of significance, depending on whether fixed effects or macro-controls are used. For each 1 dollar additional exposure to Mexico, flows are lower by 1 cent on average for any given emerging market, holding constant exposure to that given country and macro-controls. At the same time, the data point towards a generalized inflow of funds in the wake of the Mexican crisis. For each dollar of exposure to the country under consideration, flows increase by on average 13-14 cents to that country, and this effect is statistically significant. That is, the “own-country” effect is stronger than that connected to Mexico. Among macro-controls, only private sector credit growth and the current account are statistically significant and bear the anticipated sign. Trade competition is not significant in this regression, but this is not conclusive as to the role of trade competition in contagion, as the regression at hand captures only the effect on bank flows. 26

For the Asian crisis, the effect is economically significant as well as statistically significant. For each additional dollar in exposure to Thailand, flows per emerging market fall by 4 cents, on average, everything else constant. To illustrate the magnitudes involved, consider the case of Japanese banks. Exposures to Thailand were 25.5 percent of Japanese banks’ total exposure to emerging markets. This meant, according to the regression results, that Japan would have reduced its exposure (holding everything else constant) by about 1 percent (25.5*0.04) of its total emerging market exposure, on average, in each of the emerging markets where it invests. Summing over the 30 emerging markets in our regressions, this amounts to 30 percent of initial exposure to emerging markets, a very sizeable figure. 27 The coefficient on initial exposure to the country under consideration is insignificantly different from zero, indicating that there is neither a generalized inflow or outflow of funds. The real exchange rate and trade competition are statistically significant and of the correct sign. 28

Turning to the case of the Russian crisis, the results point to a generalized outflow, of some 8-9 percent of initial exposures, which is highly significant statistically. The common lender effect is not statistically significant. M2 over reserves, growth in credit to the private sector, and the real exchange rate are statistically significant and of the correct sign. 29 From this it appears possible that contagion from the Russian crisis was generalized, reflecting an increase in perceived risk or in risk aversion. Alternatively, the lack of significance on the common lender variable could be related to the limitations of the data, and we speculate how this might have worked in the conclusion.

As noted above, BIS flow data (expressed in USD) incorporate exchange rate changes and write-downs and hence do not represent true changes in volumes. This is not likely to be important in practice in most cases, because bank claims are usually expressed in foreign exchange rather than local currency (so that devaluations in borrowing countries have no effect) and because write-downs tend to be limited to the securities portfolio which has to be marked to market, with non-performing loans slower to be acknowledged. We make a rough adjustment to the BIS flow data to adjust for these factors, and check whether this makes a difference to the results. Because we do not know the currency denomination of foreign-exchange denominated loans (i.e. USD, Yen, DM, etc.), we cannot adjust for changes in cross-rates. Specifically, the adjustment we make is as follows. Let E0 and E1 denote original and final exposure. Then unadjusted flows (used in the above regressions) are equal to E1-E0. Now add back the effect of exchange rate changes and write-downs to final exposure by assuming that this effect is proportional to the product of exchange market pressure and original exposure. This effect equals this product when all of the portfolio is marked-to-market (as tends to be the case for securities or when loan loss provisioning is 100 percent) and/or when bank claims are in local currency.

Our guesstimate for the fraction of bank portfolios which is marked-to-market or expressed in local currency terms is 20 percent. Then adjusted flows equal E1-E0+0.2*pressure*E0. This guesstimate is based on BIS data on the share of debt securities in total claims30 and a selective examination of the share of (a proxy for) local currency denominated bank claims. 31 The pressure index used in this equation is an equally weighted average of percent changes in the exchange rate, reserves, and interest rates. We use the pressure index 6 months after the Mexican and Russian crises and 12 months after the Thai crisis. 32 Table 5, top panel, shows that the results are largely unchanged. We still find a significant common creditor effect in the Mexican and Asian crises and not during the Russian crisis.

Table 5.

Disaggregated Contagion Indicators Robustness Tests Coefficients and T-Statistics of OLS Estimates Dependent Variable: Adjusted Flows or Change in Flows 1/

article image
Bold entries have the expected sign and are significant at the 10 percent level or better.

As a percent of creditor j total exposure in emerging markets.

Macro- and trade-controls are included in the regression but the results are suppressed.

We report only the results based on fixed effects, because of the higher overall explanatory power; results concerning exposures are unchanged in a specification with macro-controls instead of fixed effects.

We also examine the change in flows. This helps us capture contagion which manifests itself as a decline in trend flows, rather than as an outflow. For example, flows declined to Argentina and Mexico in the wake of the Russian crisis, but remained positive.

We use the average flow during 1994 as a benchmark for flows after the Mexican crisis, and during June 95-June 97 as a benchmark for flows after the Thai and Russian crises. Table 5, bottom panel, shows that the results are similar to those obtained above and point to a strong common lender effect on the change in flows in the Asian crisis, but no such effect in the wake of the Mexican and Russian crises.

In Table 6, top panel, we add reserves to short-term debt as an additional control variable (in addition to macro- and trade-controls). This variable is never significant. In the bottom panel we add JP Morgan’s measure of market liquidity as an additional control variable. The effect of liquidity appears to have been different across crises. In the Mexican crisis greater liquidity means larger outflows; in the Asian and Russian crises, on the other hand, greater liquidity means smaller outflows. These results should be regarded as tentative, however, given that the securities portfolio (to which the liquidity measure pertains) is small compared to the loan book.

Table 6.

Disaggregated Contagion Indicator Coefficients and T-Statistics of OLS Estimates Adding Short-Term Debt and Liquidity as Independent Variables

article image
Bold entries have the expected sign and are significant at the 10 percent level or better.

Macro- and trade-controls arc included in the regression but the results are suppressed.

As a percent of creditor j total exposure in emerging markets.

In closing, it should be noted that the results in general suffer from a lack of explanatory power, with only about 20-30 percent of the variance of bank flows explained at best. This may be due to omitted variables, such as expectations about the future course of policy and of other market participants’ actions. The lack of explanatory power should not take away from the conclusions about the role of the common lender effect, however, as long as the correlation with the omitted variables is not too large.

VI. Conclusion

This paper has provided evidence from flow data of bank lending that supports the view that spillovers through common bank lenders were important in transmitting the recent Mexican and Thai financial crises. Regressions based on panel data for 11 creditor countries and 30 emerging markets point to a large and statistically significant common lender effect during the Thai crisis. The effect is somewhat smaller in the Mexican crisis and not statistically significant in the Russian crisis.

The small impact during the Mexican crisis is consistent with the lack of impact which the Mexican crisis appears to have had on developed country bank capital. In the Russian crisis, the withdrawal of funds seems to have been more generalized, pointing to the role of “wake-up calls” concerning emerging markets or a general increase in risk-aversion. Still, the absence of a common lender effect goes contrary to the widely held view of bank behavior. It could reflect the absence of some major players from the data (the BIS data exclude data on Swiss banks) or the existence of indirect exposures and guarantees not captured by the data. An alternative explanation would be that banks manipulated their off-balance sheet positions to cut their exposures, an effect which is not captured by the BIS data. Finally, because pressures to withdraw funds can appear in either quantities (flows) or prices (yields), spillovers through common bank lenders may be present even when they are not captured by the flow data. In fact, in Van Rijckeghem and Weder (1999), where we do not rely on flow data but use measures of contagion such as the exchange market pressure index, we previously found some evidence in favor of a significant common bank lender effect even in the Russian crisis. From a policy point of view these findings imply that emerging market economies could reduce their contagion risk by diversifying the sources of their funding and carefully monitoring their vulnerability through shared bank creditors. 33

Reasons for Negative Watches/Downgrades in 1995 and After Mid-1997

European Banks

article image
article image
article image
article image
article image
article image
article image
article image
article image
article image
article image

U.S. and Canadian Banks

article image
article image
article image
article image
article image

Japanese Banks

article image
article image
article image
article image
article image
article image
article image
Source: Moody’s Research Online.
Spillovers Through Banking Centers: A Panel Data Analysis
Author: International Monetary Fund
  • View in gallery

    International Bank Claims on Asian Countries Outside the Reporting Area

  • View in gallery

    International Bank Claims on Latin American Countries

  • View in gallery

    International Bank Claims on Eastern European Countries

  • View in gallery

    Change in Bank Claims for Japanese Banks During the Asian and Russian Crises, by Destination

  • View in gallery

    Change in Bank Claims for North American Banks During the Asian and Russian Crises, by Destination

  • View in gallery

    Change in Bank Claims for European Banks During the Asian and Russian Crises, by Destination

  • View in gallery

    Numerical Ratings - European Banks

  • View in gallery

    Numerical Ratings - U.S. and Canadian Banks

  • View in gallery

    Numerical Ratings - Japanese Banks