Chapter

8 Towards the Effective Supervision of Partially Dollarized Banking Systems

Author(s):
Adrián Armas, Eduardo Levy Yeyati, and Alain Ize
Published Date:
July 2006
Share
  • ShareShare
Show Summary Details
Author(s)
Jorge Cayazzo, Antonio Garcia Pascual, Eva Gutierrez and Socorro Heysen1 

8.1 Introduction

The purpose of this chapter is to contribute to the design of a prudential regulatory framework for banks operating in partially dollarized economies. The discussion is anchored on the Basel Committee on Banking Supervision (BCBS) guidance on risk management, as expressed in the 1998 Capital Accord and the subsequent Basel II revision. While BCBS recommendations are in principle applicable to all banking systems, the risks that highly dollarized financial systems face are so specific and so large that they require special attention by both the financial institutions and the supervisors. The resulting exposures create substantial systemic vulnerabilities to which, from the standpoint of financial stability, supervisory regimes need to adapt.2

Partial dollarization increases the vulnerability of financial systems to both solvency risk and liquidity risk.3 Solvency risk results mainly from foreign currency mismatches in the event of large movements of the exchange rate. Liquidity risk stems from the limited backing of banks’ dollar liabilities, and is often associated with (or triggered by) solvency risk.

Following international standards, partially dollarized countries control banks’ currency risks by imposing limits or minimum capital requirements on foreign exchange exposures. While international standards provide an adequate framework for countries with significant exposure to foreign currency, one aspect that deserves special consideration in countries with a high level of dollarization is the definition, for prudential purposes, of a riskless open foreign exchange position. We show that a matched foreign exchange position, according to the conventional definition, is not riskless in a highly dollarized country. Instead, the definition should be adapted to reflect dollarization.

Banks’ attempts to contain foreign exchange risk often lead them to increase their exposure to currency-induced credit risk. They match their currency positions by granting foreign currency loans to domestic clients whose cash flow is in domestic currency. While effectively transferring the foreign exchange risk to the borrowers, banks retain the credit risk resulting from the possibility that the borrowers’ currency mismatch affects their capacity to repay the loan in the event of large adverse exchange rate fluctuations. Exposure to credit risk also increases if the value of the collateral – denominated in domestic currency – declines in relation to the value of the dollar loan obligation it is backing, following a currency depreciation.

Implicit or explicit government guarantees distort pricing decisions and enhance the attractiveness of foreign exchange-denominated contracts. Borrowers (and banks), operating in the context of fixed exchange rate or ‘fear of floating’4 regimes, expect the exchange rate risk to remain limited within the horizon set by the maturity profile of the loans. This encourages short-term lending and limits intermediation spreads in foreign currency. As a consequence, borrowers (and banks) underprice the risks associated with holding a currency mismatch in their balance sheets. In view of the higher spreads and often volatile interest rates associated with loans in domestic currency, they prefer therefore to intermediate in foreign currency. Implicit or explicit guarantees on dollar deposits or expectations of generalized government bail-outs in the event of a large depreciation further encourage foreign currency lending and borrowing.5 The limited availability of hedging instruments in many emerging markets and the shallowness of domestic credit markets also provide a rationale for unhedged foreign currency lending. The underpricing of risk induces banks to hold insufficient reserves – in the form of provisions or capital – as a protection against the risks associated with large exchange rate fluctuations. This is a problem that bank supervisors need to address.

The limited backing of banks’ foreign currency liabilities by liquid foreign currency assets and their convertibility at par create systemic liquidity risk. When the demand for local assets falls, due to a perceived increase in country risk or banking risk, depositors’ attempt to convert their deposits into foreign currency cash or transfer them abroad, and/or foreign banks’ attempt to recall short-term lines of credit thereby creating liquidity pressures on the banks. Unless there are sufficient liquid foreign currency assets to back liquid foreign currency liabilities, banks may run out of foreign currency liquid reserves and become unable to pay off their foreign currency liabilities coming due. Similarly, central banks may run out of international reserves to provide foreign currency lender of last resort support to distressed banks. When this happens, deposit (or loan) contracts may need to be broken and disruptive or confiscatory measures taken, thereby validating creditors’ fears and justifying the run. These systemic risks are often overlooked by banks, who prefer that the cost of holding additional liquid assets be borne by central banks.6

The combination of risk underpricing and insufficient prudential buffers can have serious consequences in the event of large unexpected shocks, not only for individual banks but for the financial system as a whole. Given that the primary responsibility for bank solvency and liquidity must rest on its shareholders and management, the basic prudential objective is to ensure that these risks are internalized appropriately. To this end, countries should fully implement the Basel guidelines on the management of risks, paying special attention to the specific vulnerabilities that arise in a dollarized environment. However, qualitative guidance encouraging banks to adequately manage risks, while necessary, is unlikely to be sufficient. Supervisors need to ensure that banks’ buffers, in the form of capital requirements, provisioning requirements, or liquid asset requirements, are sufficient to cover shocks that may arise within acceptable risk-tolerance levels. The regulatory framework governing liquidity in a dollarized economy should also take into account that liquidity risk is twofold: (i) individual bank risk due to an isolated run on the bank’s liabilities; and (ii) systemic risk linked with a generalized run on banks.

This chapter is organized as follows. Section 8.2 describes current supervisory practices, showing that several partially dollarized countries have adopted measures to reduce vulnerabilities from foreign exchange and liquidity risks, but few have addressed those arising from currency-induced credit risks. Shortcomings arise from both the lack of full implementation of the Basel guidelines on risk management and the absence of sufficient prudential buffers. Section 8.3 presents a general framework to reduce the vulnerabilities of partially dollarized economies. The framework suggests how to interpret Basel guidelines on risk management, and assess the nature and size of the required prudential buffers. Section 8.4 concludes by discussing some implementation issues. This chapter does not address issues related to the causes of (or solutions to) dollarization, that fall mainly outside the purview of prudential policies and are discussed elsewhere in this volume.7

8.2 Current supervisory practice

Current supervisory practices in most partially dollarized countries are not effectively addressing the vulnerabilities that are characteristic of dollarized environments. This shortcoming stems from two sources. First, many highly dollarized countries fall short of fully implementing the BCBS guidelines for the management of key risks.8 Second, few supervisors have taken provisions to ensure that adequate prudential buffers to cover these risks are in place. While many countries have implemented measures to achieve adequate protection from foreign exchange and liquidity risks, few have sought to ensure adequate protection against currency-induced credit risk. These observations are based on the results of a survey conducted between June and September 2004 in seventeen countries in different parts of the world and with different levels of dollarization (see Appendix 8.1 for details).9

All the surveyed countries have implemented prudential regulations based on current international standards for controlling foreign exchange risks. Regulations include limits on (or capital requirements against) foreign exchange exposure. Many countries have switched from limits on foreign exchange open positions to capital requirements against these positions; several have adopted both. In some cases, these regulations entail a structural open position (as in Lebanon) or asymmetric limits on open positions that allow relatively high long open positions (as in Bolivia and Peru). As is shown in the next section, in highly dollarized countries, asymmetric limits and structural open positions are generally preferable. However, to fully protect banks’ solvency, regulators would gain from redefining what a riskless foreign exchange position is.

Most countries have also implemented prudential measures to reduce the vulnerabilities of financial systems to liquidity risk. Following BCBS guidelines, banks are required to manage this risk, to conduct stress tests on a variety of scenarios and to implement contingency plans to address liquidity problems. These scenarios are generally based on the estimation of maturity gaps, in many cases currency specific. A handful of countries have introduced limits on maturity mismatches. In addition, many highly dollarized countries utilize a combination of prudential measures, mostly minimum liquidity ratios and reserve requirements, to ensure that banks, and the banking system as a whole, have an adequate buffer of liquid assets to face stressful conditions. These minimum requirements tend to be relatively high, with rates ranging from 10 to 40 per cent, and may include high marginal rates for some liabilities. In many countries, liquidity requirements are higher for foreign currency liabilities than for domestic currency liabilities. In some countries (Croatia, Honduras and Slovenia), liquidity requirements only apply to foreign currency liabilities.

In contrast, the extent to which regulatory frameworks deal with currency-induced risk is limited and very recent. Two-thirds of the surveyed countries with moderate to high dollarization have not required banks to manage their currency-induced credit risk, nor conducted stress tests that allow them to identify the relevance of this risk for their banks. The remaining one-third, plus two formerly dollarized countries (Argentina and Poland), have generally focused their efforts on requiring that banks manage their currency-induced credit risk and measure it through stress testing. Few have achieved progress towards ensuring that these risks are adequately priced and covered with a sufficient buffer. Among the surveyed countries, only Uruguay has recently required higher capital for foreign currency assets10 and only Peru requires higher provisions for foreign currency loans. Besides Peru, five other countries (Singapore, Poland, Lebanon, Argentina and Chile) report that banks are expected to assign a higher risk rating to debtors whose capacity to repay is sensitive to exchange rate movements. Interestingly enough, several highly dollarized countries have recognized that the 8 per cent minimum capital standard does not provide sufficient cover for credit risk, and have implemented higher requirements. The most notable cases are Lebanon and Romania, which require a 12 per cent minimum capital adequacy ratio.

It is not coincidental that some countries with low dollarization levels (Argentina, Brazil and Chile) have taken administrative measures to control currency-induced credit risk, whereas highly dollarized countries have avoided them. The effectiveness of limits or prohibitions on foreign currency credit to unhedged borrowers is likely to be lower (and the costs higher) in countries that are already dollarized. Once the use of the dollar has become well established, strong prudential measures to contain the risks are more likely to induce financial disintermediation and regulatory arbitrage. Thus, it is generally easier to take preventive action when dollarization is low. While there is no easy fix to this problem of currency-induced credit risk, it is encouraging that some countries are taking the right steps anyway.

8.3 Towards good practices

The tendency to underprice risks warrants a proactive approach to prudential regulation and supervision. This approach must consider two key elements:

  • Risk-based supervision needs to be implemented along the lines of the Basel Core Principles and Guidelines for Effective Supervision, taking into account the implications of operating under highly dollarized environments. While the responsibility for managing risk lies with the banks, supervisors can induce them to better manage their risk by setting high standards for risk management. In a dollarized environment, this implies that the supervisors should seek to ensure that banks adequately manage all their risks, including currency-induced credit risk and systemic liquidity risks. These aspects are frequently overlooked.

  • Supervisors should ensure that banks have adequate buffers to protect their solvency and liquidity, including against large shocks that could occur within acceptable risk-tolerance levels. These large shocks would be left largely uncovered if unregulated and could have serious consequences, not only for individual banks but for the banking system as a whole and, thus, for financial stability. Determination of the size of the buffers should result from an assessment of the size of the shocks that could occur, and their impact on bank solvency and liquidity. The goal is to compensate for the underlying distortions that lead to the underpricing of risk, as close as possible to their source, and induce agents to better internalize and price the risks of operating in a dollarized environment. Minimum capital and provisioning requirements should be used to protect banks’ solvency against currency-induced credit risk. Minimum liquidity requirements are recommended as a buffer against systemic liquidity risk.

While the framework presented below is consistent with international standards,11 the discussion expands beyond the level of detail generally covered in these standards. In some instances, our view is that full implementation of current international standards, as reflected in Basel I, would not suffice to adequately address the vulnerabilities of a dollarized banking system. In some other cases, countries may face restrictions that prevent them from addressing their vulnerabilities while at the same time adhering to international standards.

Foreign exchange risk

International standards to manage and control market risk provide an adequate framework for countries with significant exposure in foreign currency. These standards stress that bank supervisors must be satisfied that banks have in place systems that adequately measure, monitor and control market risks. In addition, supervisors should have powers to impose specific limits and/or a specific capital charge on market risk exposures, including on foreign exchange business, if warranted.

The determination of what constitutes a ‘risk-free position’ is necessary before deciding whether foreign exchange risks are to be priced and/or limited. Traditionally a net currency position is measured as assets minus liabilities for each currency. A position in which assets and liabilities are matched (net open position = 0) is considered to be risk-free. However, while a matched foreign open position will protect the bank’s capital, expressed in domestic currency, it does not necessarily protect its capital adequacy ratio (CAR). In fact, in highly dollarized systems, exchange rate fluctuations can significantly affect the CAR of banks with a perfectly matched foreign open position. As shown in Box 8.1, the larger the difference between the dollarization ratio and the foreign exchange position as a percentage of capital, the higher the impact on the CAR. For example, a 20 per cent depreciation would reduce the CAR of a bank with a 67 per cent dollarization ratio and a matched open position (as conventionally defined) from 10 per cent to 8.8 per cent. Instead, should the bank maintain a foreign open position equivalent to its rate of dollarization, it would maintain its CAR at the 10 per cent initial level.

This problem is mitigated, but not fully resolved, by the structural position allowed by the Basel Committee.12 Banks may be allowed to protect their CAR by excluding from their net open position any position they have deliberately taken to hedge, partially or totally, against exchange rate fluctuations. However, three conditions have to be met: (i) the positions have to be of a non-dealing nature; (ii) the position does no more than protect the bank’s CAR; and (iii) any exclusion needs to be applied consistently during the life of the asset.

It would be desirable to have a more direct approach aimed at protecting banks’ CAR. Supervisors should centre the calculation of capital charges (and/or limits) on the basis of the actual level of dollarization of each bank. Under this approach, banks are required to fully hedge their CAR against exchange rate movements, i.e., the structural position is not a choice. For this approach to be effective, gains in the open position should be tax exempt. Thus, should a bank have 30 per cent of its assets in foreign currency, and supervisors establish a limit of 20 per cent of risk-tolerance, the corresponding limit for this bank would be a band of 20 per cent both below and above its current dollarization level, resulting in an open position not below 10 per cent of capital, nor above 50 per cent of capital (see Box 8.1).

Box 8.1How does a devaluation affect the CAR of a bank, depending on its foreign exchange position and asset dollarization?

The example below is a simple way to illustrate how the capital adequacy ratio of a bank is affected by a depreciation/appreciation of the domestic currency, if its foreign exchange open position, as a proportion of capital, differs from the level of dollarization of its assets. Assume the following initial situation:

Bank A (matched foreign open position; 67 per cent dollarization of assets)
FX assets = 200FX liabilities = 200
LC assets = 100LC liabilities = 70
Capital = 30
FX open position = 0 (200-200)
CAR = 10% (for simplicity, assume that all assets weight 100 per cent for capital adequacy purposes)
Bank B (mismatched, set at the level of dollarization of assets; 67 per cent dollarization of assets)
FX assets = 200FX liabilities = 180
LC assets = 100LC liabilities = 90
Capital = 30
FX open position = 20, equivalent to 67 per cent of capital
CAR = 10%
What would happen after a 20 per cent depreciation of the domestic currency?
Bank A (matched)
FX assets = 240FX liabilities = 240
LC assets = 100LC liabilities = 70
Capital = 30
CAR = 8.8%
Bank B (mismatched)
FX assets = 240FX liabilities = 216
LC assets = 100LC liabilities = 90
Capital = 34
CAR = 10%

As shown above, high dollarization makes it necessary to carefully assess the convenience of having regulatory requirements in which limits or capital charges are based upon the belief that a matched net position is risk free.

Credit risk

As explained above, unexpected exchange rate movements generate currency-induced credit risk that, in a partially dollarized banking system, tends to be highly underpriced. Most of the time, banks’ and borrowers’ expectations are validated by the behaviour of the central bank, which generally resists depreciations due to concerns about the financial stress these would cause.13 Thus, many inadequately hedged foreign currency loans are granted and banks tend to hold insufficient reserves to protect themselves from large depreciations. To address this problem, risk needs to be internalized and adequate prudential buffers introduced, in the form of capital or provisioning requirements. Generally, capital covers unexpected losses, while provisions cover expected losses, either identified (covered by specific provisions) or latent losses not yet identified (covered by general provisions). Alternative options are also discussed.

Supervision of credit risk

International standards provide a sound basis to ensure that financial institutions, including those in dollarized economies, manage their credit risk properly. BCBS ‘Principles for the Management of Credit Risk’ (1999) state that banks should operate under sound, well-defined lending criteria, which include a thorough understanding of the borrower, as well as the purpose and structure of the credit and its source of repayment (Principle 4). Banks should also have information systems and analytical techniques that allow them to measure the credit risk inherent in all activities (Principle 11). Additionally, banks should consider the potential impact of future changes in economic conditions so as to assess individual credits and credit portfolios under both normal and stressful conditions (Principle 13). Moreover, the contingent nature of market-sensitive exposures, such as foreign exchange contracts, requires that banks have the ability to assess the probability distribution of the size of actual exposure in the future, and its impact on both the borrower’s and the bank’s leverage and liquidity.

In countries with significant lending in foreign currencies, these principles imply that banks should pay particular attention to currency-induced credit risk. Issuing a specific regulation or guidelines on such a risk could be of great assistance to banks. The regulation should establish disclosure requirements and minimum standards, and define the responsibilities of directors and managers as regards risk management. Building up the supervisory capacity to assess the adequacy of banks’ credit risk management and exercising supervisory powers to ensure compliance are essential.

Supervisors should ensure that banks gather sufficient information to measure the sensitivity of a borrower’s capacity to repay under changing exchange rates. For all relevant exposures, banks should be required to gather information on borrowers’ cash flows and balance sheets, with a breakdown by currency. Even when a bank lends only in domestic currency, its borrowers’ capacity to repay could be hampered, in the event of a depreciation, if the borrower has significant liabilities in foreign currency with other counterparts. Credit bureaus could also be encouraged to provide currency-specific information on all debts, highlighting – if relevant – debts indexed to the domestic price index or the exchange rate.

Separating by currency assets and liabilities, as well as revenues and expenditures, is not a trivial exercise; banks need to have a deep knowledge of their customers and the markets where they operate. For instance, an importer who prices his products in dollars may see his capacity to repay hindered by a depreciation if this causes a contraction in the demand for his products. An exporter who sells in the domestic market could also be negatively affected by a depreciation. Inversely, a worker who earns domestic currency, but receives remittances from abroad, could maintain his capacity to repay in the event of a depreciation, particularly if the remittances come through the bank and are defined as the source of repayment.

While the exchange rate is expected to be the main source of indirect credit risk in dollarized financial systems, the significance of changes in interest rates should not be overlooked. Borrowers with domestic currency loans and floating interest rates, or with a maturity mismatch, could be an added source of credit risk, especially under fixed exchange rate regimes or fear of floating. In such cases, domestic currency interest rates tend to be volatile, affecting the borrower’s capacity to repay and the bank’s exposure to credit risk.

Supervisors should make sure that banks’ policies and procedures consider their exposure to currency-induced credit risk. Banks should define internal policies that could include limits on these exposures and, in countries where these exposures are already high, target a progressive reduction in their exposure. Risk can be limited by reducing banks’ exposure to mismatched borrowers or by encouraging borrowers to reduce their currency mismatches. Offering hedging products or charging interest rates that better reflect currency-induced credit risk could be useful in this regard. Banks should also assess their loan-pricing policies to ensure that they adequately reflect overall credit risk. Special attention could be given to setting policies regarding foreign currency loans to individuals (for consumer goods or housing).

The assessment of borrowers’ capacity to repay their loans in the event of changes in exchange rates or interest rates can be done individually for the large borrowers, using information on currency-specific cash flows and balance sheets. For smaller borrowers with homogeneous characteristics, such as individuals or small firms, the assessment can be done on a group basis. When the information on currency-specific cash flows and balance sheets is not available, it would be prudent to assume that all the income of these borrowers is denominated in domestic currency.

The selection of appropriate scenarios and assumptions for stress testing is essential in order to properly assess the credit risk embedded in the bank’s portfolio. Supervisors could let banks define the parameters for these changes (exchange rate, interest rate, output) and assess their adequacy at the time of the supervisory review. Alternatively, supervisors may provide the parameters, thereby making the results of the stress tests more comparable and easier to review. To minimize any possible confusion in the interpretation of these parameters, the supervisory authorities should clearly explain the rationale and methodology on which they are based.

Banks should be able to calculate the expected losses from their loan portfolio, including those that arise from borrowers’ currency, interest rate or maturity mismatches. These will depend on the borrowers’ probability of default, on the bank’s exposure at the time of default and on the bank’s loss-given default. While it may not be realistic to expect that all banks in all countries will have the capability to estimate default probabilities, assessing the capacity to repay should provide a basis for estimating expected losses. In turn, to estimate loss-given default, banks should be able to assess the recovery value of collateral, which depends on the market value of collateral and on the costs of recovery (taking into account the time it takes to foreclose and fully recover the collateral).

As the recovery value of collateral can also be affected by market conditions, banks should assess this value for changing market conditions. The limited availability of information on most asset prices is a difficulty that would need to be overcome in most countries. When collateral is difficult to recover, highly illiquid, or its price cannot be estimated, it may be advisable to disregard such collateral. In other cases, frequent appraisals of relevant collateral are an option, albeit costly, to obtain this information. Alternatively, previously appraised collateral values could be adjusted frequently, using simplified methods to be reviewed by supervisors. For instance, the foreign currency value of real estate collateral could be expected to decline in the event of a depreciation. In the absence of better information, banks could assume that real estate prices are set in domestic currency, and thus their fall would be proportional to the depreciation. This method would probably provide a worst-case scenario.

Banks should disclose their policies on credit-risk management, and their main exposures, including as regards currency-induced credit risk. Such a disclosure should be required at least annually, as part of the bank’s audited financial statements. The availability of this information should help foster market discipline, as the most sophisticated creditors would penalize banks that take higher risks, increasing their costs of funding, and thus encouraging more prudent behaviour. However, to prevent disclosure from causing a loss of confidence, disclosure requirements could be phased in so as to give time for improvements in risk management to be implemented.

Supervisors should conduct their own independent assessments of banks’ exposure to credit risk, including currency-induced credit risk. These assessments should be conducted for individual banks and the overall banking system. Stress-testing techniques could be used to estimate banks’ exposure to credit risk under changing market conditions. These stress tests could be conducted with information regularly provided by banks and available in the supervisory data bases. Supervisors should have access to detailed information on borrowers, such as the one provided by credit bureaus. Alternatively, supervisors should have the option to define specific stress-testing scenarios, and require banks to run them. These off-site assessments should be complemented with an on-site evaluation of banks’ main exposures to credit risk. Box 8.2 presents an example of simple stress tests that could be applied by supervisors. A more detailed presentation is included in Appendix 8.2. Although direct limits or prohibitions on foreign currency loans to inadequately hedged borrowers may have been successful for some countries, such as Brazil or Chile, their application to highly dollarized financial systems may not be advisable. Indeed, such measures tend to be used by countries where high dollarization has not yet become a problem. Administrative measures generally introduce distortions, have high costs and promote regulatory arbitrage. Hence, their applicability in countries that are already highly dollarized is likely to be limited. Instead, it is preferable to let banks manage their risks.

Expected losses

Banks’ loan classification systems should reflect the borrowers’ capacity to repay under various market conditions. Loan classification rules in some partially dollarized countries are still based on past payments performance. These need to be phased out and replaced by more forward-looking rules. In particular, a borrower should be downgraded, even if he has always been current on all payments, if his projected cash flow is not adequate to fully service all his debts under expected market conditions. The degree of downgrading should also be related to the extent of the impairment of the borrower’s repayment capacity under likely stress-test scenarios. Note that this approach does not imply that foreign currency borrowers who do not have a foreign currency income should be automatically downgraded.

Box 8.2Quantitative assessment of currency-induced credit risk and its application to off-site supervision: the case of Peru

The impact of exchange rate shocks on the FX loan portfolio (see Appendix 8.2 for details). In the case of Peru, the econometric estimation of the relationship between provisions and depreciation shows that a 1 per cent depreciation results in a 1.9 per cent increase in the growth of provisions on FX loans:

  • Growth(FX_loan_provisions/FX_loans) = 1.9*Depreciation + f[GDP growth; Lending Rate; Inflation].

The probability distribution of exchange rate shocks and the desired level of protection against currency-induced credit risk. With these econometric estimates, a bank supervisor can assess the minimum capital buffer for a bank to withstand, for example, a 99th percentile exchange rate shock. The historical distribution of exchange rate shocks in Peru since mid-1992 shows that the 99th percentile corresponds to a depreciation of 43 per cent. Clearly, the larger the shock the supervisor would like to protect against, the more costly in terms of bank capital. Consequently, the supervisor may consider capital buffers against more moderate shocks, such as a 20 per cent depreciation (90th percentile shock).

Histogram: annual depreciation

The capital buffer will depend on the size of the exposures, including the FX loan portfolio, and can be set in terms of additional provisions on FX loans or additional capital charges. The figure below shows (i) the capital buffer against an annual depreciation of ‘x’ per cent and, as alternative measure, (ii) the risk weight on FX loans that would ensure compliance with the minimum regulatory CAR of 9.2 per cent. The buffer has been calibrated to the level of RWA and FX loans of Peru’s banking system.14

Annual depreciation

Loan loss provisions should reflect the impairment of borrowers’ capacity in the event of expected changes in market conditions. A combination of specific and general provisions could be applied for this purpose. Specific provisions could be applied when expected losses are estimated for individual borrowers, and general provisions when the estimation of expected losses results from aggregate tests.

The application of such loan classification and provisioning rules may require banks and supervisors to have analytical capabilities that may not currently be in place. As developing these capabilities takes time, two simplified options could be explored in the meantime:

  • A prescriptive approach would establish automatic downgrades (and higher provisions) for types of borrowers whose capacity to repay is perceived to be highly sensitive to expected changes in the exchange rate. These rules would have to be designed for each particular country, taking into account specific market conditions and institutions. In some countries, it might be individuals that are most at risk, in others, it might be corporates in the non-tradable sector. While simple to implement, this approach has a number of drawbacks, as it creates arbitrary divisions between types of borrowers and could promote regulatory arbitrage. If badly designed, it may not be of much value in internalizing risks.

  • Alternatively, banks could develop their own systems and capabilities to determine provisions. Those banks that, in the view of the supervisor, do not have appropriate systems in place, should be required to create an additional general provision on their overall foreign currency portfolio. Provided that this provision is high enough (it would need to be higher than – or similar to – estimated specific provisions under an expected loss provisioning system), this approach has the advantages of creating the right incentive for banks to develop their own risk management capacity.

Unexpected losses

Banks should hold enough capital to cover the credit risk caused by unexpected changes in the exchange rate. Authorities must set the size of this buffer and define how to achieve it. The buffer should provide sufficient cover for an exchange rate shock within acceptable risk-tolerance levels. It should apply only to assets that constitute a currency-induced credit risk (for un-hedged borrowers). While capital (rather than provisions) would normally be required to cover unexpected losses, for reasons to be discussed below, some authorities may choose provisions. For simplicity, the decisions about the size of the buffer and its nature are presented independently.

The size of the buffer

The size of the buffer would depend on: (i) the impact of the exchange rate shocks on the value of banks’ portfolios; (ii) the probability distribution of exchange rate shocks; and (iii) the degree of protection with which authorities are comfortable. Since the shocks and their impact on the value of portfolios are subject to a large degree of uncertainty, the decision will be affected by the availability of information and by assumptions made. In turn, the degree of protection sought by policymakers should take into account that higher protection increases the cost of conducting banking business. The three following issues therefore need to be addressed:

Impact of exchange rate shocks on the value of banks’ portfolios. The quantitative assessment of currency-induced credit risk typically involves: (i) econometric estimates of the relationship between exchange rate movements and a measure of credit risk, using either aggregate bank data or the financial information of the borrowers; or (ii) forward-looking assessment of the impact of a devaluation on the repayment capacity of the borrower measured by its interest coverage ratio or other financial indicators. The first methodology relies on past information; hence, it may not provide a good indicator of the effect of a future depreciation in the event of changes in regulation or changes in the underlying quality of borrowers. The second methodology requires information on the currency composition of the borrowers’ balance sheet and income statement that might not always be available. Both methods depend on the quality of information available and the adequacy of the period of time covered by such information. For instance, currency-induced credit risk could be underestimated if the data available corresponds to a period of economic prosperity (i.e., the upturn of an economic cycle). An example of an assessment of this risk, under the first method, is presented in Box 8.2, where it is used to examine the size of the buffer needed to cover currency-induced credit risk of various depreciation levels (details are presented in Appendix 8.2).

Probability distribution of exchange rate shocks.15 To estimate the probability distribution of exchange rate movements, authorities may refer to past information on exchange rates in the local economy. This approach has the drawback that past information regarding exchange rate changes may not be a good predictor of future changes, particularly for countries with fixed or managed exchange rates, where imbalances have only recently arisen.16 An alternative could be to use historical information from similar countries that have experienced exchange rate shocks or simulations. A macroeconomic model that captures the accumulating imbalances or the distribution of exchange rate changes in countries with similar characteristics could be used. The methodology should be carefully explained to prevent simulations of exchange rate shocks being regarded as a signal of a change in the exchange rate regime or a decision to pursue policies inconsistent with maintaining this regime.17

The desired level of protection. Authorities must set a risk-tolerance level for currency-induced credit risk. Should they be protected from 99 per cent of possible levels of depreciation or only from 95 per cent? The larger the depreciation, the larger the buffer and, thus, the higher the cost of banking. Hence, the benefits from additional protection should be carefully weighed against its costs. The latter (Ci) that result from protecting against an ‘i’ per cent devaluation can be measured as the product of the size of the buffer (Bi) times the opportunity cost of capital (r) in per cent of output (GDP) or bank profits (P):18

Ci=r*Bi/GDP or Ci=r*Bi/P.

Capital or provisions?

The decision on the nature of the buffer is not an easy one. The use of provisions to cover losses against unexpected exchange rate movements departs from International Financial Reporting Standards (IFRS), shortly to be adopted by many developed and emerging countries. In fact, under International Accounting Standard (IAS) 39, assets should only be subject to specific provisions when they are impaired, i.e., when there is a known event causing the impairment of the loan. Strictly speaking, a future unexpected depreciation does not qualify as a ‘known event’. General provisions, on the other hand, can only be required for losses that can be estimated in a pool, but have not yet been individualized. Thus, under IAS 39, potential losses from an unexpected depreciation would fail to qualify under either of these definitions and would need to be covered by capital requirements.

However, raising capital requirement may be difficult when such an increment would require a change in legislation. Moreover, an increase in capital, when mandated through an increase in the CAR, is rather inflexible and does not allow for much fine-tuning across risk categories. Instead, a preferable approach is to use higher risk weights for assets that are more sensitive to currency-induced credit risk.

When risk weights are not under the full control of the supervisors or cannot be adjusted across risk categories in a sufficiently flexible manner, the alternative is to use provisioning requirements. Under this option there is no need to distinguish between expected and unexpected exchange rate movements. Specific provisions could be required when the sensitivity analysis is conducted individually, and general provisions could be required when borrowers are assessed as a group.

Liquidity risk

Liquidity risk can be: (i) idiosyncratic, affecting individual banks and generally related to their own behaviour; or (ii) systemic, in case of widespread liquidity problems, generally stemming from macroeconomic fundamentals, contagion or generalized panic. While both types of risk are present in all financial systems, dollarized financial systems tend to be more vulnerable to systemic liquidity risk. The supervisory framework of a highly dollarized banking system should thus include: (i) a risk-based supervision of liquidity risk and its management, based on BCBS guidelines; and (ii) a minimum liquidity or reserve requirement to ensure that banks adequately internalize the liquidity risk that is specific to a dollarized environment.19 As in the case of capital requirements, the goal should be to better internalize risk. Limits on maturity gaps, while effective for the management of idiosyncratic risks, may not be as effective for managing systemic liquidity risk.

Supervision and management of liquidity risk

The international standards on liquidity management, presented in ‘Sound Practices for Managing Liquidity in Banking Organizations’, produced by the BCBS (2000), are a key reference for bank supervisors, and are applicable to highly dollarized financial systems. These practices state that banks should have a strategy for the management of liquidity (Principle 1), with policies approved by the Board (Principle 2). Banks should also: (i) establish a process for the ongoing measurement and monitoring of net funding requirements; (ii) analyze liquidity under a variety of ‘what-if scenarios; (iii) review their assumptions frequently to ensure they continue to be valid; and (iv) have contingency plans to handle liquidity crises (Principles 5, 6, 7 and 9). Currency-specific issues, a key concern in dollarized financial systems, are explicitly addressed. Banks are expected to have a system to measure, monitor and control their liquidity position in the major currencies in which they are active, and undertake a separate analysis of their strategy for each currency (Principle 10). Based on this analysis, banks are expected to set and review internal limits on the size of their cash flow mismatches over particular time horizons for foreign currencies in aggregate and for each significant individual currency in which they operate (Principle 11).

Following these guidelines, banks in dollarized economies should identify funding needs in the main currencies in which they conduct operations. This entails measuring all cash inflows against all cash outflows in each main currency, from all possible sources, including off-balance-sheet items. Immediate funding requirements, as well as future needs, should be determined. A variety of methods can be used for this purpose, but banks should consider not only contractual maturity but observed behaviour. Careful profiling of behaviour is essential for banks to adequately assess funding requirements under changing market conditions and make sound liquidity decisions. This approach should be also used to monitor compliance with internal limits as well as regulatory limits or cash requirements.

Liquidity analysis should consider a variety of stress tests of individual as well as system-wide disturbances, including the two main specific sources of liquidity risks for dollarized financial systems. First, scenarios should assess the impact on liquidity of changes in market conditions, such as a currency depreciation or rising interest rates, taking into account a bank’s exposure to currency-induced and interest rate-induced credit risk. These indirect credit risks will affect a bank’s liquidity, as assets may not be repaid under their original terms and may also be difficult and costly to sell under stressful conditions. Moreover, the resulting solvency risk increases the probability of a deposit run against all the banks that are perceived to share this problem, and may lead to systemic liquidity problems. Second, scenarios should model the impact on liquidity of currency-specific asset and liability volatility experienced in the past or that are likely to occur, including possible capital flight. The evolution of a bank’s currency-specific and overall liquidity profile – as measured by a maturity ladder or by ratios of liquid assets – under a variety of scenarios can be a useful benchmark for assessing the bank’s liquidity and for determining the action to be taken to improve the bank’s performance under those conditions.

Banks’ contingency plans should: (i) address their strategy for handling individual as well as system-wide liquidity crises; and (ii) establish procedures for making up cash flow shortfalls under emergency situations, including currency-specific shortfalls and cash shortfalls in foreign currency (bills). These plans should consider market and institutional restrictions that may exist under crisis situations. While foreign exchange shortages for a particular bank might be solved by accessing the foreign exchange market, such a recourse may no longer be available under crisis situations. Moreover, the problem of shortfalls in foreign currency bills is costly to overcome at short notice and under stress. While some central banks in dollarized countries store significant volumes of foreign currency bills, others may not be willing to assume the associated storing and shipping costs. Banks should thus have clear procedures to solve shortages of foreign currency bills at short notice. As small depositors are more likely to withdraw their deposits in cash, banks with a broad deposit base are likely to require proportionally more dollar bills. Under systemic crisis scenarios, the strategies should also consider possible limitations of access to key sources of funding (e.g., debt issues in domestic or foreign markets, domestic or foreign interbank funds) and recognize that market disturbances could hamper the liquidity of markets normally regarded as liquid (e.g., domestic corporate equity or bonds, government paper).

Supervisors should thus assess the adequacy of banks’ liquidity risk strategies, policies and procedures to withstand the disturbances that are likely to occur in the financial system in which they operate. In this regard, the supervisory review should examine whether banks’ plans, policies and actions take into account the specific sources of liquidity risk of operating in a dollarized financial system, such as currency-induced credit risk, the absence of a lender of last resort, and the added potential of systemic liquidity crunches. Banks’ contingency plans should take into account shortfalls of foreign currency liquid assets and foreign currency cash. The supervisory assessment of banks’ exposure to liquidity risk should also take into account liquidity risk derived from currency-induced credit risk. Based on this assessment, supervisors should require corrective measures as necessary, including changes to liquidity risk management policies and practices, additional holdings of liquid assets or lower maturity mismatches and a reduction in exposure to liquidity risks stemming from currency-induced credit risk.

A buffer for liquidity risk

Minimum liquidity or reserve requirements are preferred to more market-based measures, such as limits on maturity gaps. While the latter are more risk-sensitive and allow banks more freedom on how to manage their liquidity, hence are less costly, they are unlikely to provide an adequate buffer to protect banks and the financial system from systemic liquidity risk in dollarized systems. Under a systemic liquidity crisis, assets that are normally liquid may become illiquid, and liabilities that are normally stable may become unstable. This is particularly true if early withdrawal of time deposits and other liabilities is possible.

Required foreign currency liquid assets should be held in foreign currency and their rates may be set higher than domestic currency requirements. The specific design of the minimum requirements has to be tailored to the needs of each particular country. However, the following general considerations should be noted:

  • Eligible assets should be liquid even under stressful market conditions. In some countries this restricts the range of eligible assets to cash, deposits at the central bank and liquid deposits in investment-grade financial institutions abroad. In others, the range of liquid assets could be somewhat wider. If pressures are likely to mount to use central bank international reserves for other purposes (e.g., fiscal), it may be advisable to have liquidity requirements, instead of reserve requirements. In turn, if there are difficulties ensuring that bank liquid assets are in fact liquid (e.g., they may be pledged as collateral for bank operations), reserve requirements may be preferable. An alternative option is to set up a trust fund abroad, with the sole purpose of being used as collateral for central bank liquidity support to the contributing banks.

  • The liability base should include the broadest range of liabilities likely to be volatile under stressful market conditions. This reduces the scope for arbitraging the regulation through liabilities that are excluded from the minimum requirements.

  • When setting the minimum rates, authorities should assess the costs of these requirements against the benefits of the insurance they provide. An example of a simplified way in which the costs and benefits of these requirements can be assessed is presented in Box 8.3.

  • Liquidity requirements should not unduly tax banks’ operations or create a captive demand for government debt. If liquid assets are required to be held in the form of central bank liabilities, these should pay market interest rates.

Box 8.3Costs and benefits of prudential requirements to control liquidity risk: the case of Peru

This box estimates the costs and benefits of two overlapping prudential requirements used in Peru to control liquidity risks: reserve requirements and liquidity requirements.

Prudential norms

Reserve requirements: Required reserves are held as vault cash or deposits at the Central Bank in the deposit’s currency denomination. In addition to a 6 per cent non-remunerated reserve requirement applied to all deposits, foreign currency deposits are subject to 30 per cent marginal reserve requirement (down from 45 per cent in 1998, but up from 20 per cent in early 2004). This regulation was recently modified to subject more credits from banks abroad to reserve requirements. Thus, average required reserves on FCD are currently 29.5 per cent. The Central Bank pays a fixed rate of 2.25 per cent on foreign currency reserves above the 6 per cent level.

Liquidity requirements: Banks are required to hold liquid assets equivalent to at least 8 per cent and 20 per cent of all their liabilities maturing during the next twelve months, in domestic currency and foreign currency respectively. Eligible assets include vault cash, deposits at the Central Bank, Central Bank certificates of deposit, deposits in first-rate foreign banks and investments in securities negotiated in centralized markets and rated as investment grade by international agencies.

Costs

Both liquidity and reserve requirements affect banks’ profits, as liquid prime assets normally earn lower returns than less liquid assets, and reserve requirements are remunerated as below market rates. Assuming that, in the absence of liquidity or reserve requirements, banks would only hold liquid assets equivalent to 5.5 per cent of local currency liabilities and 3.6 per cent of foreign currency liabilities, costs would amount to 1.6 per cent of liabilities in foreign currency (1.5 per cent for reserve requirements and 0.7 per cent for liquidity requirements) and 0.45 per cent of liabilities in domestic currency (0.2 per cent for reserve requirements and 0.4 per cent for liquidity requirements).

Benefits

The marginal contribution of the liquidity and reserve requirements to limiting liquidity risk can be estimated by subtracting the liquid assets that banks would hold voluntarily from the required liquid assets, which amounts to 15 per cent of total bank liabilities. This figure is substantially above the maximum run experienced by the Peruvian banking system from 1993 to June 2002 (maximum losses for the banking system range from 2 per cent to 8.3 per cent). However, since runs have been generally accompanied by some flight to quality, this buffer would not protect against – and perhaps should not be expected to protect against – the largest run at the individual bank level.

8.4 Implementation issues

The implementation of the proposed supervisory framework raises a number of challenges for bank supervisors. How, when and to what extent the above-mentioned adaptations should be implemented are not easy decisions. The specific characteristics of the financial system and the economy in which it operates and the market infrastructure and institutions, have to be taken into account. The following questions are at the centre of the decision process:

  • Who really needs this framework?

  • How much risk reduction is desirable?

  • What is the best sequence of implementation?

  • How can regulatory arbitrage be avoided?

  • Is the implementation of this framework independent of the actions of other national authorities? Or should it be regarded as part of a broader national plan to reduce the vulnerabilities from dollarization and, therefore, be coordinated with other authorities, such as the fiscal or monetary authorities?

  • How can these measures be integrated into the broader plans of supervisory and regulatory improvements?

There is no unique dollarization threshold beyond which countries should implement the proposed measures. It is important to note that these recommendations are not to be regarded as an all-or-nothing package; some countries may need all of them, others may need some of them; yet others may need none. Clearly, countries with dollarization levels above 50 per cent, or with somewhat lower but increasing levels of dollarization, should seriously consider the risks discussed in this chapter. On the other hand, countries with dollarization levels below 15 per cent may not need to invest resources in adapting their supervisory frameworks. In many countries, however, it may not be easy to decide if it is worth investing time and resources to insure against risks that may or may not materialize. The recommendation for countries in the grey area is to first gather information that will allow an assessment of their exposure to these risks. Stress tests, such as the ones presented in Appendix 8.2, can assist in determining how sensitive the solvency and liquidity of their financial systems are to an exchange rate depreciation and other market disruptions associated with dollarization. In some countries, high exposure to dollarization risks could be concentrated in one business segment, such as mortgage or consumer loans, so it would be best to address these risks separately.

The costs of controlling the risks of dollarization could be high for highly dollarized financial systems. In fact, the costs of implementing the proposed measures would be high if the risks that need to be addressed are also high. The measurement of the additional risks from dollarization, and the estimated risk reduction achieved by a specific prudential measure and its costs, are key for the design of prudential requirements. How much protection against the risks of dollarization is desirable has to be gauged against the costs of this protection. Calculations such as the ones presented in Boxes 8.2 and 8.3 could be of assistance for this purpose. The objective of these prudential requirements cannot be to eliminate all risks, as the enormous costs would create huge incentives for arbitrage. In some cases, a carefully phased implementation would be necessary to achieve the desired risk reduction without making the business non-profitable. A road map for implementation needs to be tailored to the specific needs, the level of bank supervision and the condition of the banking system of each particular country.

While the optimal sequence of implementation cannot be defined independently of the country’s institutions, supervisory framework and the condition of its financial system, supervisors should recognize that:

  • Currency-induced credit risk is a specific type of credit risk. Supervisory measures that attempt to address this risk without having addressed the more general weaknesses of overall credit risk management are not likely to prevent the next financial crisis. More precisely, if banks do not have adequate overall credit risk management, they are unlikely to have good currency-induced credit risk management.

  • Better disclosure of risk exposures and management policies can promote market discipline; however, when applied to vulnerable financial systems and weak risk management, it could also create confidence problems. The timing of disclosure requirements has to be carefully considered to prevent this from happening.

Tightening prudential regulation may create incentives for regulatory arbitrage. Banks could be induced to transfer risks to, or register transactions in, other entities – domestic or foreign – that are not subject to the new prudential requirements. In designing these measures, authorities should thus try to reduce the scope for regulatory arbitrage. This is particularly important when designing provisioning or capital requirements, minimum liquidity requirements and limits or prohibitions on certain mismatches or transactions. Regulatory reforms may therefore need to be accompanied by measures geared at avoiding circumvention of such regulations. For instance, supervisors should be aware that tightening norms on currency-induced credit risk may lead banks to increase intermediation in local currency with much shorter maturities, or trade one risk for another (i.e., when long-term projects are financed with short-term domestic currency loans, the currency-induced credit risk is traded for a direct credit risk).

The possibility of increasing risk in the domestic currency loan portfolio should not be overlooked. Also, whenever possible, the regulations should be applied to all members of a conglomerate. For this purpose, effective consolidated supervision is necessary. It is noted, however, that in some of the countries surveyed, consolidated supervision of conglomerates is incomplete and ineffective. Moreover, authorities may lack the power to impose prudential requirements beyond the domestic borders, particularly if the cross-border members of the conglomerate are not branches of the domestic bank. Additionally, regulatory arbitrage is not a static problem. In fact, the effectiveness of some measures may erode over time as banks find ways to circumvent their costs and, thus, supervisors should often review them to ensure their continued effectiveness.

The recommendations discussed in this chapter can be easily integrated into the current improvement plans of most supervisory agencies. The proposed adaptations are based on the Basel I framework, but are also consistent with a transition towards Basel II. In fact, being in full compliance with Basel ‘Core Principles for Effective Banking Supervision’ (1997) and implementing other elements of the Basel I or Basel II supervisory framework should help to improve the effectiveness of the proposed measures.

Since prudential supervision can only mitigate some of the risks arising from dollarization, prudential measures may need to be part of a broader medium-term plan designed to promote the use of the domestic currency. The plan would include measures such as: keeping inflation low, removing administrative ceilings on interest rates, reducing high unremunerated reserve requirements for local currency deposits, developing markets for local currency-denominated public securities and improving the efficiency of the payments system.20 While most of the prudential measures to strengthen the supervisory framework could be implemented independently of other national authorities, coordination between the monetary and prudential authorities is advisable to improve the effectiveness of their plans.

Appendix 8.1 Country practices: survey results

This section explores current supervisory practices to control the risks stemming from banks’ foreign currency activities. It is based on a survey conducted between June and September 2004, in seventeen countries at diverse levels of economic development and in diverse regions across the world.21 The level of financial dollarization22 of the countries surveyed is also diverse, including six countries with more than half of their total deposits in foreign currency (Bolivia, Croatia, Lebanon, Peru, Singapore and Uruguay), six countries with dollarization levels between 30 and 50 per cent (Costa Rica, Honduras, Latvia, Romania, Slovenia and Turkey) and five countries with low levels of dollarization (Argentina, Brazil, Chile, Poland and Sweden).

Foreign exchange risk

Traditionally, prudential regulation on foreign exchange exposures has been based on limiting banks’ foreign currency exposures. More recently, however, an increasing number of countries are imposing capital requirements against open foreign exchange positions. Requiring capital for foreign exchange exposures, in addition to that required for credit risk, makes it more difficult for weakly capitalized banks to take on new risks. However, capital requirements give a bank greater flexibility in choosing the risks it will accept by allowing managers to allocate a bank’s capital between credit and market risk, including foreign exchange risk.23

Most countries surveyed have both capital charges and limits on foreign exchange exposures (Table 8A1.1). Three countries (Poland, Singapore and Sweden) have only capital charges on foreign exchange exposures, while six countries have only limits on these exposures (Argentina, Bolivia, Chile, Costa Rica, Honduras and Uruguay). The remaining eight countries have both. Capital charges on foreign exchange risk vary between 8 per cent (minimum Basel recommendation) and 12 per cent (in the case of Lebanon). A special case is Brazil, where the capital requirement is 50 per cent of foreign exposures exceeding 5 per cent of capital. Some countries (Bolivia and Peru) have asymmetric limits on foreign exchange exposures, whereas others permit structural positions (Lebanon).

Table 8A1.1Foreign exchange risk practices for selected countries
Prudential rulesSupervisory assessment and preventive action
FX risk capital charge requirement1Type2Limits on FX exposure3Long open position limit4Short open position limit4Type5Supervisory guidelinesFX exposure data collection
Specific FX risk management guidelines6Disclosure requirementsMost common FX position7Spot and forward markets reasonably active8
CountryIn per centI or CY/NIn per centIn per centI or CFrequency9
ArgentinaNn.a.Y30-30INYLA/ND
BoliviaNn.a.Y80-20IYNLA/ND
Brazil50CY30-30CYNLA/AD
ChileNn.a.Y20-20INNLA/AW
Costa RicaNn.a.Y100-100INYLA/NM
Croatia10I,CY20-20IYYLA/AD
HondurasNn.a.Y5-5IYNsN/NM
Latvia10IY--IYYLA/AM
Lebanon12CY--I,CYYLA/NM
Peru9.1IY100-2.5IYYLA/AM
Poland8I,CNn.a.n.a.n.a.YYLA/AD,M
Romania8I,CY20-20INNLA/ND
Singapore8I,CNn.a.n.a.n.a.NYn.a.A/AM
Slovenia8I,CYn.a.n.a.n.a.YYLA/NQ
Sweden8I,CNn.a.n.a.n.a.NYMA/AQ
Turkey8I,CY20-20IYYsA/NW,M
UruguayNn.a.Y--INYLA/ND,M

Over half of the countries surveyed have specific risk management guidelines pertaining to foreign exchange risk. In Peru, guidelines are quite specific regarding the methodology for their internal control systems, including value at risk, scenario analysis, back testing and stress testing. In other cases, there are general guidelines on risk management; however, there are no specific guidelines on foreign exchange risk.

Credit risk

Data collected by supervisory authorities

Some dollarized countries report collecting debtor information that distinguishes domestic and foreign currency claims (Bolivia, Lebanon, Peru and Uruguay). Detailed information on the loan portfolio, by debtor or by operation, is generally collected for large exposures. Argentina, Brazil, Chile, Croatia and Poland, for example, collect individual information for commercial or medium to large debtors, and gather aggregated information at the portfolio level, for consumer loans or small companies. Others (Bolivia, Costa Rica, Honduras, Peru, Turkey and Uruguay) gather individual information, also for small exposures.

Four countries (Poland, Costa Rica, Honduras and Uruguay), two of them with low dollarization levels, report having a working definition of the un-hedged borrower.24 However, these are internal definitions of the supervisory agency, not published or shared with financial institutions. Only Costa Rica requires banks to report their exposure to credit risk from un-hedged borrowers. Costa Rica reports gathering quarterly information on the capacity of large borrowers to generate a foreign currency cash flow and estimating, as a residual, large borrowers that do not have foreign currency revenues. As of March 2004, 28 per cent of all loans in the Costa Rican financial system are estimated to be granted in dollars to borrowers who do not generate foreign exchange revenue. Other supervisors report using information on debtors’ activities, the purpose of the loan and on the sector composition of exports and imports to make inferences about banks’ exposure to un-hedged borrowers (Uruguay). On this basis, it is reported that Uruguayan banks granted 74 per cent of their loans to un-hedged borrowers in June 2004.

Prudential rules

Few countries have implemented prudential rules aimed at controlling banks’ exposures to currency-induced credit risk. None of the responding countries report using higher capital requirements for foreign currency assets, relative to domestic currency assets.25 However, some highly dollarized countries have capital requirements that are higher than the 8 per cent minimum CAR recommended by the BCBS (Figure 8A1.1). Only one country (Peru) reports requiring higher provisions for foreign currency loans relative to domestic currency ones. In addition to Peru, five other countries (Singapore, Poland, Lebanon, Argentina and Chile) report that banks are expected to assign a higher risk rating to debtors whose capacity to repay is sensitive to exchange rate movements, in the context of their overall risk assessment of borrowers.

Figure 8A1.1Financial dollarization (FD) and capital requirements (CAR) for selected countries, 20041,2

Source: IMF.

Note:

1. Capital requirement ranges from 8 to 10 for Argentina and from 8 to 12 for Slovenia.

2. Foreign currency loan data not available for Brazil and Singapore. Information on foreign currency deposits for Lebanon as of 2003.

Limits or other restrictions are used by some countries with low degrees of dollarization (Brazil, Chile and Honduras) and by Argentina, formerly highly dollarized until the legal pesification (see Table 8A1.2). For instance, Brazil prohibits banks from granting loans in foreign currencies, but allows foreign currency-indexed loans. Chile requires that banks approve and report to the Superintendency, internal policies for the management of these credits prior to engaging in this type of business. Honduras has a limit of 15 per cent of foreign currency deposits that can be allocated to grant foreign currency loans to non-exporters. Argentina’s current legal framework26 stipulates that funds from foreign currency deposits should be allocated to foreign trade-related financing, inter-financial loans or Central Bank bills. In case of misallocation, there is an increase in liquidity requirements deposited at the Central Bank.

Table 8A1.2Credit risk regulations for selected countries
Specific rules for credit risk exposures to un-hedged borrowers
Capital requirements for credit riskBanks required to approve internal policies and inform supervisory agency
CountryCapital requirementsType1Higher capital requirementsHigher provisionsHigher generic provisionsLimits or prohibitionsHedging requirementsHigher risk rating expected
Argentina8 to 10I,CNNNYNYN
Bolivia10INNNNNNN
Brazil11CNYNYNNN
Chile8CNNNNNYY
Costa Rica10INNNNNNN
Croatia10I,CNNNNNNN
Honduras10INNNYNNN
Latvia10I,CNNNNNNN
Lebanon12CNNNNNYN
Peru9.1I,CNYYNNYN
Poland8I,CNNNNNYN
Romania12CNNNNNNN
Singapore10I,CNNNNNYN
Slovenia8 to 12I,CNNNNNNN
Sweden8I,CNNNNNNN
Turkey8I,CNNNNNNN
Uruguay10I,CNNNNNNN
Source: IMF.

Supervisory guidelines

Six of the surveyed countries have explicitly required banks to manage and control their currency-induced credit risk. For instance, in Lebanon and Peru, banks are required to gather data and perform an analysis that allows them to assess the extent of their borrowers’ exposure to a currency depreciation. General guidelines recommend that banks assess the borrower’s ability to generate foreign currency cash flow to hedge against a possible change in the exchange rate. In Poland, the General Inspector of Banking Supervision has issued guidance letters on the management of risks stemming from foreign currency lending.

Few countries conduct regular stress testing to estimate banks’ exposure to credit risks derived from borrowers’ currency, interest rates or maturity mismatches (see Table 8A1.3). Peru requires financial institutions to conduct annual stress tests to measure the impact of a 10 per cent and 20 per cent depreciation of the domestic currency on the repayment capacity of debtors in the loan portfolio. Uruguay and Romania also report conducting some stress tests to measure the impact of a currency depreciation on the loan portfolio, but these do not appear to be prepared on a regular basis. None of the responding countries report having conducted stress tests to measure banks’ exposure to credit risk derived from borrowers’ interest rates or maturity mismatches. However, four countries perform scenario analysis measuring the impact of changes in several macroeconomic variables on banks’ balance sheets, including, among others, changes in exchange rates, interest rates, inflation and GDP growth (Singapore, Poland, Slovenia and Lebanon). In Singapore and Slovenia the supervisory authority has issued recommendations for banks to conduct these tests. In contrast, in Poland and Lebanon these stress tests are mandatory for banks.27

Table 8A1.3Credit management and stress-testing practices
Supervisory guidelinesStress tests
CountryCredit risk management guidelinesGuidelines on managing credit risk of borrowers’ currency mismatchesStress tests for credit risk from borrowers’ currency mismatchesSupervisory guidelines for the stress testsStress tests for credit risk from borrowers’ interest rate mismatchesStress tests for credit risk from other market variablesScenario stress tests (simultaneous shocks: i.e., GDP, reduced credit access)Disclosure requirements on credit risk exposures to un-hedged borrowers
ArgentinaYYNNNNNN
BoliviaYNNNNNNN
BrazilNn.a.n.a.n.a.NNn.a.n.a.
ChileNNNNNNNN
Costa RicaNNNNNNNN
CroatiaNNNNNNNN
HondurasNNNNNNNN
LatviaYNNNNNNY
LebanonYYYYNNYY
PeruYYYYNNNN
PolandYYYYNNYN
RomaniaNNYNNNYN
SingaporeYYYYNNYN
SloveniaYNYNNNYN
SwedenYn.a.n.a.n.a.NNYn.a.
TurkeyYNNNNNNN
UruguayYYYNNNNN
Source: IMF.

Supervisory assessment and preventive action

Banks’ credit risk management policies and practices are generally examined during on-site examinations. In this process, some countries report assessing the overall credit risks of banks (Croatia, Poland, Slovenia, Lebanon and Honduras) and others report conducting a specific assessment of the exposures to credit risk from foreign currency loans to un-hedged borrowers (Singapore, Lebanon and Uruguay). In Singapore, the supervisory authorities perform independent assessments of banks’ exposure to credit risk from foreign currency loans during both off-site review and the on-site examination process. In Lebanon, for large borrowers (with facilities exceeding 15 per cent of the bank’s capital, or $5 million, whichever is less), this assessment includes credit risk from foreign currency loans to unhedged borrowers. Uruguay makes a general classification of a bank loan portfolio in the ‘tradable’ vs. ‘non-tradable’ sector, and analyzes the portion of loans in the non-tradable sector that is denominated in foreign currency.

Most countries report that hedging instruments are available in their markets for borrowers to hedge against foreign currency risks; however, the extent of their use by borrowers is not known. Only a few countries have very active markets (Singapore, Sweden, Poland and Brazil) with a wide variety of available instruments. Most of them have shallow markets, with one or two main instruments offered, mostly forward contracts with the exchange rate as the underlying variable (Croatia, Romania, Slovenia, Lebanon and Peru).

Most dollarized countries surveyed report that banks appear not to be pricing the foreign currency risk derived from un-hedged currency mismatches (Croatia, Slovenia, Turkey, Honduras and Uruguay). Banks tend to operate under the assumption that the authorities will keep the current exchange rate regime. The fact that this could be very costly and may not always be true fails to be priced in. Uruguay explicitly mentions this as a problem, and reports seeking to discourage this behaviour by requiring banks to take this mismatch into consideration when rating and provisioning borrowers. Four countries (Singapore, Sweden, Lebanon and Costa Rica) report that interest rates charged by banks are a function of the overall risk of the borrower, which may include risks associated with the sensitivity of their capacity to repay in the event of exchange rate movements. In turn, three countries (Latvia, Poland and Romania) report that they have specifically identified that banks charge higher interest rates to borrowers with currency mismatches.

Liquidity risk

Data collected by supervisory authorities

Fourteen of the seventeen countries report collecting some information on liquidity risks on a currency-specific basis. Eleven of these collect information on maturity mismatches in each of the currencies that are significantly important, including three countries with low financial dollarization. Some countries collect reports on maturity mismatches based on contractual maturity and adjusted behaviour profiling (Singapore and Chile), others (Peru) do it based on adjusted behaviour only (Table 8A1.4).

Table 8A1.4Liquidity risk management practices
Prudential rules
Data collection1Liquidity requirementsReserve requirements
On currency-specific liquidityOn currency-specific maturity gapsOn concentration of liabilitiesOn sources of fundingOn marketability of assetsMarginal rate for domestic currencyMarginal rate for foreign currencyOther
Rates for domestic currencyRates for foreign currencyDifferentiated by maturityRates for domestic currencyRates for foreign currencyDifferentiated by maturityInterest paidLimits on maturity gaps
ArgentinaYYYYYn.a.n.a.n.a.0 to 180 to 35YNNYN
BoliviaYNYYN0 to 100 to 10Y0 to 20 to 2Yn.a.n.a.YN
BrazilNNYYN8 to 45n.a.n.a.15 to 45n.a.Nn.a.n.a.N(*)N
ChileYYYYYn.a.n.a.n.a.3.6 to 93.6 to 9<1 yearn.a.n.a.YY
Costa RicaYYNFSNn.a.n.a.n.a.1010Nn.a.n.a.NN
CroatiaYNYYYn.a.35N1919N2424YN
HondurasYNNYNn.a.38N1212Nn.a.n.a.NY
Latvian.a.YNYN3030<30 days44n.a.n.a.n.a.YN
LebanonYYYYY1010n.a.n.a.15<1 yearn.a.n.a.YN
PeruYYYYY820<1 year66Nn.a.30YN
Polandn.a.YYYNn.a.n.a.n.a.3.53.5Nn.a.n.a.YN
RomaniaYYYNNn.a.n.a.n.a.1830<2 yearsn.a.n.a.YY
SingaporeYYNYFS1818N33n.a.n.a.n.a.NN
SloveniaYYNYNn.a.100<6 months2 to 4.52<2 yearsn.a.n.a.YY
SwedenNNNFSFSn.a.n.a.n.a.n.a.n.a.n.a.n.a.n.a.n.a.N
TurkeyYYNYYn.a.n.a.n.a.611Nn.a.n.a.YN
UruguayNNNFSFS3030<1 year4 to 1719 to 25Yn.a.n.a.YN
Source: IMF.

Prudential rules

Measures have been taken to reduce the vulnerabilities of financial systems to liquidity risks that could arise from financial dollarization. The specific modalities of these arrangements vary across countries, though commonly countries utilize a combination of prudential measures. The most common combination is that of minimum liquidity ratios and reserve requirements.28 Some dollarized countries have high minimum requirements to build a buffer for liquidity risks. Others also apply higher requirements for foreign currency relative to domestic currency liabilities.

All the countries surveyed have minimum reserve requirements, with the exception of Sweden, and nine of them also have minimum liquidity ratios. The design of these instruments varies, though typically most countries require that liquid assets are held in the same currency as the liabilities they are expected to cover. Some countries apply higher rates for shorter maturity liabilities (Bolivia, Brazil, Chile and Uruguay) and others require higher rates for foreign currency deposits relative to domestic currency ones (Romania, Turkey, Lebanon, Argentina, Bolivia, Peru and Uruguay). A combination of average and marginal reserve requirements (Croatia and Peru) is used by some countries. For instance, in Peru there is a 30 per cent marginal reserve requirement for all foreign currency deposits and some liabilities with foreign financial institutions. Reserve requirements are compensated in most countries, albeit usually at below market interest rates. Some countries have established minimum liquidity requirements that apply only to foreign currency liabilities (Croatia, Slovenia and Honduras).

Four countries (Romania, Slovenia, Chile and Honduras) impose limits on maturity mismatches of banks’ assets and liabilities and one (Argentina) is considering imposing such a limit. In all these cases, the limits appear to be defined in terms of residual contractual maturities. Limits can be defined for one or two particular time buckets, as in Slovenia, Chile and Honduras, or for all time buckets on a cumulative basis, as in Romania. All of these countries define limits for the overall mismatches of domestic and foreign currencies. However, two countries set an independent limit – for one currency only – on the mismatch for the 30 days time band: foreign currency in Chile and domestic currency in Honduras.

Supervisory guidelines

Most countries have issued regulations or some sort of supervisory guidelines requiring banks to set appropriate policies and practices to manage liquidity risks. Some countries, such as Croatia, have issued these in the context of general risk management rules. Others, such as Slovenia and Bolivia, have issued specific regulations on liquidity risks. Only two countries, Latvia and Lebanon, have set explicit recommendations to manage foreign currency liquidity risks. In Lebanon, for instance, the ‘Generic Risk Management Manual’ includes specific requirements to assess any form of mismatch in each foreign currency, determine alternative sources of financing and consider committed lines of credit in foreign currencies.

In nine of the seventeen countries, banks are required to use stress-testing techniques to estimate the impact of market and other changes on their liquidity. Some regulators have explicit requirements for these stress tests. For instance, in Singapore, banks are required to examine their cash flows under bank-specific crisis and general market crisis scenarios. In Sweden, the FSA guidelines on these scenarios require banks to measure payments and to analyze liquidity risks for each currency separately where the bank is exposed. In Lebanon, banks are expected to simulate various scenarios considering market changes in terms of currency, instruments, volumes, maturity, rates and products. In most countries the parameters and specific conditions for these scenarios are to be defined by the banks. An exception is Peru, where the regulation sets an explicit stress scenario for banks to run. In this latter case, while the scenario conditions are equal for both currencies, the liquidity risks are to be assessed independently for each currency.

Contingency plans for adverse liquidity conditions are required in thirteen out of the seventeen countries. Wherever stress testing is required, the plan is generally designed to solve the specific conditions and vulnerabilities identified in the stress scenarios. Most supervisors review contingency plans during the on-site examination process, including Bolivia and Uruguay, which have no formal requirements for these plans to be formulated. In Poland, Turkey and Peru, the contingency plans are reviewed during the off-site as well as the on-site supervisory processes. One country (Sweden) restricts this review to the systemically important banks.

Supervisory assessment and preventive action

Most countries conduct an off-site review of banks’ liquidity on the basis of reports submitted by banks, and assess liquidity risk management during on-site examinations. These assessments generally focus on overall liquidity risks, and rarely look into specific aspects related to foreign currency operations. An exception is Poland, where supervisors conduct a specific assessment of liquidity risks arising from banks’ foreign currency operations for cases in which these are considered significant. In particular, Poland’s assessment of risks associated with foreign currency operations looks into the status and prospects of sources of funding, off-balance-sheet operations impacting liquidity risk levels, the impact of subsidiary cash flows on the bank’s liquidity and the impact on liquidity of foreign currency-induced credit risk, among others. Besides examining liquidity risks of individual institutions, some supervisors also conduct an assessment of systemic liquidity conditions and risks (Croatia, Lebanon, Brazil, Peru and Uruguay) and in the case of Latvia this assessment is done by the Central Bank. The Central Bank of Brazil, for instance, conducts system-wide stress tests for all financial institutions in the financial system, identifying vulnerable institutions and their related entities and feeding back these results to bank supervisors, in order to design appropriate corrective actions.

Institutional framework and market conditions

Under normal conditions, banks have access to a wide variety of sources of funds in foreign currency, but some of these may not be accessible under adverse liquidity conditions. To protect against liquidity risk, some countries with a significant share of foreign currency liabilities hold high levels of international reserves of the central bank and or commercial financial institutions (see Figure 8A1.2). While reserves of commercial banks are immediately available to attend to banks’ liquidity needs, central bank international reserves are made available through foreign exchange operations or liquidity facilities. Thus, some central banks provide liquidity in foreign currency, under normal or under exceptional circumstances. In six countries liquidity facilities are available only in domestic currency (Singapore, Croatia, Latvia, Argentina, Brazil and Honduras). Lebanon, Bolivia, Chile and Peru have open liquidity facilities in foreign currency that can be accessed regularly by banks. In Lebanon, these include: discount of commercial bills or of foreign currency reserve requirements, repos of Lebanese Eurobonds, overdrafts collateralized with commercial bills, gold or securities and purchases of bills or government bonds. In Bolivia, the Central Bank provides liquidity against banks’ reserves deposited abroad and through repos of Central Bank or government securities. In Peru and Chile liquidity in foreign currency is provided against banks’ reserve requirements deposited at the central bank. In Sweden and Slovenia the central bank provides liquidity in foreign currency through currency swaps. Several countries (Sweden, Poland, Lebanon, Costa Rica, Peru and Uruguay) provide lender of last resort credit in foreign currency, against eligible collateral. Eligible collateral can be limited to government or central bank paper and other first-class securities. Some countries, however, accept loans as collateral (Sweden, Bolivia, Costa Rica, Peru and Uruguay).

Figure 8A1.2Financial dollarization (FD) and international reserves (minus gold) for selected countries, 2004

Source: IMF.

Appendix 8.2 Currency-induced credit risk in selected banking systems

The purpose of this section is to provide a measure of the currency-induced credit risk in selected Latin American and European banking systems. To that end, using aggregate bank data, we estimate a relationship between exchange rate movements and indicators of credit risk, such as the NPL and provisioning ratios.

The econometric estimation of the effect of exchange rate movements on the NPL ratio requires controlling for other macroeconomic variables likely to affect the credit quality of borrowers, including GDP growth, interest rate on bank loans and inflation. To avoid marked seasonality effects, all the variables are expressed in annual growth rates except for the interest rate, where the annual average rate is used.29 The choice of a particular dynamic-specification for the regressors was guided by minimizing the sum of the squared errors. For all countries, the preferred specification includes one lag of the independent variables. However, a more general dynamic specification, allowing for lags up to one year, was also estimated, with similar results. Hence, given data availability constraints a more parsimonious dynamic specification was chosen.30

Table 8A2.1 reports the results of the estimation of indirect credit risk for selected countries with publicly available time series information on the ratio of NPLs to total loans. As expected, a depreciation of the domestic currency increases the growth rate of the NPL ratio in Peru, Bolivia and Poland. However, a depreciation has no statistically significant effect in Brazil, Chile and Slovakia. In all countries we found a significant effect of output deceleration and rises in interest rates on the growth of NPLs. To the extent that inflation reduces the real value of loans and facilitates their repayment, a negative relationship is to be expected between inflation and NPL growth. This effect is found statistically significant in Peru; however, for the other countries it is found not to have a significant effect with the exception of Bolivia, where the opposite effect is found.31

Table 8A2.1Estimates of annual NPL growth rates in selected banking systems
BrazilBolivia1ChilePeruPolandSlovakia
Depreciation-.066.9**-.021.57**.47**-.01
Production growth-1.45**-7.5**-4.73**-.91*-.64**-1.13*
Interest rate1.0**3.3**4.0**5.0**3.0**1.0**
Inflation.601.5*.49-5.4**.61-.24
Adjusted R2.55.58.67.82.66.30
Observations98/01-90:Q1-97/02-94/12-99/11-96/01-
04/0104:Q304/1004/0904/0404/09
Note: Staff’s own estimates based on monthly data, except for Bolivia, where quarterly data are used. ‘*’, ‘**’ indicate statistical significance at the 90 and 95 per cent level respectively, based on Newey-West heteroskedasticity-autocorrelation consistent variance-covariance matrix.

Alternatively, credit risk in the banking system can be proxied by the ratio of provisions to total loans. Following the same methodology used in the estimation of the annual growth of NPLs, we estimate the impact of a depreciation on the annual growth of provision expenditures as a share of loans. Table 8A2.2, in column 3, reports the estimation results, which are qualitatively similar to the results of the estimations of the NPL growth rate (information on the ratio of provisions to total loans was only available for Peru).

Table 8A2.2Peru: estimates of annual provision to loan growth rates
NPL to loan ratioForeign currency NPL to foreign currency loan ratioProvisions to total loans ratioForeign currency provisions to foreign currency loan ration
Depreciation1.57**1.95**1.32**1.90**
Production growth-.91*-1.23**-.51-.81**
Interest rate5.0**2.0**5.0**2.0**
Inflation-5.4**-5.47**-5.3**-5.14**
Adjusted R2.82.85.82.78
Observations12/94-09/0412/94-09/0412/94-09/0412/94-09/04
Source: Staff’s own estimates.‘*’, ‘**’indicate statistical significance at the 90 and 95 per cent level respectively, based on Newey-West heteroskedasticity-autocorrelation consistent variance-covariance matrix.

Disaggregated information on the credit quality of loans by type of currency and by type of loan (i.e., consumer, mortgage or corporate) allows for a more accurate estimation of the effects of depreciation. The estimated effect of depreciation on the total NPL or provisioning ratio would help predict future effects of devaluation only if the degree of banking system dollarization remains broadly stable. For example, if dollarization has been increasing, future depreciations will have a bigger impact since more borrowers will be negatively affected in the event of a depreciation. Changes in the composition of the bank’s foreign lending portfolio are also important. Even if the degree of dollarization is stable, when the proportion of consumer and mortgage loans increases, the indirect exchange rate risk is also likely to increase, since retail borrowers are typically un-hedged. Table 8A2.2, in columns 2 and 4, shows the results for foreign currency NPLs and provisions. The regressors are also modified accordingly by replacing the average lending rate for the lending rate on foreign currency loans. The results are qualitatively similar to those of total NPLs and provisions, yet, as expected, the effect of depreciation turns out to be more pronounced in both cases.

The quantitative effect on credit risk of an exchange rate shock varies substantially across countries and appears to be a ‘threshold effect’ related to the degree of dollarization. In particular. Table 8A2.3 shows a very large effect for Bolivia – a 1 per cent depreciation leads to a 6.9 per cent increase in NPLs. The large effect appears to be related to (i) the large proportion of dollar lending (97 per cent of foreign currency loans), (ii) very high corporate debt and (iii) a relatively low share of tradable goods (especially when abstracting from hydrocarbon-related exports). Peru also presents a relatively large effect (1.6 per cent) as well as a high degree of dollarization (79 per cent). In Poland, with a moderate dollarization rate, the quantitative impact of depreciation is much less pronounced (0.5 per cent). In contrast, those countries where the level of dollarization is comparatively low – Brazil (13 per cent), Chile (18 per cent), Slovakia (13 per cent) – an exchange rate shock has no statistically significant effect on credit risk. Also, in most countries, the size of the tradable sector (imperfectly measured by the share of exports in GDP) seems to correlate negatively with the quantitative impact on credit risk.

Table 8A2.3Effect of an exchange rate shock (ERS) on NPLs’ and provisions’ annual growth rate
Indicators of currency mismatch and corporate

leverage
Effect of ERSAverage loan dollarizationExport/GDPCorporates’ debt/asset
Total NPLs
Brazil10.013.210.633.0
Bolivia6.997.020.843.0
Chile10.017.630.530.0
Poland0.522.727.417.0
Slovakia10.012.964.9-
Case study: Peru
Total NPL ratio1.679.314.933.0
Total provisioning ratio1.379.314.933.0
Foreign currency NPL ratio2.079.314.933.0
Foreign currency provisioning ratio1.979.314.933.0
Note:

The parameter estimates can be added to the set of early warning tools of risk management and bank supervision. Figure 8A2.1 shows the results of stress testing credit risk exposures to various-sized exchange rate shocks for Peru. The parameter estimate for foreign currency provisions in Table 8A2.3 together with information on profits, capital and risk-weighted assets (information available to risk mangers as well as to supervisors) can be combined to assess the effect of a given exchange rate shock on the capital adequacy ratio. According to our estimates, a 30 per cent devaluation will reduce the CAR from 14 to 12 per cent.32 Furthermore, a 70 per cent devaluation will push the CAR below the minimum regulatory level (9.21 per cent). A similar analysis can also be used to determine the minimum CAR level necessary to withstand a given devaluation. For example, according to our estimates, the minimum CAR level to withstand a 20 per cent exchange rate shock would be 10.6 per cent.

Figure 8A2.1Peru: effect of currency-induced credit risk on the solvency of the banking system

References

    Abrams, R. and P.Beato (1998) ‘The Prudential Regulation and Management of Foreign Exchange Risk’, IMF Working Paper 98/37 (Washington, DC: International Monetary Fund).

    Basel Committee on Banking Supervision (1980) ‘Supervision of Banks’ Foreign Exchange Positions’ (August).

    Basel Committee on Banking Supervision (1988) ‘International Convergence of Capital Measurement and Capital Standards’ (July).

    Basel Committee on Banking Supervision (1997) ‘Core Principles for Effective Banking Supervision’ (April).

    Basel Committee on Banking Supervision (1998) ‘Amendment to the Capital Accord to Incorporate Market Risks’ (January1996, updated to 1998).

    Basel Committee on Banking Supervision (1999) ‘Principles for the Management of Credit Risk’ (July).

    Basel Committee on Banking Supervision (2000) ‘Sound Practices for Managing Liquidity in Banking Organizations’ (February).

    Basel Committee on Banking Supervision (2003) ‘The New Basel Capital Accord’ (April).

    Basel Committee on Banking Supervision (2004) ‘Implementation of Basel II: Practical Considerations’ (July).

    Calvo, G. and C.Reinhart (2002) ‘Fear of Floating’ Quarterly Journal of Economics, Vol. 117, No. 2, pp. 379-408.

    de Nicoló, G., P.Honohan and A.Ize (2005) ‘Dollarization of Bank Deposits: Causes and Consequences’, Journal of Banking and Finance, Vol. 29, No. 7, pp. 1697-727.

    Escobar, F. (2004) ‘Efectos de las variaciones del tipo de cambio sobre las actividades de inter-mediación financiera del sistema financiero boliviano 1990-2003’, Working paper, Central Bank of Bolivia.

    GuideA.-M., D.H.Hoelscher, A.Ize, D.Marston and G.de Nicoló (2004) ‘Financial Stability in Dollarized Economies’, IMF Occasional Paper No. 230 (Washington, DC: International Monetary Fund).

    Ize, A. and A.Powell (2004) ‘Prudential Responses to De Facto Dollarization’, IMF Working Paper 04/66 (Washington, DC: International Monetary Fund). Revised version published in the Journal of Policy Reform, Vol. 8, No. 4 (2005), pp. 241-62.

    Reinhart, C., K.Rogoff and M.Savastano (2003) ‘Addicted to Dollars’, NBER Working Paper No. 10015 (Cambridge, MA: National Bureau of Economic Research).

    Tornell, A. and F.Westermann (2002) ‘Boom-Bust Cycles in Credit Constraint Economies: Facts and Explanation’, IMF Staff Papers, Vol. 49, Special Issue, Second Annual IMF Research Conference (29-30 November 2001).

Notes

The authors are grateful to Silvia Ramirez, Vania Etropolska, Moses Kitonga and Marie Carole St Louis for their valuable assistance. They also wish to thank Peter Hayward, Alain Ize, David Marston and Philip Turner, as well as the participants at the April 2005 Lima conference and an IMF seminar for valuable comments and suggestions. They are especially indebted to the supervisory authorities from the seventeen countries that responded to the questionnaire and provided comments on an earlier version of this chapter.

See Ize and Powell (2004), for a presentation of the need for prudential measures to reduce the vulnerabilities from dollarization.

These vulnerabilities have been extensively discussed in Guide et al. (2004). See also de Nicoló, Honohan and Ize (2005).

Tornell and Westermann (2002) note that the incentive structure is sufficiently strong that small firms belonging to the non-tradable sector borrow more intensively in foreign exchange in periods of boom encouraged by bail-out guarantees and sometimes real exchange rate appreciation. This explains the increase of the non-tradable to tradable output ratio in these periods.

Ize, Kiguel and Levy Yeyati (Chapter 9 in this volume) show that requiring banks to hold a minimum level of foreign currency liquidity in proportion to their foreign currency liabilities is a second-best policy that ensures that banks do not free-ride on the central bank LOLR, and that weaker banks do not benefit at the expense of stronger and more prudent banks.

Current standards are contained in various documents of the BCBS (including 1997, 1999 and 2000).

The survey included six countries with more than half of total deposits in foreign currency (Bolivia, Croatia, Lebanon, Peru, Singapore and Uruguay), six countries with dollarization levels between 30 and 50 per cent (Costa Rica, Honduras, Latvia, Romania, Slovenia and Turkey) and five countries with low levels of dollarization (Argentina, Brazil, Chile, Poland and Sweden).

In August 2005, Uruguay approved higher capital requirements for foreign currency loans, by establishing a 125 per cent weight on these assets. This norm is scheduled to become effective in July 2006.

Current standards are contained in various BCBS documents BCBS (including 1997, 1999 and 2000).

The calculation of the capital buffer assumes that the net FX open position in the banking system is similar to the degree of loan dollarization. However, if the degree of loan dollarization is above the net open position, a larger capital buffer would be needed (see Box 8.1 for details).

Banks’ exposure to credit risk may be simultaneously affected by several market variables, including the exchange rate, interest rates, inflation and the level of economic activity. Authorities may consider these shocks individually or jointly, taking into account the correlations between these variables. While acknowledging this, the focus here is put on exchange rate shocks.

A long enough history, which includes the events leading to dollarization, is likely to also contain large exchange rate variations. However, these events might be hard to replicate under the improved monetary management applied in most highly dollarized countries in the recent past.

An additional complication occurs when authorities are committed not to devalue, as is the case in currency board regimes. From a prudential point of view, authorities may still want to build up a buffer against this unlikely event. The way this buffer is communicated is particularly important in this case, so as not to create mixed signals or self-fulfilling prophecies.

Other potential costs may arise as banks seek to elude the added cost of doing banking business. These include the possibility of disintermediation and regulatory arbitrage, which are more difficult to assess.

In the first case, the liquid assets are managed by banks, whereas in the second case, these are managed by the central bank, and are mostly constituted by central bank liabilities.

For a detailed presentation of these issues, see Guide et al. (2004).

Four responding countries are members of the European Union (EU) (Latvia, Poland, Slovenia and Sweden) and three countries are European non-EU members (Croatia, Romania and Turkey). There is also one Asian country (Singapore), one country from the Middle East (Lebanon) and eight countries from the western hemisphere region (Argentina, Bolivia, Brazil, Chile, Costa Rica, Honduras, Peru and Uruguay).

Financial dollarization is measured here as the share of foreign currency deposits over total deposits. The term dollarization is used for all countries, although in some of these countries the foreign currency of choice is not the US dollar, but the euro.

For instance, Uruguay considers ‘borrowers receiving loans in foreign currency, whose cash flow to repay loans is in local currency’. Poland regards an un-hedged debtor as one ‘that does not have the natural hedging, e.g., cash inflows denominated in foreign currencies, and does not secure his exposure on derivative market’. In contrast, Honduras and Costa Rica have definitions that refer only to the capacity of the debtor to generate foreign currency.

In Uruguay, higher capital requirements for foreign currency loans are to become effective in July 2006, by establishing a 125 per cent risk weight on these assets. A similar approach has been adopted by Georgia, a country not in the survey, where a 200 per cent risk weight is applied to foreign currency assets for the calculation of the minimum required capital to risk weighted assets ratio.

Article 23 of Decree 905/02 and related Central Bank regulations.

Although some countries require banks to disclose to the public their credit policies and, in general, the major risks that they are exposed to, there is no specific requirement to disclose credit risks emerging from lending to un-hedged borrowers.

Reserve requirements have been traditionally regarded as a monetary policy instrument to assist authorities control the money supply, as eligible assets required for compliance – cash and deposits at the central bank – are also central bank liabilities (base money). However, they have also been viewed as a special type of liquidity ratio and, as such, have been used as a prudential tool, operating as a buffer stock to face liquidity shocks.

The rationale is that interest rate levels are more relevant than interest rate changes in explaining changes in NPLs and provisions.

The econometric estimates are based on monthly data on NPL of the banking system, total loans of the banking system, average lending rates, real GDP (when not available, industrial production index was used instead) and the exchange rate with respect to the dollar or to the euro (for pre-1999 data, the euro rate was replaced with the Deutsch mark rate). Data were collected from January 1990 to the latest date for which observation are available for Brazil, Bolivia, Chile, Peru, Poland and Slovakia. The choice of this particular country set was guided by data availability among those emerging or developing economies with dollarized/eurorized banking systems. In most cases, complete data sets were only available from the mid-1990s. For Bolivia, only quarterly data were available (1990:Q1-2004:Q3). For Peru, data on total provisions, foreign currency NPLs and provisions were also collected. All data were downloaded from the international financial statistics (IMF), the web pages of the corresponding central banks, supervisory agencies and national statistical offices.

One possible rationale for the insignificant effect of inflation on NPL growth in Chile (and, to a lesser extent, in Brazil and Poland) could be the use of inflation-indexed contracts.

Simulations assume that additional provision requirements due to depreciation cannot be met with profits, only with existing capital.

    Other Resources Citing This Publication