Building Monetary and Financial Systems
Chapter

Chapter 13. Albania: Stress Testing for Banking Supervisors

Author(s):
International Monetary Fund
Published Date:
October 2007
Share
  • ShareShare
Show Summary Details
Author(s)
Karl Driessen

In April 2005, the Bank of Albania (BoA) hosted a three-day Workshop on Stress Testing for Banking Supervisors. The workshop drew over 30 participants from 14 central banks and supervisory agencies from Central and Eastern Europe (CEE). The workshop program was developed and presented by a panel of IMF staff and experts.1

Workshops are effective capacity building tools for larger audiences. They can supplement existing in-house training programs, in particular on more specialized topics for which it is difficult to develop training material. Hands-on exercises facilitate implementation of the course material in the local environment, and encourage deeper understanding. Regional workshops provide the additional benefits of sharing experiences, establishing professional networks, and fostering peer dynamics. The IMF has increasingly made use of workshops for delivering its technical assistance on monetary and financial issues. IMF-organized workshops can take different formats, including: (i) workshops delivered through a regional training institute, such as the Joint Vienna Institute (JVI) or the Singapore Training Institute (STI); (ii) regional workshops co-hosted by a central bank or supervisory agency; or (iii) workshops benefiting one central bank or agency.

The objectives of the Workshop on Stress Testing for Banking Supervisors were to familiarize bank supervisors with basic stress testing techniques and advocate its use in a risk-based supervision environment. The workshop—the first specifically targeted to banking supervisors—was intended to de-mystify stress testing as a possible extension to offsite analytical tools. Bank supervisors are often not involved in stress testing exercises, even if the data used for stress tests typically are collected through regulatory reporting mainly for banking supervision purposes. The workshop was also intended to buttress the risk-based approach to supervision by using a tool that centers on risk and its financial consequences. Although banking supervisors do not necessarily put financial stability on the top of their agenda, but rather the soundness of individual institutions, stress testing can be used both for monitoring system-wide and individual institution exposure to risk.

The workshop consisted of a series of presentations and exercises on stress testing topics. Several country representatives made presentations on stress testing practices in their respective agencies. A roundtable at the end of the workshop provided an opportunity for all participants to describe and comment on their practices. Each day featured a group exercise session focusing on stress testing one major risk (see Box 13.1.).

Although participating countries had different transition experiences, their respective financial systems shared many common features. Some participating countries had recently been admitted as members of the European Union (Czech Republic, Estonia, Hungary, Lithuania, Poland, Slovak Republic). Others were still in an earlier stage of transition, and continued to have a sizeable state presence in the economy.

At the time of the workshop, most countries had participated in the joint IMF/WB Financial Sector Assessment Program (FSAP), and had encountered stress testing in that context. Financial systems in most countries shared some of the following common traits: (i) a significant presence of foreign banks; (ii) a moderate to high degree of dollarization/euroization; (iii) continued state presence in the banking sector; and (iv) the presence of banks with poorly performing loan portfolios with exposure to state—or socially-owned enterprises.

Box 13.1.Stress Test Exercises—General

The exercises used during the workshop were based on the fictional financial system of Bankistan developed by IMF staff1

The spreadsheets contain partial balance sheet data for a set of financial institutions. These institutions are grouped in the following categories: state-owned banks, foreign-owned banks, and domestic private banks.

The worksheet develops stress tests for the following risks: (i) credit risk; (ii) interest risk; (iii) foreign exchange risk (direct/indirect); and (iv) interbank contagion.

1 Martin Cihák. “Stress Testing 101: A Hands-On Exercise,” IMF paper/unpublished. The exercise file and the accompanying documentation are available upon request from mcihak@imf.org.

This chapter provides an overview of the workshop contents, highlighting the effectiveness of workshops in technical assistance delivery. The second section, “Overview of Stress Testing,” provides an introduction to stress testing. This is followed by the identification of the vulnerabilities in financial systems that form the basis for stress testing.

The next section, “Implementing Stress Testing,” covers the main risks for which stress tests are usually deployed: market risk (interest and exchange rate risk) and credit risk. Principal approaches(bottom-up—using individual financial institution data as a starting point, or top-down—using a more aggregate econometric approach) are also described. Last section concludes.

Overview of Stress Testing

Stress tests can be defined as a set of statistical techniques to help assess the vulnerabilities of financial institutions and financial systems to exceptional but plausible events. The commonality of the techniques is that they address some “what if” questions: “What happens if interest rates suddenly increase?” “What if the exchange rate suddenly depreciates?” There is no best practice for stress testing. Stress test techniques (i) range from very simple to quite complex in methodology; (ii) have different coverage; (iii) differ in how they claim plausibility.2

Stress tests differ in terms of their methodological complexity. Broadly speaking, one can identify the following three methodologies:

  • Sensitivity analysis, where the effect of varying one parameter (e.g., interest rates) on a (set of) financial indicator(s) (e.g., capital adequacy) is observed.
  • Scenario analysis, where the joint effect of varying multiple parameters (e.g. interest rates, GDP growth rates, etc.) on a (set of) financial indicator(s) is observed.
  • Contagion analysis, where the effects of varying parameters are traced through a financial system based on cross-institution exposures.

Stress tests differ in their degree of aggregation. To analyze systemic risk, a significant part of the financial system must be included in the analysis. However, for banking supervision purposes, it would be useful to include all supervised institutions. Most often, stress tests are limited to the banking system, although in highly developed financial markets, where banking and life insurance products are sometimes sufficiently similar and integrated banking and insurance conglomerates dominate, the coverage of stress tests may be wider. For system-wide stress tests, there are two broad approaches: (i) “bottom-up,” where the stress test effects are aggregated over individual financial institutions; and (ii) “top-down,” where the focus is on correlations between more aggregated variables such as system-wide nonperforming loans (NPLs) and GDP growth. Often, data availability determines the approach followed.3

The risks that are typically stress-tested include credit and market risk, but other risks are also analyzed if relevant and feasible. The principal market risks—unforeseen changes in market prices for asset classes on banks’ balance sheets that affect asset valuation as well as income streams—that are identified in stress tests are (i) interest rate risk; and (ii) exchange rate risk. Credit risk includes both the direct credit risk embedded in the default probabilities of the loan portfolio, and indirect credit risk, which depends on the market risk carried by the bank client, but will affect the default probabilities if not properly hedged (e.g., a large devaluation will significantly increase the repayment obligation of a foreign-exchange denominated loan, and thus reduce the repayment capacity in domestic currency if not hedged).

There are different views on what constitute “exceptional but plausible” events. Perhaps the simplest way is to use a standard shock size (e.g., 100 basis point increase in the yield curve) for an interest rate stress test. This works well in a straightforward stress-test model where the effect of the shock is proportional to the size of the shock. Such an approach facilitates comparison across time, institutions, and different shock sizes.

An alternative method is to use a specific historical outcome (e.g., the largest depreciation witnessed in the last decade); this is sometimes referred to as a “worst-possible case”. A third possibility is to construct hypothetical scenarios based on analysis of past volatility and correlations, which is particularly useful if more than one parameter is varied, or the effects of a range of possible parameter values are desired.

Stress tests are by now part and parcel of the financial stability toolbox. With the growing interest in financial stability in the aftermath of the Asian Crisis of the late-1990s, and the widespread participation in the FSAP, many central banks and supervisory agencies now perform stress tests on a regular basis. They have also encouraged, or at times regulated, that commercial banks integrate stress testing into internal risk management practices.

Results of stress tests are now regularly reported in financial stability publications. Čihák (2006), in a study on how central banks report on financial stability, finds that over half of the Financial Stability Reports (FSR) examined publish the results of stress testing.4 Stress test results can be presented along various dimensions; the most common are the effect of the shock on the capital adequacy ratio, and on earnings.

Stress testing is a multi-stage process, and is continuously evolving. In the initial stage, the main macroeconomic and market risks as well as the main exposures to these risks need to be identified (see upcoming section on implementing stress testing). This is followed by a determination of the coverage of and the data needs for the proposed stress tests. A methodology must be selected, and the size of shocks for the sensitivity or scenario analysis must be determined. After running the stress test, the results must be analyzed and interpreted. Stress testing has developed over time: during FSAPs, authorities have become more involved in its implementation; commercial banks are also being increasingly requested to run stress test scenarios, not just as part of internal risk management, but as part of a system-wide stability exercise on the basis of parameters provided by the supervisory authorities; lastly, the scope of stress tests has increased, and including nonbank financial institutions has become more common.

As an analytical tool, stress testing has advantages and limitations, and should be used in conjunction with other analytical methods and information. Its principal strengths are its ability to provide a forward-looking assessment of risks and greater analytical precision—typically at the level of the individual financial institution—than alternative approaches such as the analysis of Financial Soundness Indicators (FSIs). The drawbacks include the robustness of the results with respect to the assumptions that have been made, the view of financial institutions as static portfolios rather than dynamic risk managers, and the omission of non-quantifiable factors. Therefore, it is prudent to combine the use of stress testing with further analysis, including of FSIs and of nonquantitative information such as the legal, regulatory, accounting, corporate governance, and tax frameworks.

Box 13.2.Stress Testing and Basel II

Stress testing is becoming increasingly relevant for banking supervisors. With the gradual move toward Basel II comes a more risk-based framework for supervision, which is fully compatible with stress testing. Some instances of where the Basel II Accord refers to risk include the following: (i) the framework’s inclusion of market risk requires stress testing to assess capital needs; (ii) supervisors should consider how a bank provides for unexpected events in evaluating capital levels (para. 750); (iii) banks are expected to stress-test their internal risk models (para. 527j); (iv) there are explicit references to concentration risk (para. 775), and liquidity and credit risk (e.g., see para. 158).

Identifying Vulnerabilities

What are the risks that banks and other financial institutions are exposed to? Banks are in the business of taking and managing risks, and thus minimizing exposure to risks is not the objective, neither for banks nor for supervisors. Rather, the level of risk should be proportional to the buffer that can be used if risks materialize—capital—and risks should be properly managed. (Figure 13.1.) illustrates the sources of risks to banks. Shocks can emanate from the real economy (real GDP growth, inflation, etc.), from government policies such as taxes and government payment behavior, or through the relations with the rest of the world (exchange rates, terms of trade, capital flows, etc.). Shocks can also originate with those that lend money to banks (withdrawal or rollover behavior), and with borrowers (repayment behavior).

Figure 13. 1.Origins of Shocks to the Banking System

Banks’ exposure to different risks is illustrated in (Table 13.1)The exposure to interest rate risk depends on the maturity mismatch of assets and liabilities, both on—and off-balance sheet. Exposure to exchange risk is limited to the net open position (assets minus liabilities) in each currency. Exposure to other market risks (e.g., stock prices) is a function of the net holdings of non-interest bearing securities. Credit risk exposure depends on the quality of the borrowers, and other factors that influence default probabilities. The exposure to liquidity risk depends on the predictability of asset and liability flows, which in turn are related to asset quality and interest rate volatility, among others.

Table 13.1.Banks’ Exposure to Different Risks
Interest

Risk
Exchange

Risk
Other Market

Risk
Credit RiskLiquidity Risk
Maturity mismatch; interest rate sensitivityNet open position by currencyNon-interest bearing securities positionDefaultsMaturity mismatch; cash flow; asset quality

During the workshop, regional data were presented to indicate the ranges of exposure to various risks. Average capital adequacy ratios of banking systems varied widely (from 12 to 32 percent at end-2003 for a sample of eleven participating countries), suggesting a different ability to withstand large shocks.

Several banking systems showed a rather high degree of maturity transformation, with long-term loans funded by short-term deposits; this points to a growing importance of interest risk—although maturity mismatches leave open the crucial question of whether interest rates are fixed or floating, which determines who (bank or borrower) carries the interest risk. Some banking systems had still largely underdeveloped lending activity (at least, to the private sector), limiting credit risk exposure. NPLs also varied greatly, leading to different assessments of asset quality.

Exposure to exchange risk depends to a large extent on the exchange rate regime. (Figure 13.2.) shows the exchange rate regime classification of a sample of CEE countries in 1999 and 2003. Managed float was the preferred regime followed by a conventional peg, and currency board regimes. Since managed float regimes tend to reduce exchange rate volatility and hence underprice exchange risk, both banks and their clients are encouraged to take larger open positions, increasing the exposure to exchange risk. Since pegs are vulnerable to speculative attacks, similar exposure build-up may take place over time.

Figure 13.2.Exchange Regimes for Central and Eastern European Countries, 1999–2003

Another indication of exposure to exchange risk is the degree of dollarization/euroization. (Figure 13.3.) shows the share of foreign exchange loans (deposits) as a share of total loans (deposits) at end—2003 for a sample of eleven countries. High shares may be indicative of exchange risk in bank products which could eventually hurt banks through deteriorating credit risk resulting from increased default probabilities of bank borrowers.

Figure 13.3.Share of Foreign Exchange Loans and Deposits, End-2003

Sources: IMF Country Reports, World Economic Outlook, and International Financial Statistics.

Implementing Stress Testing

The mechanics of stress testing depend on the methodology used. In the simplest form, where one starts with a financial institution balance sheet, the effect of a stress test is the product of (i) the shock that is applied and (ii) the balance sheet exposure. This will be illustrated below for foreign exchange risk.

Market Risk

Market risk encompasses all risks related to the fluctuation of prices of assets. This includes changes in (i) the exchange rate (foreign exchange risk); (ii) interest rates; and (iii) prices of quoted securities. Financial institutions have traditionally employed stress testing to manage their trading portfolios, which are exposed to these risks. The value—at—risk (VAR) model is a widely used method that calculates what the maximum losses on a portfolio are during a given period, given a confidence interval of, for example, 95 or 99 percent.5 Since CEE banks do not typically hold significant amounts of traded shares, this section will only discuss foreign exchange and interest risk. An example of market risk stress test practices—including on foreign exchange risk—performed in Albania is provided in Box 13.3.

Box 13.3.Stress Testing at the Bank of Albania—Market Risk

The Bank of Albania has gradually gained exposure to stress testing, including through various technical assistance and FSAP missions to Albania that took place just before the Workshop in 2005. The Banking Supervision Department conducts quarterly stress test exercises for all supervised banks.

Foreign exchange risk: Banks’ portfolios carried a high share (80 percent) of dollar/euro loans in total loans. Foreign exchange risk stress tests are applied with prospective devaluation of the Albanian lek versus the U.S. dollar (50 percent) and the euro (40 percent). Since banks limit their open position, and generally maintain a slightly long position in foreign currency, the exposure is limited, although there is some exposure to adverse dollar—euro movements.

Interest risk: An important source of risk in the Albanian banking system is interest risk: at end’2004, over 50 percent of assets were held as fixed—rate treasury bills. This was due to a prohibition in lending by the largest bank’ Savings Bank—imposed after large-scale lending problems. The standard interest risk stress test calculates the effect on the capital adequacy ratio of a 300 basis point shift in the yield curve.

Credit risk: Since in early 2005 levels of credit were still very low, credit risk was not considered an important source of risk. However, there has been rapid credit growth since, and stress testing practices must evolve as risk exposures change. It was estimated that 63 percent of borrowers taking out loans denominated in foreign exchange were unhedged in 2005. However, no indirect credit risk stress test was performed that would gauge the effect of a devaluation on the repayment capacity of (unhedged) borrowers.

Foreign exchange risk

The net open position in foreign exchange determines the direct foreign exchange risk exposure of a financial institution. For each currency, long positions (assets) are netted against short positions (liabilities) to calculate the net open foreign exchange position in that currency. Since most regulatory frameworks impose prudential requirements in this area, this information is typically readily available to banking supervisors for all banks. If the data are available for each currency, separate shocks (e.g., for the U.S. dollar and the euro) can be set in a stress test. Alternatively, the stress test can analyze the effects of a depreciation or appreciation of the domestic currency against all other currencies. Typical shock sizes that are used in sensitivity tests are (i) a ten percent depreciation against all currencies; (ii) the largest observed historical devaluation over some time period; (iii) a shock that will deplete capital for the most exposed banks. In addition to the direct foreign exchange risk that banks take on, often more important is the indirect foreign exchange risk that banks’ borrowers are exposed to. Since this impacts the repayment capacity of borrowers, it is classified under credit risk.

Interest risk

The exposure to interest risk depends on the maturity mismatch of assets and liabilities of financial institutions. The first step in an interest rate stress test is to group assets and liabilities (both on—and off—balance sheet) into a number of “maturity buckets” (e.g., 0–3 months; 3–6 months; 6–12 months; over one year) corresponding to their remaining maturities (if the assets and liabilities have fixed interest rates), or until they are repriced (if the assets and liabilities are subject to variable interest rates). In a typical bank case, most liabilities are relatively short—term, and assets are longer—term. When interest rates rise, banks will suffer losses, since they face higher interest expenditures on deposits, but must wait before they can increase lending rates. To limit exposure to interest risk, banks may pass on the risk to customers by shortening the repricing period. This creates indirect interest risk that will be borne by bank borrowers.

Credit Risk

Credit risk—and hence associated stress testing—is gaining in importance in CEE countries as financial deepening continues. At the same time, credit risk is quite difficult to model, as willingness and ability to repay is determined by many different factors, including the legal framework, credit culture, and liquidity of collateral, as well as macroeconomic factors such as economic growth, interest rates, and fiscal policy.

A simple approach to credit risk stress testing simulates a worsening credit portfolio. The scenario may indicate that all classified assets move down in some pre—determined way. For example, if there several classes of NPLs, the stress test may assume that a proportion of loans transition to the next lower class; this affects the need for provisioning and write—offs, and capital adequacy.

As noted above, credit risk can be a function of interest and foreign exchange risk borne by borrowers. However, these are difficult to quantify using bottom—up methods, and lend themselves better to top-down stress tests. In top—down stress tests for credit risk, macroeconomic data are used to estimate a relationship between real economic activity and credit quality. One of the advantages of the top-down approach is that one can use models of the economy that are built to maintain internal consistency, which allows for their use with a range of macroeconomic policy (e.g., monetary and fiscal) scenarios. Box 13.4. describes aspects of the modeling approach of the framework used in the United Kingdom.

Box 13.4.Credit Risk Stress Testing—An Example from the Bank of England

In the context of the United Kingdom FSAP in 2002, the Bank of England developed a framework for top—down stress testing, which is documented in Bunn et al. (2005).

The stress test consists of six steps: (i) an initial shock; (ii) its impact on macroeconomic variables; (iii) the impact on borrowers; (iv) the impact on loss rates; (v) the impact on earnings; and finally (vi) the impact on profit and loss accounts of financial institutions.

Step 2 generates the consistency between the correlation of variables that impact credit quality. It is performed using a Bank of England macro forecasting model. Steps 3 through 5 use different estimated models of aggregated household and corporate balance sheets, and the banking sector, including the preponderance of credit card and mortgage arrears, corporate liquidation rate, and write-off rates, as functions of variables such as the unemployment rate, retail prices, and sectoral leverage (“gearing”) rates (debt-to—income ratios).

The scenarios used in the FSAP were (i) a 35 percent decline in global equity prices; (ii) a 12 percent decline in UK residential and commercial property prices; (iii) 1.5 percentage point unanticipated increase in average earnings growth; and (iv) 15 percent unanticipated depreciation in the trade-weighted exchange rate.

Conclusions

Stress testing can be seen as an extension of the risk management framework that is becoming increasingly important for purposes of bank supervision. To a large extent, this reflects the advent of Basel II, with its references to banking as a risk management business, and the need for banking supervisors to have a thorough understanding of risk management principles. Deeper understanding of stress testing by supervisors will also benefit those involved in financial stability analysis, as it will allow a more technical dialogue that may result in more accurate systemic analysis, as well as provide a basis for supervisory action against specific vulnerable institutions.

Workshops are convenient forms of technical assistance: they reach a wide audience, permit dissemination of recently gained insights by industry practitioners, and allow for country experiences to be shared. This holds true particularly for fast-moving topics such as stress testing, whose analytical frameworks still have to fully crystallize.

References

    BlaschkeWinfredMatthew T.JonesGiovanniMajnoni and Soledad MartinezPeria2001Stress Testing of Financial Systems: An Overview of Issues, Methodologies, and FSAP ExperiencesIMF Working Paper 01/88 (Washington, DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    BunnPhilipAlastairCunningham and MathiasDrehmann2005Stress Testing as a Tool for Assessing Systemic RiskFinancial Stability ReviewJune2005 (London: Bank of England).

    • Search Google Scholar
    • Export Citation

    ČihákMartin2004aStress Testing: a Review of Key ConceptsCzech National Bank Internal Research and Policy Notes 2004/2.

    ČihákMartin2004bDesigning Stress Test for the Czech Banking SystemCzech National Bank Internal Research and Policy Notes 2004/3.

    • Search Google Scholar
    • Export Citation

    ČihákMartin2006How Do Central Banks Write on Financial Stability?mimeo (Washington, DC: International Monetary Fund).

    Committee on the Global Financial System2005Stress Testing at Major Financial Institutions: Survey Results and Practice” (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation

    JonesMatthew T.PaulHilbers and GrahamSlack2004Stress Testing Financial Systems: What to Do When the Governor CallsIMF Working Paper 04/127 (Washington, DC: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    SorgeMarco2004Stress Testing Financial Systems: An Overview of Current MethodologiesBIS Working Paper 165 (Basel: Bank for International Settlements).

    • Search Google Scholar
    • Export Citation
1The workshop team consisted of Marcos Pineschi Teixeira (Central Bank of Brazil), Marco Sorge (Bank for International Settlements), Gudrun Johnsen (formerly IMF), and Karl Driessen (IMF).
2The literature on stress testing is vast and growing. Recent surveys include Jones, Hilbers, and Slack (2004), and Sorge (2004).
3Coverage can also be extended to financial institutions on a consolidated basis, including foreign branches and subsidiaries.
4Čihák (2006), in Table 7 of his article, reports that 55 percent of the FSRs examined contained the results of stress testing, with credit risk stress testing being the most frequently reported (in 55 percent of the cases), followed by interest rate risk (in 45 percent of the cases), and exchange rate risk (in 33 percent of the cases).
5For a survey of bank stress testing practices, see Committee on Global Financial Stability (2005).

    Other Resources Citing This Publication