This paper discusses Bulgaria’s prospects for converging to the living standards of the more advanced members of the European Union (EU). The unfavorable economic environment of the early 1990s and the economic crisis in 1996–97 hurt Bulgaria’s output, employment, and investment. Following the crisis, structural reforms and a sound macroeconomic framework set the stage for a sustained recovery. The structure of the Bulgaria economy has shifted markedly over the last decade, and investment has become the main engine of growth.


This paper discusses Bulgaria’s prospects for converging to the living standards of the more advanced members of the European Union (EU). The unfavorable economic environment of the early 1990s and the economic crisis in 1996–97 hurt Bulgaria’s output, employment, and investment. Following the crisis, structural reforms and a sound macroeconomic framework set the stage for a sustained recovery. The structure of the Bulgaria economy has shifted markedly over the last decade, and investment has become the main engine of growth.

V. Bulgaria—BNB Stress Tests of the Banking Sector72

A. Introduction

127. “Stress testing” refers to the use of various techniques to estimate potential vulnerabilities of financial systems to exceptional, but plausible, events. Stress tests of the banking sector require the specification of risks, usually, credit, interest rate and exchange rate risks. More complicated stress tests could include shocks to liquidity, equity prices, property prices and non-interest income, among other variables. Stress tests estimate the impact of a single shock (sensitivity analysis) or multiple shocks (scenario analysis).73

128. Stress tests tend to have wide coverage of the banking sector, and tend to comprise simple sensitivity analyses. According to Cihak and Hermanek (2006), they usually cover either all, or almost all, banks in terms of market share.74 Other segments of the financial sector are rarely covered. Further, stress tests tend to be very rudimentary, as many central banks are still only in the early part of their work in this area. 75 Credit risk is covered in almost all stress tests; interest rate risk is also covered in most of them (Table 1). Exchange rate risk is covered in some, but in many cases, is analyzed only in terms of open positions, and not as an explicit stress test. Some stress tests include scenario analysis, based on historical (a significant event experienced in the past) or hypothetical (a plausible event that has not yet occurred) scenarios. Only a few utilize econometric models, and even then, the models tend to be relatively simple.

Table 1.

Comparison of Stress Tests in Selected European Financial Stability Reports with the BNB Model

article image
Source: Cihak and Hermanek (2005) and Bulgarian National Bank.

Credit substitutes and derivatives.

129. In Bulgaria, stress testing of the banking system is performed by the Banking Supervision Department (BSD) of the Bulgarian National Bank (BNB). Presently, individual banks in Bulgaria are not required to perform their own stress testing. That said, some big banks do so as part of their risk management exercise, and these results are made available to the BSD’s on-site examiners.76 Banks which choose to apply the internal ratings-based (IRB) models under the Basel II capital requirements (likely to be implemented in 2007) must present their stress tests, as well as internal models for supervisory approval.77

130. The BNB’s stress testing framework is consistent with existing models at other central banks. It includes key shocks, such as to credit quality, interest rates and exchange rates, and takes into account shocks to derivatives positions (Table 1). Scenario analysis is also included in the BNB tests. Further, the BNB takes a very conservative approach to its stress tests. By including crisis-like scenarios in the stress tests—based on experiences during the 1996-97 financial crisis—the authorities are aware of the extreme case outcomes for the banking sector at all times.

131. The BNB also complements its stress tests with an Early Warning System (EWS).78 In this exercise, more than 100 ratios are calculated—well above the FSAP recommended set of core and encouraged financial stability indicators (FSIs); these ratios are analyzed on a time-series and cross-sectional basis, and are also used in their CAEL/CAMELS ratings.79

132. The BNB also performs regular on-site inspections, the frequency of which is based on the CAMELS ratings of individual banks, or if the off-site analyses discovers issues of concern. Full-scope audits are performed by on-site teams, which include information technology (IT) expertise.

133. The objective of this paper is to provide an overview of the BNB’s stress testing of the banking sector. Section B presents the stress test framework. This is followed by a discussion of the aggregate results for the banking sector in section C, along with the potential implications of test outcomes. Section D concludes with a few recommendations for development of the BNB’s stress testing framework moving forward.

B. The Stress Test Framework

134. The BNB uses a “bottom up” approach in its stress test of the banking sector, using detailed data which are submitted by banks on a regular basis. The effects of shocks on individual banks are initially estimated on a bank-by-bank basis, then on groups of banks and finally, on the banking system as a whole.80

135. The BNB’s model involves introducing shocks to selected variables in the profit and loss account and the balance sheet. Three types of risks are evaluated in the stress tests, namely: (i) credit risk, which is considered the most important risk for the local banking system; (ii) interest rate risk; and (iii) exchange rate risk. Each risk variable is initially “shocked” separately.

136. The BSD also tests separate scenarios incorporating all three risk variables. Three scenarios are assumed, each representing a progressive deterioration in the environment. The shock levels for exchange rates and interest rates remain the same in each scenario, while shocks to credit quality, derivatives and credit substitutes are intensified. A summary of the individual shocks and the three scenarios are presented in Table 2.

Table 2.

Bulgaria: Stress Test Variables and Shock Scenarios

(Shocks in percentage terms)

article image
Source: BNB.

137. The BNB’s current stress tests are based on very conservative assumptions of extreme outcomes. For the 2001-02 period, the stress tests are based on assumptions agreed between the BNB and the IMF during the 2002 FSAP. Since 2003, the BNB has tested sensitivities and scenarios based on its own research into the 1996-97 financial crisis. The tests assume a recurrence of the most severe movements in the exchange and interest rates, and the strongest deterioration in credit quality ever recorded for Bulgaria. For example, the assumption of a 60 percent exchange rate shock is based on the experience during the 19966-97 crisis when the exchange rate declined by the same magnitude against other currencies in the space of one day. Similarly, the most severe interest rate shock in history was a 192 percentage points rise in local interest rates in one day during that crisis. The design and selection of the parameters are based on research into different stress test models by a working group at the BNB.

Credit Risk

138. The test for credit risk examines the impact of a shock to credit quality on banks’ capital adequacy.81 In calculating the capital adequacy ratio following a credit shock, losses are deducted from the capital component (numerator) and from the risk-weighted assets component (RWA). Given that the capital amount is smaller, the impact of any loss has a greater impact on capital relative to RWA. Specific weights—based on historical evidence from the 1996-97 financial crisis—are applied to different categories of loans.82 The exposure in each loan category is weighted to determine the amount that would be subjected to a credit shock, which is defined by different migration percentages of standard loans. The stress tests also take into account the currency structure of the weighted loans portfolio, by applying multipliers to the currency exposures (including in euro), on a currency-by-currency basis.

139. For the standard loans, the BNB assumes a shock of a 10 percent migration to loss. Three scenarios with different migration percentages are also estimated: 1, 3 and 5 percent. The one percent shock scenario is considered the most realistic, but the others are estimated to be conservative. Each percentage of migration is equal to the amount of provisions additionally expensed to the Profit and Loss Account of a bank. If the net profit/loss is a loss, it is deducted from Own Funds. The rate of migration from each classification to loss was raised in Q2 2004, specifically, for watch (30 percent to 45 percent) and substandard (50 percent to 75 percent) loans.

Interest Rate Risk

140. A gap analysis or “mismatch model,” based on the difference between the flow of interest earned by a bank on its assets and the flow of interest paid on its liabilities, is used to determine interest rate risk. Since January 2003, the Off-Site Supervision Directorate has obtained detailed data on the interest rate sensitive assets and liabilities of banks—both on- and off-balance sheet—by maturity “buckets” as well as by individual currencies. Selected shocks are applied on six-month cumulative maturity mismatches in the leva (192 percentage points), the euro (2 percentage points) and the U.S. dollar (2 percentage points), and to a group of all other currencies (5 percentage points). The interest rate shocks for both the euro and the U.S, dollar, of 2 percentage points each. The BNB considers the magnitude of this shock to be significant in the context of the monetary policies adopted by the European Central Bank and the U.S. Federal Reserve, where policy adjustments are usually 25 basis points at a time.83

Exchange Rate Risk

141. The exchange rate shock represents the biggest one-day depreciation in the leva in history. The 60% shock is applied to the net open foreign exchange position of each bank. With the euro excluded from the calculation of the net open position, it virtually consists of the U.S. dollar open position since the amounts in other currencies (the British pound, the Japanese yen, etc.) are negligible.

Other Risks

142. Both credit substitutes and derivatives are also tested for credit risk. They are also subject to foreign exchange shock, if they are part of the net open currency position, while interest sensitive derivatives are subject to interest rate shocks as well. Exposures to credit substitutes and derivatives are assigned higher shock multipliers, in Scenarios 2 and 3.84 The objective is to appropriately compensate for the underdeveloped markets for such products.

143. The BNB estimates market risk in banks’ balance sheets to be insignificant. The share of the market risk equivalent in total risk-weighted assets (RWA) is only 3.8 percent. As such, this risk is omitted from the stress test model.

C. Stress Test Results


144. Stress test results are provided by the BNB for the 2003-05 period. The results are presented for Group 1 and Group 2 banks, and for banks grouped according to their foreign and domestic ownership (Table 3). The results for 2001 and 2002 are not presented, since they are not comparable given the different assumptions used during that period, and the information available at the time.

Table 3.

Bulgaria: Summary of BNB Stress Test Results by Group and Ownership, 2000-2005

article image
Source: Bulgarian National Bank.

Assumes 10 percent migration of standard loans to loss.

Notes: 1. Group 1 comprises the 10 largest banks; Group 2 is represented by 18 medium-sized and small banks. 2. Highlighted results denote outcomes with CARs below the minimum requirement of 12 percent.

145. The results show a stronger deterioration in the CAR of banks in 2004 and 2005, compared to 2003, for the same levels of credit shock. Further, the deterioration in 2004 is slightly worse than for 2005. In both cases, the CAR of the banking sector would have dropped well below the required minimum 12 percent.

146. Separate exchange rate and interest rate shocks appear to have little effect on the CAR of banks. The results are robust for all three years. In each case, shocks to either variable have resulted in changes in CAR of up to 2 percent, and all groups have remained well above the required 12 percent.

147. The scenario tests suggest that the capital in the banking sector would have been severely affected if a crisis of extreme magnitude had occurred in 2004 or 2005. The outcome for Scenario 1 suggests that some capital injection into the banking system would have been necessary. In both Scenarios 2 and 3, the CAR would have dropped well below zero.

148. The shocks consistently have a more negative effect on domestic banks relative to foreign banks. In each shock scenario, the impact on the CAR of domestic banks is substantially greater compared to their foreign counterparts, suggesting that domestic banks are more vulnerable, in aggregate, notwithstanding the fact that some domestic banks within the group are sound.

149. Group 1 banks have become increasingly more vulnerable to shocks than the Group 2 ones. This outcome became more distinct for 2005, compared to a relatively mixed outcome in 2004.

Potential Outcomes

150. The capital injection required to return banks to the minimum capital requirement of 12 percent has been calculated by the BNB. Specifically, estimates for the different magnitudes of credit shock are calculated for the aggregate banking sector, and then isolated to banks whose CAR would drop below 12 percent (Table 4). The BNB’s tests show that the most extreme credit shock outcome (10 percent migration of standard loans to loss) could cost up to 4.2 percent of GDP in terms of capital injections into problem banks. However, it should be noted that this estimate does not take into account the fact that banks have had to exclude current profits from their CAR calculations since April 2005.85 This amounted to an additional buffer of BGN 600 (1.5 percent of GDP) being excluded from CAR as at end-2005.

Table 4.

Bulgaria: Estimates of Capital Injection Required After a Credit Shock, 2005

(In percent of GDP)

article image
Source: BNB.

151. The required capital injection for scenarios 1, 2 and 3 are not presented. This is because the necessary amounts are not likely to be meaningful, given the extremity of the combined shocks/crises.86 That said, the authorities have made such estimates internally, and are fully cognizant of the amounts needed to recapitalize individual banks and the banking system as a whole, in the event that one of the extreme scenarios does occur.

152. In the event of one of the more extreme scenarios, the BNB would be unable to act as a lender of last resort, given the existing currency board arrangement. According to the Law on Banks, article 21 paragraph 2:

“The Central Bank shall in all cases revoke the license issued to a bank due to insolvency, where: (1) the bank fails to pay its obligations due for more than 7 days; or (2) the total of its liabilities exceeds the total of its assets”

Any decision subsequently taken by the government with regard to potential capital injections for banks would likely be made on a case-by-case basis.

D. Recommendations

153. The BNB’s “bottom up” stress tests of the banking sector include both sensitivity and scenario analyses. The model involves introducing certain shocks to selected variables in the profit and loss account and the balance sheet. These shock factors are consistent with those used by other central banks in that key risks—namely, credit, interest rate and exchange rate—are taken into account. In each test, very stringent assumptions are adopted, based on the extreme shocks to the banking system which have occurred in Bulgaria’s history. As a result, the BNB’s stress tests are very conservative as they essentially test for potentially severe crisis situations.

154. The BNB has indicated that it is constantly trying to enhance its modeling of the financial sector. This is in line with the practices by central banks in advanced economies, such as the U.K. and Norway.87 As a first step, the authorities plan to calibrate the stress test assumptions as applicable, and as more historical data become available.

155. Several enhancements would improve the robustness of the tests:

  • Testing for liquidity risk. This would provide a clear understanding of the potential stresses on the inter-bank market, given the limited capacity of the central bank to intervene under a currency board arrangement. From an individual bank perspective, it would be useful to determine the extent of a bank’s capacity to sustain a liquidity drain.88 As a subsequent step, the bank run could be combined with a sudden stop in capital inflows. The latter scenario is particularly pertinent, given banks’ dependence on overseas funding.89

  • Testing for other different scenarios. These could be based on the observed historical volatility of the variables being shocked (for example, two standard deviations of exchange rate fluctuations), in addition to using the 1996-97 crisis as a benchmark.

  • Testing for the market risk of a range of instruments (local and foreign bonds and equities). This will become more important, as they increasingly account for a bigger share of banks’ investment and trading portfolios.

156. The stress tests could also include broader scenarios of financial sector contagion, as more reliable data become available over time with the maturing of the economy and the development of the banking sector. For instance, they could incorporate contagion from bank runs, failure of a systemically important bank, or the impact of macro-economic shocks. Possible enhancements include:

  • Incorporating macro-economic shocks.90 One example is an adverse supply-side shock, such as a sharp increase in oil prices, which could impact domestic and foreign demand. This would likely result in the deterioration in the credit quality of corporates and, subsequently, households. The analysis could examine the direct impact on individual banks and through the inter-bank relationships.

  • Testing for contagion within the banking sector. Assume the failure of a systemically important bank within the system (e.g., from the second scenario above). The analysis could examine its impact on other banks, and the channels through which it occurs.

Table A.1.

Examples of Stress Tests in Recent European Financial Stability Reports

article image
article image
Source: Cihak and Hermanek (2005), based on central banks’ recent financial stability reports. Austrian National Bank: Financial Stability Report 7, June 2004. Danmarks Nationalbank: Financial Stability 2003. Deutsche Bundesbank: Report on the Stability of the German Financial System, Monthly Report, October 2004. De Nederlandsche Bank: Overview of Financial Stability in the Netherlands, December 2004, Issue No. 1. Hungarian National Bank: Report on Financial Stability, June 2003. National Bank of Poland: Financial Stability Review, First Half of 2004. Norges Bank: Financial Stability, 2004:1, June 2004. Sveriges Riksbank: Financial Stability Report, 2004:2.1)The latest FSR contained the stress tests carried out by (or in collaboration with) an FSAP mission.2)Based on end-2003 FSR. The subsequent two FSRs present only the “stress CAR,” which shows a bank’s financial position in a situation where all NPLs are written off.


  • Bank for International Settlements, 2005, “Stress Testing at Major Financial Institutions: Survey Results and Practice,” Committee on the Global Financial System (Basel, January).

    • Search Google Scholar
    • Export Citation
  • Blaschke, Winfrid, Matthew T. Jones, Giovanni Majnoni and Soledad Martinez Peria, 1988, “Stress Testing of Financial Systems: An Overview of Issues, Methodologies and FSAP Experiences,” IMF Working Paper No. 01/88 (Washington, International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Bunn, Philip, Alastair Cunningham and Mathias Drehmann, 2005 “Stress Testing as a Tool for Assessing Systemic Risks,” Financial Stability Review, June 2005 (London: Bank of England).

    • Search Google Scholar
    • Export Citation
  • Cihak, Martin, 2004, “Designing Stress Tests for the Czech Banking System,” CNB Internal Research and Policy Note No. 3/2004 (Prague: Czech National Bank).

    • Search Google Scholar
    • Export Citation
  • Cihak, Martin and Jaroslav Hermanek, 2005, “Stress Testing the Czech Banking System: Where are We? Where are We Going?” CNB Internal Research and Policy Note No. 2/2005 (Prague: Czech National Bank).

    • Search Google Scholar
    • Export Citation
  • Eklund, Trond, Kai Larsen and Eivind Bernhardsen, 2001, “Model for Analyzing Credit Risk in the Enterprise Sector,” Economic Bulletin 3/2001 (Oslo: Norges Bank).

    • Search Google Scholar
    • Export Citation
  • Goodhart, Charles and Lea Zicchino, 2005, “A Model to Analyze Financial Fragility,” Financial Stability Review, June 2005 (London: Bank of England).

    • Search Google Scholar
    • Export Citation
  • Jones, Matthew T., Paul Hilbers and Graham Slack, 2004, “Stress Testing Financial Systems: What to Do When the Governor Calls,” IMF Working Paper No. 04/127 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Norges Bank, 2004, Financial Stability (Oslo, June).


Prepared by Li Lian Ong, with input from Martin Cihak (both MFD) and the Banking Supervision Department of the Bulgarian National Bank.


In 2004, the Committee on the Global Financial System (CGFS) initiated an exercise on stress tests undertaken by banks and securities firms. The objectives of the exercise were to determine the main risk scenarios for financial institutions, and to explore how stress testing practices have evolved over time (BIS, 2005).


Cihak and Hermanek (2006) provide a cross-country comparison of stress tests presented by central banks in their recent financial stability reports (FSRs). They review 36 recent FSRs, focusing on the features of the respective stress tests (see Table A.1). Virtually all the stress tests presented in the respective FSRs are based on bank-by-bank data. Countries in the survey include Australia, Austria, Belgium, Brazil, Canada, Chile, Denmark, Euro Area, Estonia, Finland, France, Germany, Hungary, Hong Kong, Indonesia, Iceland, Indonesia, Ireland, Italy, Israel, Latvia, Luxembourg, Netherlands, New Zealand, Norway, Philippines, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden, Switzerland and the United Kingdom.


Blaschke, Jones, Majnoni and Martinez Peria (1988) discuss the stress-testing analyses conducted for FSAPs.


That said, not many banks actually do their own stress-testing, and even then, the models are not sophisticated. Some banks also apply value-at-risk (VaR) models, which are also available to on-site supervisors.


According to the BNB, it would strongly discourage the use of IRB models, with the advent of Basel II, since most banks in the system are unlikely to have sufficient historical data to credibly implement these models just yet.


EWS could, to some extent, be used as an input into stress tests (Cihak, 2004).


CAEL refers to: Capital adequacy, Asset quality, Earnings and Liquidity.

CAMEL refers to: Capital adequacy, Asset quality, Management, Earnings and Liquidity.


See Table 3 for bank groupings.


It should be noted that any deterioration in loan quality leads to a change in RWA. RWA are affected when testing for changes in credit risk; RWA remains unchanged when interest rate and exchange rate risks are tested.


For example, loans to budget (mostly loans to municipalities, amounting to 0.2 percent of total loans), commercial real estate and construction loans, other commercial loans, agricultural loans, consumer loans, residential mortgage loans to individuals, among others.


The stress-testing model agreed between the BNB and the 2002 FSAP mission incorporated a 50 basis points change in the leva interest rate only.


Derivatives instruments include foreign exchange, interest rate, equity, commodity and other derivatives contracts. Shocks are applied to the positive (marked-to-market) values of these instruments.


This is one of the measures introduced by the BNB to restrain credit growth.


The recapitalization amount for Scenario 1 appears to be somewhat similar to the credit shock of a 10 percent migration of standard loans to loss.


The Bank of England’s stress-test model has changed significantly since the 2002 FSAP (see Bunn, Cunningham and Drehmann, 2005; Goodhart and Zicchino, 2005). The new model is built from micro foundations, with core (theoretical) and non-core (set of equations which fit data better and pick up correlations) components, whereas the old model was more data driven. Norges Bank (2004) indicates that the central bank plans to develop its existing SEBRA model (see Eklund, Larsen and Bernhardsen, 2001), which predicts bankruptcy probabilities based on annual accounts figures for all Norwegian limited companies, by incorporating some market indicators.


According to the BNB, it currently excludes explicit testing of liquidity risk in the banking sector, given that liquidity in the system is very high. According to BNB estimates, the banking system would be able to cover up to around 30 percent of deposits at present, in the event of a bank run.


According to the BNB, its stress tests take this potential scenario into account by excluding overseas funding from the calculation of relevant ratios.