Greece: Selected Issues

This Selected Issues paper examines three areas of key policy importance for Greece in the coming years. The paper documents the loss of international competitiveness in recent years, as well as the accompanying widening of the current account deficit. It analyzes fiscal consolidation episodes in advanced economies, and confirms the conclusion found in the literature that, for durable consolidation, control of current spending is superior to revenue increases. The paper also estimates European banks’ vulnerability to rapid credit growth and economic slowdowns.

Abstract

This Selected Issues paper examines three areas of key policy importance for Greece in the coming years. The paper documents the loss of international competitiveness in recent years, as well as the accompanying widening of the current account deficit. It analyzes fiscal consolidation episodes in advanced economies, and confirms the conclusion found in the literature that, for durable consolidation, control of current spending is superior to revenue increases. The paper also estimates European banks’ vulnerability to rapid credit growth and economic slowdowns.

III. Credit Growth and Bank Vulnerability in the Euro Area and Greece31

A. Introduction

1. Since the mid-1990s, Greece has experienced a sizable increase in private sector credit. Between 1995 and 2005, real private sector credit in Greece rose by an average of 14½ percent a year, more than any euro-area country except Ireland. This rapid expansion has reflected a variety of factors, including low levels of financial development, pent-up demand pressures following years of credit controls, financial sector deregulation, and low real interest rates.

A03ufig34

Real Private Sector Credit

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

A03ufig35

Financial Catching-Up

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Sources: European Central Bank; IMF, International Financial Statistics; and staff calculations.

2. Although Greece’s financial depth remains below that in the euro area, the rapid credit expansion of recent years has raised concerns about risks. Greece’s private sector credit-to-GDP ratio, at 60 percent in 2005, is the lowest in the euro area, but nonperforming loans (NPLs) are twice as high as in the euro area.32 This could indicate that strains may have already risen as a result of the fast credit expansion.

3. This chapter analyzes the effect of rising credit growth on the vulnerability of the Greek banking sector in relation to that in the rest of the euro area. A growing body of the literature has found that rapid credit growth is associated with episodes of financial distress.33 The argument is that during a lending boom leverage increases and bad projects obtain financing, either because monitoring becomes more difficult, due to the large volume of lending, or because the borrower’s higher net worth turns out to be due to an asset price bubble. As exposure increases, the asset quality declines and the banking system becomes increasingly vulnerable. To test this hypothesis, this chapter examines the empirical relationship between credit growth and the vulnerability of banks in the euro area during 1994-2005, and addresses the following issues:

  • Are Greek banks more vulnerable to the pace of credit growth than other euro-area banks?

  • Does the relationship between credit growth and bank vulnerability change during downturns?

  • Are financial risks higher for credit growth rates above certain threshold values?

  • What are the main risks factors of the Greek banking system?

4. The main results are as follows:

  • In contrast to banks in other euro-area countries, the vulnerability of Greek banks increases with credit booms.

  • Credit growth has a larger and more immediate negative effect on Greek banks during real downturns and following an equity market bust.

  • The impact of credit growth is more pronounced for those banks in the euro area with faster credit growth. However, Greek banks become more vulnerable at lower credit growth rates than do others in the euro area.

  • Greek banks are exposed mainly to credit risks and persistently high NPLs suggest that banks’ risk management practices need to improve. In addition, the expansion into southeastern Europe may heighten risks over the medium term as banks start lending to local firms and households whose credit worthiness is less certain.

5. Promoting proper credit assessment and risk management by banks should, therefore, continue to be a policy priority for the authorities, given Greece’s rapid credit growth. The BoG has taken a proactive approach but should continue to carefully supervise banks’ risk management practices, ensure sufficient provisioning of NPLs, and maintain close cooperation with its counterparts in southeastern Europe.

6. The rest of the paper is organized as follows. Section B describes trends in credit to the private sector in Greece and compares them with those of other euro area countries. Section C discusses the model specifications and describes the data. Section D presents the empirical results. Finally, Section E analyzes the key risks in the Greek banking system.

B. The Greek Boom in Context

7. The lending boom in the Greek banking sector started in the mid-1990s, and since then, Greece has witnessed the most spectacular credit expansion among the euro area countries, after Ireland (Figure 1). In Greece, the credit expansion was not initially related to euro accession since no significant decrease in real interest rates occurred prior to that event (see, for example Brzoza-Brzezina (2005)). However, considerable financial liberalization of household lending took place. In particular, in 1991, commercial banks entered the mortgage market, which has previously been restricted to specialized banks and building societies. In 1993, consumer loans were allowed up to a total amount of €587, which in 1994 was extended to €23,500, including credit cards, although personal loans (i.e., loans without documents) could not exceed €2,935 (restrictions that were eliminated in 2003). In addition, foreign exchange controls were removed in 1993-94, resulting in a surge of lending (Honohan, 1999). Upon euro adoption, real rates dropped from 5½ percent in 1999 to about 1 percent in 2000 and into negative territory the following years. Strong credit growth in the last decade also reflects financial catching-up, given Greece’s low levels of financial intermediation compared to other countries in the euro area, and increased competition among financial institutions, both of which were fostered by liberalization.

A03ufig36

Greece: Private Sector Credit Growth

(In percent)

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Source: MF, International Financial Statistics.
Figure 1.
Figure 1.

Private Sector Credit Growth in the Euro Area, 1998-2006

(Annual percent change)

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Source: European Central Bank.

8. Initially, credit expansion was driven by both corporate and household lending although corporate loans added more because of their higher level (Figure 2). Since 2002, however, corporate credit growth has slowed, partly reflecting a switch to bonds. As a result, the ratio of business credit-to-GDP has remained at about 30 percent of GDP, one of the lowest in the euro area. Meanwhile, household credit has taken a leading role, with very high growth pushing the ratio of household debt to GDP from 14 percent of GDP in 2001 to 29 percent in 2005 (Figure 3). The increase in household debt of recent years has been driven by both mortgage and consumer credit, but, because of its low starting level, mortgage debt is one of the lowest in the euro area. By contrast, consumer credit is one of the highest, thanks to the unprecedented growth rates of personal loans and loans against documents. Since 2003, personal loans have increased by an average of 51 percent per year, while loans against documents have increased by 36 percent. Part of this lending has probably substituted for other sources of credit, in particular, credit cards. However, the data may not be fully comparable across countries: lumping consumer and other credit in one category, Greece has one of the lowest levels of this type of credit in the euro area.

Figure 2.
Figure 2.

Greece: Private Sector Credit, 1998-2006

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Source: Bank of Greece.
Figure 3.
Figure 3.

Euro Area: Private Sector Credit, 2005

(In percent of GDP)

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Sources: European Central Bank; IMF, World Economic Outlook; and staff’s calculations.

C. Methodology

Model

9. A bank’s vulnerability, or alternatively its stability, is measured by their distance to default (DD), which is increasingly used to assess soundness (De Nicoló and others, 2005; and Maechler, Mitra, and Worrell, 2006) because it is directly related to the probability of a loss exceeding equity capital. It can be summarized as

DD=k+μσ,(1)

where k is equity capital as percent of assets, μ is the average return on assets, and σ is the standard deviation of returns on assets, a proxy for return volatility.34 DD measures the number of standard deviations a return realization has to fall in order to deplete equity under the assumption that returns are normally distributed. Therefore, a higher level of DD implies less vulnerability, or greater stability.

10. Stability is modeled as a function of credit growth and various macroeconomic and bank-specific factors. A lagged dependent variable is included to allow for the possible persistence of financial stability. Also, two lags of the credit growth variable are included since it takes time for the vulnerabilities created by a credit boom to surface.35 Following Tamirisa and Igan (2006), a parsimonious baseline specification is selected by sequentially testing the significance of various macroeconomic and bank-specific variables identified in the recent literature as determinants of bank soundness.36 The parsimonious baseline specification is

FSi,t=αi+δt+β1FSi,t1+β2Crediti,t1+β3Crediti,t2+β4Crediti,t1*dGRC+β5Crediti,t2*dGRC+β6Macrojt1+β7Banki,t1+ui+ϵi,t,(2)

where i indexes banks, j indexes countries, and i, indexes years. FSi,t is the indicator of stability (i.e. the DD for each bank at each observed point); Credit is the real credit growth; Macro is a set of macroeconomic variables (real GDP growth and GDP per capita); Bank is a set of bank-specific variables (cost-to-income ratio and bank size); μi are firm-specific fixed effects; and εit is a serially uncorrelated error term. As discussed above, the effect of credit growth is expected to change over time. In the short term, credit growth could have a positive effect on financial stability if new loans are highly profitable and risks take time to materialize. Over time, however, loan growth, if not properly managed, may increase credit risks. Positive macroeconomic conditions (measured by higher real GDP growth) and higher level of economic development (real GDP per capita), greater bank efficiency (lower cost-to-income ratio), and larger bank size should also increase stability.

11. Equation (2) is estimated with the generalized method of moments (GMM) difference estimator. The fixed-effects estimator explicitly controls for bank-specific effects. However, even though the within transformation eliminates the ui S, by construction the transformed error term (ɛi,t1TΣt=1Tɛi,t) is still correlated with the lagged dependent variable. The bias (which influences all variables) is a function of T, and as only T tends to infinity the within estimator of β becomes consistent. In addition, some regressors are endogenous. In particular, credit growth is subject to two-way causality, as banks’ financial stability supports loan growth, and loan growth, in turn, may create vulnerability.37 To solve these issues, a GMM-difference estimator developed by Arellano and Bond (1991) is used. This estimator takes the first difference of each variable to eliminate the bank-specific effects and then uses lagged levels of the variables as instruments. Consistency depends on the assumption of no serial correlation in εi,t and the validity of instruments.38 Two tests suggested by Arellano and Bond (1991) are used to check the validity of our assumptions. The first test looks at whether the error term is second-order serially correlated. The second test is a Sargan test of overidentifying restrictions, where the null hypothesis is that the instruments are uncorrelated with the residuals. Failure to reject the null hypothesis of both tests would support the assumptions.

Data

12. Bank-specific data come from the Bankscope database and cover the period 1990-2005, which after accounting for the lagged variables, leaves an estimation period of 1994-2005. All commercial banks located in the euro area for which data for the period were available are included.39 This yielded a total of 1,009 banks, with about six observations per bank on average, although coverage for Greece is more limited, with only about five observations per bank, and there are no observations for 2005 because of a change of accounting standards (Table 1). The number of banks is substantially lower at the beginning of the period, largely because Bankscope is a relatively new data set with improving coverage over time. Macroeconomic data on real GDP and GDP per capita were taken from the IMF’s World Economic Outlook. House prices data come from the BIS with the exception of Austria, Portugal and Greece, which come from national sources. Table 2 presents summary statistics of the variables used in the econometric analysis. As with most empirical work at the level of disaggregation of this chapter, results should be interpreted with caution in light of issues related to data quality and consistency. Concerns relate partly to comparability of the data (different accounting standards, for example) and to potential biases related to the entry and exit of banks in the sample.

Table 1.

Number of Banks in the Sample by Country and Year

article image
Sources: Bankscope; and staff estimates.
Table 2.

Summary Statistics

article image

Excluding Greece.

Sources: Bankscope; and staff estimates.

13. Mean values of distance to default seem to indicate that the Greek banking system became more vulnerable after the introduction of the euro (Figure 4). Although vulnerability rose in the euro area as well, probably as a result of the global slowdown, the increase was not as pronounced as in Greece. Nevertheless, Greek banks have since recovered some of their lost ground. Incidentally, Greece’s bank-by- bank credit growth took off dramatically following the recovery from the 2000 slowdown, aided by easing global liquidity and low interest rates, and remained above the euro area average. Bank efficiency, measured by the cost-to-income ratio, seems to be comparable to the euro area average, but bank size is larger. Finally, real GDP growth has been stronger than in the euro area; this reflects the catch-up of the Greek economy, as GDP per capita is still below the euro area average.

Figure 4.
Figure 4.

Greece and the Euro Area: Mean Values by Year 1/

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Sources: Bankscope; IMF, World Economic Outlook; and staff estimates.1/ Legend applies to all figures. For the definition of variables, see Appendix I.2/ Excluding Greece.

14. There is a significant dispersion in the DD and credit growth at the bank level. The distribution of DD is asymmetric and skewed toward positive values (Figure 5). Although, the distribution of credit growth values is more symmetric, it also has fat tails, reflecting rapid credit expansion. The data also suggest little relationship between DD and credit growth for euro area banks. However, there seems to be a positive, though weak, relationship between DD and the first lag of credit growth of Greek banks (Figure 6).

Figure 5.
Figure 5.

Histograms for Distance to Default and Bank Credit Growth, 1994-2005

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Figure 6.
Figure 6.

Correlation between Distance to Default and Bank Credit, 1994-2005

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Sources: Bankscope; and staff’s calculations.1/ Excluding Greece.

D. Econometric Results

Baseline specification

15. Table 3 reports the estimation results of the dynamic financial stability model described in equation (2), focusing on one-stage robust estimates that have been corrected for heteroscedasticity, using a maximum of two lags of each of the explanatory variables as instruments.40 For all regressions, there is no sign of second-order serial correlation of the first-differenced residuals, and the Sargan test accepts the null hypothesis that the overidentifying restrictions are valid. The only exception is the baseline specification estimated for the period 1999-2005, where the null hypothesis of no overidentifying restrictions is rejected and the estimator is therefore inconsistent.

Table 3.

Baseline Specification 1/

article image
Notes:

Robust z statistics in parentheses.

+ significant at 10%; * significant at 5%; ** significant at 1%.

Dependent variable is FSt Column (1) reports the estimates for the whole sample, while column (2) reports estimates for the period 1999-2005. All bank-specific and macro variables are lagged by one year to avoid simultaneity problems.

Sargan test is the χ2 statistics of a test of the null hypothesis that the overidentifying restrictions are valid. Statistics are based on two-step estimator. P-values are reported in parentheses.

Serial correlation is the Ζ-statistic from a test of the null hypothesis of no second-order serial correlation in the residuals. Ρ-values are reported in parentheses.

16. The estimates suggest that credit growth boosts the profitability of Greek banks in the short term but erodes stability over a longer horizon. That is, stability increases with the first lag of credit growth, as with euro area banks, but deteriorates two years after credit increases, although the effect is small (Table 3). Segoviano Basurto, Goodhart, and Hofman (2006) also find that there is a lag between the time a lending boom takes place and the time financial fragility materializes. These results suggest that, while banks in the euro area may be able to contain risks and/or build up sufficient capital as credit growth increases, that is not the case in Greece. As expected, stability shows some persistence: banks that are sound today are more likely to be sound tomorrow. Surprisingly, however, vulnerability rises with cost efficiency and size, although this result is consistent with De Nicoló (2000), who finds that insolvency risk increases in size for banks in industrialized countries. Finally, GDP per capita and real GDP growth increase stability. These results largely hold for the period 1999-2005, although recall that these estimates are inconsistent.

17. House price inflation has a short-term positive effect on the financial stability of Greek banks. Credit growth in Greece has largely been driven by household loans, of which 65 percent are mortgages, while real property prices have risen by an average of 5.8 percent since 2001. To analyze whether a combination of lending and property price increases may raise the likelihood of financial fragility, an interaction term between credit growth and house price inflation is included in equation (2). The results suggest that the stability of Greek banks increases as credit growth and house price inflation increase, but no effect is evident for euro-area banks (Table 4). One explanation is that the Greek credit boom does not appear to be related to a housing price bubble.

Table 4.

Baseline Specification including Real House Price Inflation 1/

article image
Notes:

Robust z statistics in parentheses.

+ significant at 10%;* significant at 5%; ** significant at 1%.

Dependent variable is FSt. Columns (1) reports the estimates for the whole sample, while column (2) reports estimates for the period 1999-2005. All bank-specific and macro variables are lagged by one year to avoid simultaneity problems. Rt is a dummy variable that indicates if there was a stock market downturn at time t.

Sargan test is the Χ2 statistics of a test of the null hypothesis that the overidentifying restrictions are valid. Statistics are based on two-step estimator. Ρ-values are reported in parentheses.

Serial correlation is the Ζ-statistic from a test of the null hypothesis of no second-order serial correlation in the residuals. Ρ-values are reported in parentheses.

Asymmetric effects

18. To test the presence of asymmetric effects of credit growth over the cycle, two types of downturns are considered: real sector downturns and stock market busts. Following Ruiz-Arranz (2003) and Vermeulen (2002), a real downturn occurs if industrial production falls. Table 5 presents data on industrial production for the euro-area countries, with the shaded areas identifying downturn years. For example, Greece experienced industrial production declines in 2001 and 2005. A stock market bust is defined as a decline in the stock market index. According to this measure, Greece experienced an equity market downturn in 1995, and during 2000-2003 (Table 6). Asymmetric effects are captured by modifying equation (2) to allow a different parameter on credit growth during a downturn:

Table 5.

Industrial Production Growth Rates

(In percent)

article image
Sources: Eurostat; and National Statistical Offices.
Table 6.

European Stock Market Indices

article image
Source: Bloomberg.
FSi,t=αt+δt+β1FSi,t1+(β22Rt)Crediti,t1+(β31+β32Rt)Crediti,t2++(β41+β42Rt)Crediti,t1*dGRC+(β51+β52Rt)Crediti,t2*dGRC+β6Macrojt1+β7Banki,t1+ui+ϵi,t,(3)

where Rt is a dummy variable that indicates a downturn at time t and βk2 measures the existence of asymmetric effects.

19. Credit growth seems to have a larger and more immediate negative effect on the stability of Greek banks during real downturns. Table 7 (column 2) indicates that Greek banks with higher credit growth are more likely to suffer a deterioration in their financial stability following a downturn, but only in the 1999-2005 period, perhaps owing to the lack of a real downturn before that. These results should be interpreted with caution, particularly, because industrial production might not be a good measure of economic downturns in countries like Greece, where the manufacturing sector is small and where real GDP growth has been consistently strong. Econometric results using a measure of the output gap (not shown) indicate that credit growth has a larger negative impact during downturns, but the effect is not immediate. Surprisingly, credit growth seems to improve the stability of euro-area banks during downturns (Table 7, columns 1 and 2). One possible explanation is that banks anticipate the slowdown and tighten their credit standards.41

Table 7.

Asymmetric Effects over Real Sector Downturns 1/

article image
Notes:

Robust z statistics in parentheses. + significant at 10%; * significant at 5%; ** significant at 1%.

Dependent variable is FSt Columns (1) and (3) report the estimates for the whole sample, while columns (2) and (4) report estimates for the period 1999-2005. All bank-specific and macro variables are lagged by one year to avoid simultaneity problems. Rt is a dummy variable that indicates if there was a real downturn at time t. LD denotes the ratio of loan to deposit and short-term funding.

Sargan test is the Χ2 statistics of a test of the null hypothesis that the overidentifying restrictions are valid. Statistics are based on two-step estimator. Ρ-values are reported in parentheses.

Serial correlation is the Ζ-statistic from a test of the null hypothesis of no second-order serial correlation in the residuals. Ρ-values are reported in parentheses.

20. The outstanding amount of lending seems to be an important determinant of financial stability of banks in the euro area. As loan growth increases, it may outstrip deposit growth, resulting in an increased reliance on nonretail and potentially more costly and volatile funding sources, such as capital market issues and borrowing in the interbank market. This outcome may increase banks’ vulnerability by cutting margins and increasing dependence on cross-border flows with risks of availability and cost. To test the effect of the amount of lending, equation (3) is re-estimated allowing the interaction of credit growth with the ratio of loans to deposits and short-term funding (Table 7, columns 3 and 4). After controlling for lending size, the credit growth of euro-area banks reduces stability during downturns. That is, the higher the level of lending relative to deposits, the more likely credit growth will weaken bank soundness during a downturn. The results for Greek banks are not statistically different from those of banks in other euro area countries.

21. Credit growth seems to have a larger and more immediate negative effect on Greek banks’ stability during stock market busts as well. Table 8 indicates that the first lag of credit growth has a negative impact on the stability of Greek banks when stock market valuations are falling, contrary to the findings for euro-area banks. However, the second lag of credit growth has a positive effect on the financial stability of Greek banks during stock market downturns (but only for years after 1998).

Table 8.

Asymmetric Effects over Equity Price Cycles 1/

article image
Notes:

Robust z statistics in parentheses. + significant at 10%; * significant at 5%; ** significant at 1%.

Dependent variable is FSt. Columns (1) reports the estimates for the whole sample, while column (2) reports estimates for the period 1999-2005. All bank-specific and macro variables are lagged by one year to avoid simultaneity problems. Rt is a dummy variable that indicates if there was a stock market downturn at time t.

Sargan test is the Χ2 statistics of a test of the null hypothesis that the overidentifying restrictions are valid. Statistics are based on two-step estimator. Ρ-values are reported in parentheses.

Serial correlation is the Ζ-statistic from a test of the null hypothesis of no second-order serial correlation in the residuals. Ρ-values are reported in parentheses.

Threshold effects

22. The effect of credit growth may become more intense when credit growth exceeds a certain threshold. To test this hypothesis, the banks with the fastest credit expansion, that is, those banks in the upper decile of the credit growth distribution, are isolated:

FSi,t=αi+δt+β1FSi,t1+(β21D075C+β22D7590C+β23D90100C)Crediti,t1+(β31D075C+β32D7590C+β33D90100C)Crediti,t2+(β41D075C+β42D7590C+β43D90100C)Crediti,t1*dGRC++(β51D075C+β52D7590C+β53D90100C)Crediti,t2*dGRC++β6Macrojt1+β7Banki,t1+ui+ϵi,t,(4)

where D075c,D7590c, and D90100c are dummy variables for observations below the 75th percentile, between the 75th and 90 percentiles, and above the 90th percentile, respectively, of the distribution of credit growth.42

23. The results suggest that credit growth has a more pronounced impact on vulnerability for banks with higher credit growth (Table 9). Those with credit growth above the 90th decile have a lagged negative effect on the financial stability of Greek and euro area banks. This result, however, is not robust to changes in the estimation period in the case of euro-area banks. In addition, the immediate effect of credit growth above the 75th percentile on financial soundness is positive for Greek and euro-area banks but negative in the longer term for Greek banks. This difference between Greek and euro-area banks could indicate that lending standards are more relaxed in Greece and, therefore, vulnerabilities start to build up in Greek banks at lower rates of credit growth.

Table 9.

Threshold Effects1/

article image
Notes:

Robust z statistics in parentheses. + significant at 10%; * significant at 5%; ** significant at 1%.

Dependent variable is FSt. Columns (1) reports the estimates for the whole sample, while column (2) reports estimates for the period 1999-2005. All bank-specific and macro variables are lagged by one year to avoid simultaneity problems. D1 D2 D3 are dummy variables for observations below the 75th percentile, between the 75th and 90 percentiles, and above the 90th percentile, respectively, of the distribution of credit growth.

Sargan test is the Χ2 statistics of a test of the null hypothesis that the overidentifying restrictions are valid. Statistics are based on two-step estimator. Ρ-values are reported in parentheses.

Serial correlation is the Ζ-statistic from a test of the null hypothesis of no second-order serial correlation in the residuals. Ρ-values are reported in parentheses.

E. Key Risks

24. Financial soundness indicators suggest that the banking system is profitable and well capitalized (Table 10). Solvency remains satisfactory, and profits are robust, driven by rising lending volumes in Greece and southeastern Europe, wide margins, and some cost cutting through the rationalization of branch networks (resulting from mergers), investments in IT, and the implementation of voluntary retirement plans in some banks. Solvency ratios have been supported by the increase in capital of certain credit institutions and the gradual increase of subordinated and hybrid capital in total own funds. Capital adequacy ratios fell slightly in the first half of 2006, however, because risk-weighted assets grew faster than own funds.

Table 10.

Core Set of Financial Soundness Indicators for Deposit-Taking Institutions 1/

(1998-June 2006, unless otherwise indicated)

article image
Source: Bank of Greece.

These may be grouped in different peer groups based on control, business lines, or group structure.

June 2006 figures refer to Greek quoted banks unless otherwise indicated.

June 2006 figures refer to all banks

Data on a consolidated basis.

This figure does not include ATE bank and rescheduled loans. If rescheduled loans were included, then the relevant ratios would become 18.7 for 2005 and 20 for June 2006.

This figure does not include ATE bank and rescheduled loans. If rescheduled loans were included, then the relevant ratios would become 5.6 for 2005 and 5.6 for June 2006.

2006 figures refer to July 2006 and to all banks on a nonconsolidated basis (i.e., commercial, cooperative, and foreign branches).

On a nonconsolidated basis. From 2004 on, in accordance with IFRS.

25. However, the strong credit expansion of recent years, which is well above deposit growth, has increased Greek banks’ exposure to liquidity and refinancing risks. Although liquidity is satisfactory—the loan-to-deposit ratio was 97 percent in Greece in 2005, compared with 122 percent in the euro area in 2004—Greek banks are relying increasingly on the interbank money markets, as well as on the issuance of bank bonds and other debt securities for funding (Figure 7). Nevertheless, Greek banks have started to diversify their funding into new products like euro commercial paper, euro medium-term notes, securitization, and hybrid capital with longer maturities.

Figure 7.
Figure 7.

Greek Banks: Funding Structure

(In percent of total, unless otherwise indicated)

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Source: Bank of Greece.

26. Credit risk exposure is also a concern, as credit growth continues to increase at a fast pace. Private sector credit growth strengthened in the first nine months of 2006, driven by the continued acceleration in residential mortgages that was evident in 2005.43 Strong credit expansion to the private sector, notably households, boosts banks’ profitability, but is also a potential source of credit risk, especially in the event of an economic downturn or a further rise in euro area interest rates. Since floating-rate loans make up the bulk of loans to households, an interest rate increase would directly push up loan servicing costs for many inexperienced borrowers and, therefore, could worsen banks’ credit risk. Also, according to the 2005 household indebtedness survey, 12 percent of households accounting for 30 percent of debt (largely housing loans) had debt- service ratios above 40 percent in 2005. These households could be under significant financial stress if interest rates rise sharply, magnifying the impact of the shock.

A03ufig37

Debt service-to-income ratio

(In percent)

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

Source: Bank of Greece.

27. Moreover, the NPL ratio has remained stubbornly high. After improving in 2005, the NPL ratio of Greek commercial banks, particularly for consumer loans, deteriorated in the first half of 2006 (while provisioning declined), suggesting that banks continue to lend to poor risks may be to increase their market share.44 This is particularly worrisome, considering the large volume of new loans that should be performing and the fact that the cycle has not yet taken a turn for the worse. In addition, although the business loans have grown moderately, the latest migration matrix data for first half of 2006 show a decline in the creditworthiness of the corporate sector, with net downgrades accounting for 6 percent of total corporate loans. Large exposures in Greek banks increased from 145 percent in 2005 to 151 percent in the first nine months of 2006. Nonetheless, some of the larger corporates are utility companies with relatively stable revenue sources and subtle government backing.

A03ufig38

Greek Banks: NPLs and Provisioning

(Ratios in percent)

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

A03ufig39

Greek Banks: Nonperforming Loans

(Ratios in percent)

Citation: IMF Staff Country Reports 2007, 027; 10.5089/9781451816280.002.A003

28. The expansion of Greek banks into southeastern Europe through mergers and acquisitions carries benefits but can also involve risks. In 2005, Greek banks accounted for 15.3 percent of southeastern Europe’s market share. These acquisitions have expanded the deposit base and could help diversify earnings and improve Greek banks’ scale and efficiency. However, such geographic expansion involves funding, operational, and country risks:

  • Funding risk. The major banking groups have funded their foreign acquisitions mainly through their capital accumulation and the issuance of subordinated debt and hybrid capital. So far, most of the acquisitions of Greek banks have been gradual (extended over a few years) and the disbursements made in relatively small quantities, not requiring equity issuance. Although securitization has so far been minimal, Eurobank EFG has expressed its intention to securitize part of its small and medium-sized enterprise (SME) loan portfolio (around €2 billion) and use some of the proceeds to finance acquisitions abroad. Going forward, however, the expansion into new markets may require greater reliance on more volatile wholesale funding.

  • Operational risk. This risk arises from errors in trading activities or outright fraud that goes undetected because of a lack of proper internal controls. To mitigate this risk, parent banks in Greece should carefully monitor internal controls and corporate governance of the foreign subsidiaries. According to IMF(2006), the Bank of Greece (BoG) inspectors have found that risk management capabilities in the Greek banks’ subsidiaries may be insufficient.

  • Country risk. Southeastern Europe is a riskier banking environment, rendering Greek banks more vulnerable to adverse developments in the region.

29. The risks highlighted above are somewhat mitigated by the following factors:

  • Liquidity risk. The BoG introduced two compulsory minimum liquidity ratios in 2005. First, the liquid asset ratio, stipulates that the ratio of banks’ liquid assets maturing in up to 30 days to the cumulative balance of bank deposits maturing in up to 12 months, should exceed 20 percent. Second, the mismatch ratio, stipulates that the ratio of the difference between banks’ total assets and total liabilities maturing in up to 30 days to the cumulative balance of bank deposits maturing in up to 12 months should be higher than -25 percent. During the first half of 2006, the Greek banking system was well above those limits.

  • Credit risk.

    1. Recent stress tests conduced by the BoG indicate that a 30 percent increase in the probability of default would cause losses that exceed the supervisory provisions, but also, given the buffer provided by the high levels of capitalization, the banking system as a whole would remain resilient.

    2. The BoG has intensified its monitoring of credit developments (loan-to-value ratios for mortgages, approval ratios and overrides for consumer loans and credit cards, and the ratio of monthly installments to disposable income). Also, the BoG monitors banks’ exposures above €1 million to groups of companies.

    3. The BoG has increased the risk weights of mortgages with loan-to-value ratios above 75 percent. In addition, the debt-service ratio for new household loans should not exceed 30-40 percent of disposable income. Moreover, the BoG has increased provisioning requirements and write-offs.

  • Expansion into southeastern Europe.

    1. BoG closely monitors the expansion of Greek banks into the region through a multilayered supervisory framework. Among other initiatives, the BoG requires Greek banks that are active internationally to develop methodologies for assessing country risk and to form provisions in order to cover the relevant risks, (i.e., liquidity and foreign exchange) that are not being covered as yet under the harmonized European legislative framework.

    2. Recent stress tests, conducted for the loan books and bond portfolios of Greek banks in the region show that, even under a extreme scenario, the impact on the regulatory capital would be limited.

    3. Greece has signed memoranda of understanding with most southeastern European countries that allow the exchange of information, reports, and prudential returns, as well as on-site examinations by BoG supervisors.

Annex I. Data Sources and Definitions

The data set used in this paper was created in several steps:

1. Compile information for all banks in the euro area included in the Bankscope database published by the Bureau van Dijk.

2. Bankscope includes both consolidated and unconsolidated balance sheet data. To make sure that observations are not duplicated for the same bank, the following procedure was applied to include information from only one of the balance sheets. First, using the “consolidation code” variable in Bankscope we choose institutions, which will provide one balance sheet for each institution at the highest level of consolidation available. In a second step, we add those banks not included in the first step for which data are available.

3. Exclude outliers and unrealistic observations for the variables used to estimate the base specification. In particular, exclude individual observations where:

  • Equity is negative

  • FSi,t < 1st percentile or FSi,t > 99th percentile

  • Crediti,t-1<1st percentile or Credit i,t-1> 99th percentile

  • Crediti,t-2<1st percentile or Crediti,t-2> 99th percentile.

Variable definitions

FS: Financial stability. Return on average assets plus equity as a percent of assets divided by the standard deviation of return on average assets.

Credit: Bank credit growth. Annual percentage change in total loans in real terms (credit deflated by the GDP deflator).

Cost-to-income: Cost-to-income ratio. Total operating expenses divided by total operating income.

Size: Bank size. Logarithm of total real assets in U.S. dollars.

LD: Loan-to-deposit-and-short-term-funding ratio.

Real GDP per capita. Real GDP per capita, in U.S. dollars.

Real GDP growth. Annual growth rate of real GDP.

HP: House price inflation. Annual growth rates of the real house price index.

References

  • Arellano, M., and S., Bond., 1991, “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an application to Employment Equations,” Review of Economic Studies, Vol. 58 (April), pp. 27797.

    • Search Google Scholar
    • Export Citation
  • Asea, P., and S.B. Blomberg, 1997, “Lending Cycles,” NBER Working Paper No. 5951 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation
  • Brozá-Brzezina, M, 2005, “Lending Booms in the New EU Member States: Will Euro Adoption Matter?,” ECB Working Paper No. 543 (Frankfurt: European Central Bank).

    • Search Google Scholar
    • Export Citation
  • Cottarelli, C., G. Dell’Ariccia, and I. Vladkova-Hollar, 2005, “Early Birds, Late Risers, and Sleeping Beauties: Bank Credit Growth to the Private Sector in Central and Easter Europe and in the Balkans,” Journal of Banking and Finance, vol.29 (January), pp. 83104.

    • Search Google Scholar
    • Export Citation
  • Demirgiic-Kunt, A., and E. Detragiache, 1998, “The Determinants of Banking Crises: Evidence from Developing and Developed Countries,” IMF Staff Papers, Vol. 45 (March), pp. 81109.

    • Search Google Scholar
    • Export Citation
  • De Nicolo, G., 2000, “Size, Charter Value and Risk in Banking: An International Perspective,” Federal Reserve International Finance Discussion Paper No. 689 (Washington: Federal Reserve Bank).

    • Search Google Scholar
    • Export Citation
  • De Nicolo, G., 2005, “European Financial Integration, Stability and Supervision,” in Euro Area Policies: Selected Issues, IMF Country Report No. 05/266 (August), by A. Annett and others (Washington: International Monetary Fund), pp. 11346.

    • Search Google Scholar
    • Export Citation
  • Goldfajn, I., and R. Valdés, 1997, “Capital Flows and the Twin Crises: The Role of Liquidity,” IMF Working Paper 97/87 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Gourinchas, P., R. Valdés, and O. Landerretche, 2000, “Lending Booms: Some Stylized Facts” (unpublished; Princeton, New Jersey: Princeton University).

    • Search Google Scholar
    • Export Citation
  • Honohan, P, 1999, “Consequences for Greece and Portugal of the Opening-up of the European Banking Market” (unpublished; Washington: The World Bank).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2006, Greece: Financial System Stability Assessment, IMF Country Report No. 06/6 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Kaminsky, G., and C. Reinhart, 1999, “The Twin Crises: The Causes of Banking and Balance of Payments Problems,” American Economic Review, Vol. 89 (June), pp. 473500.

    • Search Google Scholar
    • Export Citation
  • Maechler, A., S. Mitra, and D. Worrell, 2006, “Exploring Financial Risks and Vulnerabilities in New and Potential EU Member States” (unpublished; Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Ruiz-Arranz, M., 2004, “Corporate Balance Sheets and Investment: Empirical Estimates for Portugal and Other EU Countries,” Portugal—Selected Issues, IMF Country Report No. 04/81 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Segoviano Basurto, M., C. Goodhart, and B. Hofman, 2006, “Default, Credit Growth, and Asset Prices,” IMF Working Paper 06/223 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Tamirisa, N. and D. Igan, 2006, “Credit Growth and Bank Soundness in New Member States” (unpublished; Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Vermeulen, P., 2002, “Business Fixed Investment: Evidence of a Financial Accelerator in Europe,” Oxford Bulletin of Economics and Statistics, 64, 3, pp. 213231.

    • Search Google Scholar
    • Export Citation
31

Prepared by Marialuz Moreno-Badia. I thank Nikolaos Stavrianou and Thomas Vlassopoulos for their valuable comments.

32

The national accounts were revised in September 2006, raising the level of output in 2000-2005 by about 26 percent. Data presented in sections A and B reflect the new GDP. The econometric results are based on the old national accounts, but the conclusions are the same if we use the revised data.

33

See, for example, Cottarelli, Dell’Ariccia, and Vladkova-Hollar (2005); Goldfajn and Valdes (1997); Demirguc-Kunt and Detragiache (1997); Kaminsky and Reinhart (1999); and Gourichas et al (2000).

34

To calculate this indicator, the market value of equity and assets and shareholders’ profits should be used. However, due to lack of data, the book values of all variables, derived from balance sheet data, are used instead.

35

In a recent study, Segoviano Basurto, Goodhart, and Hofman (2006) find that a combination of bank lending and property booms increases the likelihood of financial fragility two to three years after a boom.

36

The original set of macroeconomic variables considered was (i) GDP per capita; (ii) real GDP growth; (iii) real interest rates; (iv) real exchange rate depreciation; (v) credit-to-GDP ratio; (vi) real house price inflation; and (vii) unemployment. For the bank-specific variables, measures of bank profitability, liquidity, efficiency, and risk (proxied by the net interest margin, liquidity ratio, cost-to-income ratio, loan-to-deposit ratio, and ratio of loan loss reserves to gross loans) were used.

37

Although credit growth is lagged in the baseline specification, the Sargan test of overidentifying restrictions suggests that this variable is not exogeous.

38

If εi,t is not serially correlated, there should be evidence of significant first-order correlation in difference residuals (ɛi,t1TΣt=1Tɛi,t) but no evidence of second-order correlation in the differenced residuals.

39

For a detailed description of the data set and definitions, see Appendix I.

40

Using the full possible instrument matrix adds little explanatory power but may lead to finite sample bias.

41

There is evidence that banks change their lending standards (from laxity to tightness) systematically during the real business cycle. See, for example, Asea and Blomberg (1997).

42

In our sample, the 75th–90th range percentile is defined by credit growth rates between 18.9 and 37.2 percent.

43

Housing credit growth started to accelerate at end-2005 and has slowed only in recent months. This acceleration is partly due to the introduction of value-added tax (VAT) in newly built residential properties and the readjustment of objective values of the housing stock as of January 1. Consumer credit growth has started to accelerate in the second half of 2006 and remains high.

44

The decline in NPLs in 2005 was partly attributable to an impressive increase in bad-loan write-offs. Banks took advantage of a decision of the Bank of Greece that allowed them to set off a part of these write-offs against the provisioning shortfall, which, in turn, is deducted from own funds for the calculations of the capital adequacy ratio. Net of this effect, total NPLs (excluding restructured loans) would have risen by 22.7 percent year on year.