Risk-taking by Russian banks: do location, ownership and size matter? Текст научной статьи по специальности «Экономика и бизнес»

Risk-taking by Russian Banks: Do Location, Ownership and Size Matter?

Fungacova Z., Solanko L.

The Russian banking sector has experienced enormous growth rates during the last 6-7 years. The rapid growth of assets has, however, contributed to a decrease in the capital adequacy ratio, thus influencing the ability of banks to cope with risk. Using quarterly data spanning from 1999 to 2007 on all Russian banks, we investigate the relationship between bank characteristics and risk-taking by Russian banks. The analysis of financial ratios reveals that, on average, the risk levels are still below those observed in Central and Eastern Europe. Combining the group-wise comparisons of financial ratios and the results of insolvency risk analysis based on fixed effects vector decomposition, three main conclusions emerge. First, controlling for bank characteristics, large banks have higher insolvency risk than small ones. Second, foreign-owned banks exhibit higher insolvency risk than domestic banks and large state-controlled banks are, unlike other state-controlled banks, more stable. Third, we find that the regional banks engage in significantly more risk-taking than their counterparts in Moscow.

Keywords: bank risk-taking, banks in transition, Russia.

1. Introduction

Banking sectors in most countries of the Commonwealth of the Independent states (CIS), Russia included, have experienced nearly phenomenal growth rates during recent years. As a consequence of the dramatically improved macroeconomic situation

We are grateful for the valuable comments and suggestions we have received from Stephan Barisitz, Randall Filer, Michael Funke, Iftekhar Hasan, Esa Jokivuolle, Iikka Korhonen, Aaron Mehrotra, Tuomas Takalo, Laurent Weill, participants of the BOFIT seminar in Helsinki (March 2008), the IX International Academic Conference in Moscow (April 2008), the Sixth ESCB Workshop on Emerging Markets in Helsinki (May 2008) and the 10th EACES conference in Moscow (August 2008).

All opinions expressed are those of the authors and do not necessarily reflect the views of the Bank of Finland.

Fungacova Z. - Bank of Finland Institute for Economies in Transition (BOFIT). E-mail: zuzana.fungacova@bof.fi

Solanko L. - Bank of Finland Institute for Economies in Transition (BOFIT). E-mail: laura.solanko@bof.fi E-mail: laura.solanko@bof.fi

Статья поступила в Редакцию в январе 2009 г.

and important legislative changes, the ratio of banking sector assets in Russian GDP grew annually by more than 2 percentage points between 2001 and 2007. This ratio exceeded 60 percent by the end of 2007. Simultaneously, bank credit to the private sector has more than doubled to 30 percent of GDP.

With the rapid growth of total assets, deposits and loan stocks, Russian banks are increasingly assuming their role as financial intermediaries channeling household deposits and foreign borrowing into domestic corporate credits. This necessarily causes changes in the banks' assets and liability structures, attitudes towards risk-taking and risk management. Rapid credit growth is likely to increase (potential) banking sector risks. On the other hand, the ongoing financial deepening also indicates that the Russian banking sector is beginning to have an impact on private sector (both corporate and individual) behaviour and investments. That is, banks in Russia as well as in most other transition economies, are starting to look like banks elsewhere. They are by no means problem-free, but the challenges they need to tackle are similar to what banks in other emerging economies face. Given their growing role in economic development, surprisingly little is known about these banks' risk-taking behaviour.

The development of the banking sector in transition economies, as well as the financial sector in general, have been studied extensively. Barisitz (2008) and Bonin and Wachtel (2003) [5, 11] provide excellent recent overviews. Many studies focus on the effects of bank privatization on their performance in transition countries [8, 9], but until recently risk-taking by banks in transition has been a largely neglected area of research. Recent literature on the Russian banking sector has focused on bank supervision and the introduction of the deposit insurance system [14, 18, 42], market discipline and deposit interest rates [31, 39] and the efficiency of banks [1, 2, 19, 33] .

A handful of recent papers provide cross-country evidence on bank risk-taking in emerging economies. Haselmann and Wachtel (2007) [28] use several accounting measures of bank risk to examine the risk-taking behaviour of banks in 20 transition countries including Russia. They analyze differences in risk measures by bank ownership, size and market share. Using survey data from the EBRD, they complement the analysis with various measures of institutional quality. The results suggest that there is no group of banks with excessive risk-taking and that an unsound institutional environment leads to higher capital holdings and less credit risk-taking by banks. Maechler et al. (2007) [36] examine the effect of various types of financial risks on the bank stability in 18 Central and Eastern European economies. Their results indicate that foreign banks tend to have a higher risk profile than domestic ones but there is no significant difference between the risk profiles of larger and smaller banks. Furthermore, credit growth relates to greater bank stability and only the acceleration of growth seems to add vulnerability.

To the best of our knowledge, no study on bank risk-taking has focused on Russia or any other CIS country. However, with its 1100 banking institutions, Russia in particular provides an extremely rich test case for analyzing risk-taking. Additionally, the large number of bank failures (more than 300 since year 20001)) highlights the fact that banking in Russia is still riskier than in most developed countries. Therefore examining the determinants of risk-taking is crucial for understanding the prospects for future economic growth. Furthermore, if Russia is to become a global financial centre,

1) For more details see: www.banki.ru.

a goal clearly stated by, e.g., President Medvedev in spring 2008, we need to know much more about the behaviour of Russian banking institutions.

Currently the Russian banking sector is extremely fragmented, with a few large banks and a great number of very small ones. Especially in comparison with Central European transition economies, the state has retained a large share of control whereas the role of foreign banks has been very limited. These two structural features have often been mentioned as the main hindrances to further banking sector reform and growth. In this paper we discuss the extent to which the characteristic features of the sector determine the risk-taking behaviour of Russian banks.

We use a large panel of practically all Russian commercial banks covering the post-1998-crisis period, from April 1999 to April 2007. The large, Moscow-based and state-controlled banks form the backbone of the Russian banking sector. In line with previous literature, we therefore focus on the effects of bank size and ownership structure on bank risk-taking. Furthermore, we control for the location of the banks to see if Moscow-based banksdiffer in their risk-taking habits. Additionally, we are able to examine the influence of what probably was the most important institutional change during the period, the introduction of a deposit insurance scheme, on the risk-taking of Russian banks.

In measuring risk-taking, we use two approaches. First, we conduct a univariate analysis of traditional financial risk ratios based on accounting data. Second, we run a regression analysis of bank insolvency risk measured by the z-score indicator. The two approaches produce similar results. First, risk-taking increases with size. Second, controlling for other bank characteristics, banking institutions located outside Moscow tend to bear higher risks. And finally, ownership does matter for risk taking. Surprisingly, foreign-owned banks are found to be more risk-taking than other banks.

The next section provides a brief overview of the Russian banking sector. Section three describes the data and provides group-wise comparisons of financial risk measures by size and ownership categories and by location, as well as by inclusion in the deposit insurance scheme. Section four complements the previous results with a z-score analysis and section five concludes the analysis.

2. Banking industry in Russia

After the crisis-ridden 1990's, especially the deep recession and financial collapse of 1998, the Russian economy has grown annually by more than six percent since 2000. The banking system has experienced rapid growth since 2001, when the sector recovered from the insolvencies and the complete lack of trust created by the 1998 turmoil. Trust in counterparties is still fairly low especially at the interbank markets and the sector is prone to rumors. This was exemplified in the summer 2004 when rumors and tight liquidity created a «mini-crisis» in the banking industry. The effects were, however, not long-lasting. Bank credit to the private sector as a ratio to GDP has more than doubled during the last decade. This is very rapid growth even compared to the fast-growing emerging economies of Central and Southeastern Europe. The resulting financial deepening has been supported by a stable macroeconomic environment, increasing incomes and institutional reforms.

Continuous economic growth, rising real incomes, declining inflation and public sector surpluses have enabled fast increases in the private sector credit share. The majority of credits are financed by private sector deposits, which have increased by

10 per cent annually during the last six years [16]. Also net foreign borrowing has increased, even though the level of total foreign liabilities in Russian banks is still relatively modest at on average below 20% of total liabilities.

Table 1.

Banking system indicators , % of GDP

2004 2005 2006 2007 2008

Total assets 42,1 41,7 44,8 51,9 61,0

net foreign asset position -1,4 -1,9 -2,7 -5,9 -9,0

credit to the private sector 20,2 22,8 25,2 29,9 37,2

o/w enterprises 18,3 19,6 20,3 22,9 28,2

o/w households 1,9 3,2 4,9 7,0 9,0

deposits by the public 23,6 24,4 27,3 32,0 37,0

o/w households 11,5 11,6 12,8 14,2 15,6

Note: Data concerns beginning of each period.

Source: Central Bank of Russia.

Furthermore, a number of important institutional reforms have undoubtedly helped fuel banking sector growth. The most important one was the introduction of the deposit insurance system (DIS). The federal law on compulsory deposit insurance was adopted in December 2003. The law made the formerly implicit guarantee of state-controlled banks explicit and outlined clear rules for banks entering the system. The Deposit Insurance Authority began its operations in 2004, and by the end of March

2005 the first 824 banks were admitted into the system. Most of the rejected banks were small, as the banks already admitted accounted for 98 percent of household deposits. This did raise some concerns on the entry requirements not being interpreted rigorously enough.

By the end of September 2005, when the deadline for joining the system expired, 927 banks out of the 1150 applicants were admitted [14]2). During 2006-2007 Central Bank of Russia (CBR) gradually revoked the licenses to attract household deposits from banks not included in the system. Initially private deposits up to RUR 100000 were covered in full. Later the coverage limit was raised to RUR 190000 in August

2006 and to RUR 400000 in March 20073). During 2003-2005 also several other important laws, e.g., clarifying the rules for mortgage lending and mortgage-backed securities, were enacted. The law from 2005 gave the framework for the operations of private credit bureaux.

2) In order to pacify depositors during the mini-banking crisis of summer 2004, the government enacted a law granting temporary deposit insurance to all banks. Therefore, irrespective of possible inclusion in the deposit insurance system, all Russian banks were guaranteed blanket deposit insurance for deposits up to RUR 100000 from July 2004 until the end of 2006.

3) The limit was further raised to RUR 700000 in October 2008. See: http://www.asv.org.ru/in-surance/.

During the last few years Russian banks have intensively diversified into household lending, especially mortgages, as well as lending to SMEs. Credit maturities have also increased and maturities of over three years are not uncommon. The volumes of mortgage lending are, however, still tiny as less than 10% of homes in Russia are bought using a mortgage (Interfax, 2008). Another remarkable recent trend is the continuing de-dollarization of banking assets and liabilities. Like many transition countries, Russia was heavily dollarised and immediately after the 1998 crisis the use of dollars was very widespread. The share of foreign currency loans has now stabilized at below 25% of corporate loans. Corporate borrowers typically have a significant portion of their earnings in foreign currencies, so currency mismatches should not pose a systemic risk.

In light of all these changes, the structure of the Russian banking sector has remained surprisingly unchanged. The large, state-controlled banks still dominate the market. Even though the number of banks has decreased from 2084 at the end of 2000 to a mere 1243 by the end of 2007, the great majority of the banks are still tiny and can hardly be called banks. At the end of 2007 some 900 banks had the right to attract household deposits and only 300 banks had a general banking license. The foreign ownership share remained fairly limited as evidenced by the Table 2 below. There were 202 banks with a foreign ownership at the end of 2007, 62 of them fully foreign-owned.

Table 2.

Bank ownership in selected countries in 2005

Number of banks Number of foreign-owned banks, % of total Asset share of foreign-owned banks, % of total Domestic credit to private sector (% of GDP)

Estonia 13 77 99,4 57

Slovak Republic 23 70 97,3 34,7

Czech Republic 36 75 84,4 35,7

Lithuania 12 50 91,7 41,3

Hungary 38 71 82,6 49,8

Poland 61 82 74,2 29,2

Latvia 23 43 57,9 59

Slovenia 25 36 22,6 56,4

Russia 1253 4 8,3 26,1

Source: EBRD Transition Report 2006.

Our dataset ends in April 2007, just before the the global credit crunch caused by the subprime market problems in the US started to evolve. Initially the Russian banking industry was only mildly affected, in large part thanks to increasing crude oil prices that provided ample liquidity in the domestic market. Along with falling crude oil prices and drastically deteriorating situation at the international financial markets also the Russian banking sector began to face serious problems by the end of 2008.

3. Measuring risk - financial and regulation ratios

3.1. Data

Our dataset covers most of the banks operating in Russia over the period of April 1999 - April 2007. It consists of banks' quarterly balance sheets and profit and loss accounts. Regulatory ratios calculated by the Central Bank of Russia (CBR) are also partially included in our data and we use them in the analysis to support our main results. The data are provided by the financial information agency Interfax and originated in the Central Bank of Russia. For a more detailed description of the dataset used, see Karas and Schoors (2005) [32]. As the sample period starts in 1999, our results are not directly influenced by the financial crises of August 1998. The data constitutes an unbalanced panel, because there were banks entering and leaving the market due to mergers or failures. A brief overview of the main variables based on summary statistics is provided in Table A.1 in the appendix.

The banks are divided into different subgroups by size, ownership and location as well as inclusion in the deposit insurance system. We use the book value of total bank assets as a measure of size4). Bank size is especially important in Russia, where a handful of the largest banks account for most of the banking sector assets. At the end of 2006, large state-controlled banks accounted for about 40% of the sector assets [15]. Taking into account the overly concentrated nature of the Russian banking sector, we separate for the three largest banks (Sberbank, VTB and Gazprombank). In general, due to more possibilities for diversification and better access to financial markets, large banks are supposed to be less risky. Nevertheless, as Demsetz and Strahan (1997) [24] point out, large banks offset their potential benefits from diversification with lower capital ratios and more risky loan portfolios. Also empirical evidence on the relationship between size and risk has produced slightly mixed results [28, 29].

As for ownership, we distinguish among three ownership groups to determine majority ownership: state-controlled, foreign and domestic private banks. The foreign ownership dummy variable is based on the CBR data on 100% foreign-owned banks published quarterly. State-controlled banks are defined using the list provided in Ver-nikov (2007)5). Due to its special role as a state development bank, we do not include Vneshekonombank (VEB).

Ownership may be important for risk-taking behaviour for various reasons. State-owned banks are often assumed to take higher risks than the private ones. The underlying reasons differ according to one's view on the character of state-owned banks. Sapienza (2004) [41] distinguishes three alternative views. The social view suggests that state banks intervene to correct for the market failure caused by private

4) We first separate the three largest banks as a group of their own. The rest of the banking sector is divided into three groups. Small banks are those with total assets below 33rd per-centile, medium banks have assets between 33rd and 66th percentiles and the large ones have total assets above the 66th percentile in every time period. Alternative measures of size based on the market share of the aggregate domestic credit as well as participation in the interbank market provide us with a very similar distribution of banks into subgroups and therefore we only use total assets as a proxy for bank size.

5) This list largely overlaps with the other lists of state-controlled banks used by Karas et al. (2008) [33]. Moreover, our number also corresponds to the number of government-controlled banks in the Bank Supervision Report (2006).

banks, which «cherry-pick» the best customers and would leave the not very profitable ones without financial services. This view implies that state banks are engaged in more risky and less profitable operations but possibly enjoy soft budget constraints. The political view sees state banks as well as state enterprises more as a mechanism for pursuing politicians' private interests, such as maximizing employment or delivering favours for political protégées. This view implies that state banks may be forced to lend on a non-commercial basis i.e. due to political or other reasons. The agency view sees state banks as basically benevolent maximizers of social welfare but plagued by corruption and misallocation. Recent evidence from industrialized countries [20, 29] suggests that state-owned banks typically exhibit higher risk than other types of banks.

Studies on transition economies have, however, produced mixed results [21, 36]. In transition economies state-owned banks may be less efficient and more risk-prone due to Soviet legacies, unrestructured management or soft budget constraints. These findings, usually based on Central European countries (see e.g. [8]), are challenged by Karas et al. (2008) [33], who show that in Russia state-owned banks are not less efficient than domestic private banks.

Foreign-owned banks may have a different risk profile due to less local expertise and fewer local connections compared to the domestically owned banks. Their operations may also be less risky since they might often be able to cherry pick the most creditworthy borrowers in an emerging market [7]. Additionally, these banks can often rely on strong parent companies to provide them with access to better risk management techniques and possible diversification of country risk. On the other hand, foreign ownership may aggravate risks if parent banks tend to stress rapid credit growth in order to relieve tightening interest margins at home. Moreover, integration into the global financial system has also highlighted new issues related to risk management and financial vulnerability.

Foreign bank entry has been one of the decisive factors shaping banking sector development in Central and Eastern European transition countries. The available empirical evidence supports the common view that foreign-owned banks are more efficient than other types of banks in these countries ([5, 8, 9] and references therein). Furthermore, there is a growing literature exploring the effects of the presence of foreign-owned banks on domestic credit markets in emerging economies6). The role of foreign-owned banks in Russia has been dramatically different from those in the Central European banking sector. The share of foreign capital in the Russian banking sector was tiny up until spring 2007 as no major privatizations had taken place. The Russian banking sector is clearly more distant (both geographically and culturally) and therefore less attractive than the new and prospective EU member countries. Moreover, acquiring a large market share is not as easy as it was in Central Europe. Nevertheless, the foreign-owned banks operating in Russia may be extremely important as a benchmark for domestic ones and it is therefore most interesting to examine if they differ in their risk-taking.

6) Mostly the results on the benefits of the foreign bank presence are mixed. Detragiache et al. (2008) [25] show that banks give fewer loans after being acquired by a foreign investor. Clarke et al. (2005) [17] find that foreign banks make more loans to SMEs than domestic ones. Foreign banks may be reluctant to lend to opaque borrowers, but induce domestic banks to lend to them [22]. Giannetti and Ongena (2008) [26] suggest that foreign banks enhance access to credit, especially where financial development is low.

The division by ownership and size is rather standard. A bank's location within a single country and its inclusion in the deposit insurance scheme are more specific to Russia. Economic developments in different parts of Russia vary a lot. About half of the Russian banks are located in Moscow. The other half, located in the other regions of the Russian Federation, are mainly small banks constituting only 15% of the total banking sector assets. It has been occasionally argued that regional banks are more inclined to lend to local enterprises and to small and medium-sized businesses, thereby promoting growth more than Moscow-based banks. Moscow-based banks, on the other hand, are more active in interbank money markets. If true, this should also be reflected in differences in risk measures. Therefore we split the sample into two depending on the location of the bank's headquarters in Moscow or elsewhere in the Russian Federation. The division into regional and non-regional banks is unavoidably somewhat arbitrary as a large number of banks headquartered both in and outside Moscow have wide networks outside their home region. But the division used is the best available approximation for Moscow and non-Moscow banks. If the banks do not differ in their risk-taking based on the location of their headquarters, the division should not be significant in our analysis. But, as will be seen, the statistically significant result survives all our robustness checks.

Russia adopted a deposit insurance system in 2004 with the majority of banks screened and admitted into the system by end-March 2005. The deposit insurance system was expected to increase the confidence in and stability of the banking sector, as well as to level the playing field between large and small banks. The academic literature on deposit insurance increasingly emphasizes that explicit deposit insurance has the potential to affect bank risk-taking. Since it reduces depositors' incentives to monitor banks, it may encourage risk-taking and imprudent banking practices. The Russian data offers us a unique opportunity to test whether the introduction of a deposit insurance system affects bank risk-taking in the short run. We consider two groups of banks based on the point at which they entered the system. We create a dummy variable indicating if the bank was included into the system in the «first wave», by end - March 2005. Inclusion of the banks in the deposit insurance system is defined using the information from the Russian Deposit Insurance Agency.

3.2. Risks faced by banks and corresponding financial ratios

Banking is by nature a business of balancing risks. There is, however, no single, universal measure that could be used to assess risk-taking behaviour by banks. Thus, we rely on two different approaches. The first one is based on a univariate analysis of financial risk ratios, which are either calculated using the accounting data or belong to the regulatory ratios used by the central bank. We analyze different categories of financial risk separately by employing the relevant financial ratios as well as regulation ratios used by the CBR (for definitions, see Table A.8 with a description of variables in the appendix). Furthermore, we also test the significance of the differences in financial risk ratios among different subgroups of banks7). The second approach, discussed in section four, relies on the regression analysis of bank insolvency risk as measured by the z-score indicator.

7) We use a nonparametric K-sample test on the equality of medians.

Capitalization

Capitalization is calculated as a ratio of equity to total assets and it serves to measure leverage risk. Due to rapid asset growth, the level of capitalization declines during the period analyzed (see Table A.2 in the appendix). Capitalization is, however, still higher than in most other transition countries as reported in Haselmann and Wachtel (2007) [28]. On average, capitalization decreases with size and thus small banks tend to have higher capital ratios than larger banks. This is in line with the «too big to fail» hypothesis as well as with the perceived difficulties smaller banks face in accessing interbank markets in Russia. Larger banks in general have better opportunities for risk diversification and thus also benefit from lower costs of funding [37].

The capitalization of private banks is significantly higher than that of state and foreign banks during the whole period under review. These banks, unlike state-controlled or foreign banks, usually do not have a kind of «backup» in the form of the state or a strong parent company abroad. That is most probably the reason why they hold a higher proportion of equity capital. Foreign banks are slightly better capitalized than state banks, which is consistent with the results for the CIS in [21]. Banks located outside Moscow tend to maintain lower equity, but the gap between regional and Moscow banks has decreased since 2006 and thus the difference between these two groups of banks is no longer significant. Banks included in the DIS maintain a significantly lower equity than the other banks. There are two possible explanations for this. The first one concerns moral hazard issues connected with the participation in the deposit insurance scheme. The other is selection bias. It indicates that the banks entering the system were the better ones, which, based on their results, were obvious candidates for inclusion immediately when the system was introduced.

The CBR regulation ratio N1 used to assess capital adequacy8) confirms these trends as well. Even though the capital adequacy ratio has declined in recent years, its average value of 14,5% for November 2006 [15] still clearly exceeds the minimal requirements set by the central bank9). This indicates that Russian banks on average tend to keep slightly higher capital buffers than banks in the EU-25 countries as Jokipii and Milne (2008) report [30]. It is, however, clear that relatively large capital buffers at the beginning of our sample period are a natural reaction to the uncertainty following the crisis of 1998. The gradual decrease of capital buffers is then to a certain extent the result of the improvements in the macroeconomic environment. Nevertheless, it may also indicate that the operations of Russian banks are becoming more efficient or that the institutional environment is improving [10, 28]. The unfavourable global development resulting from the sub-prime crisis and liquidity problems in the second half of 2007 made banks more cautious again and the majority of banks increased their capital adequacy ratios towards the end of 2007 [16].

8) Unlike the indicator of capitalization, the N1 ratio is for most of the banks available only until 2005.

9) The Financial Stability Report 2006 issued by the central bank reports that according to Bank of Russia Instruction № 110_I, dated January 16, 2004, the minimum capital adequacy ratio for a bank (N1) is 10% if the bank has a capital of at least 5 million euros and 11% if the bank has a capital of less than 5 million euros. Only 11 credit institutions violated the capital adequacy ratio in 2006 and 19 in 2005 (Bank of Russia Financial Stability Report, 2006).

Credit risk

Analyzing credit risks is becoming increasingly important in Russia due to its rapid credit growth. The increase in the loans to total assets ratio (see Table A.3 in the appendix) suggests that the growth of lending has been higher than the growth in total assets, implying a gradual shift towards riskier operations of banks. Domestic banks have significantly higher lending ratios than foreign banks, whereas regional banks tend to lend more than Moscow-based ones10). On average, however, the total loans to total assets ratio in our sample is comparable with the sample of transition economies as reported in Haselmann and Wachtel (2007) [28]. Similar to our expectations, banks that belong to the deposit insurance system lend more. There are again two possible explanations for this. The first one suggests that banks in the DIS may take more risks as they are backed up by the system. The latter indicates that insured banks are on average better and more efficient and therefore they are able to bear higher risks.

One of the most commonly used indicators of credit risk is the ratio of nonper-forming loans (NPL) to total loans. The share of NPLs in Russia has indeed increased during the last years, but the levels are not yet anywhere close to becoming alarming. The median levels based on our calculations (see Table A.4 in the appendix) are still below the quality level of 1,5 per cent recommended by Grier (2001) [27]. It is, however, necessary to bear in mind that this is an ex post measure of the risks assumed by banks. When considering banks by ownership, state-controlled banks exhibit a significantly higher ratio of nonperforming loans than others. One might take this as indirect evidence of state-controlled banks' lending, willingly or unwillingly, to any customer, also to the uncreditworthy one. It is, however, interesting to note that the share of NPLs among the state-controlled banks has stayed basically unchanged in recent years. The recent increase in the NPL share has been caused mainly by private domestic banks. On the other hand, foreign banks have the lowest level of NPLs, which may reflect their relatively short period of operation on the Russian market, better credit risk management, or both.

The ratio of NPLs is increasing with the bank's size, which suggests that larger banks are able to sustain a larger proportion of NPLs. The difference between small and large banks is, however, gradually decreasing. The shrinking of this gap is the result of both an increase in the NPL ratio of small banks and a decrease among the large ones. Despite this development, the variation between banks of different sizes still remains significant. There are significant differences in the proportion of NPLs by location as well. Even though regional banks still tend to have a larger ratio of NPLs, similar to when we account for size, the gap between Moscow and regional banks has decreased recently. There are also differences between banks that are part of the deposit insurance system and the ones that are not. The ones included in the scheme have in general higher nonperforming loan ratios, which can be a natural consequence of higher lending by these banks.

Since banks with nonperforming loans are obliged to make loan loss provisions, a comparable measure of credit risk is the ratio of loan loss reserves to total loans. Its development basically corresponds to changes in the proportion of nonperforming

10) The underlying reasons for the different asset structure of regional and Moscow-based banks may include variations in fixed assets like buildings and branch-office networks. This issue would clearly merit a study of its own.

loans (see Table A.4 in the appendix). The proportion of loan loss reserves in total loans is the lowest for the foreign-owned banks. Even though the proportion of loan loss reserves was the highest for the three largest banks in 1999, nowadays this ratio is basically the same for banks of all sizes. This seems to serve as evidence for the special position of these state-controlled banks. The loan loss indicator further suggests that the deposit insurance scheme implementation contributed to changes in loan loss reserves. Before the deposit insurance scheme was implemented, loan loss reserves were significantly higher for the banks that later entered the scheme. However, with the implementation of the scheme, reserves in the banks not included in the system increased and they are higher compared to the banks that are part of the DIS.

Maximum large credit risk is a regulation ratio that measures the proportion of the total amount of large credit risks11) in a bank's equity capital. It increases over time and tends to be higher for the state-controlled banks and for the regional banks. This could indicate that these banks have close connections with large state-controlled or regional companies. The maximum large credit risk ratio is also higher for larger banks with the exception of the three largest ones. Moreover, it is significantly lower for the banks outside the deposit insurance system, which once again indicates that banks that are part of the system are able to engage in relatively more risky activities.

Even though our analysis of credit risk measures suggests that the operations of state-controlled banks tend to be relatively riskier than the others, the comparison of the credit risk indicators to the corresponding figures in other countries as well as to the critical values indicated in the literature suggest no excessive risk-taking. Our results are thus in line with the CBR [15] in that, on average, the credit risk of Russian banks remains moderate.

Liquidity risk

The Russian banking sector's liquidity as measured by the ratio of liquid to total assets has decreased slightly in recent years, but its level, reported in Table A.6 in the appendix, is still comparable to the other transition countries as well as to the quality level recommended by Grier (2001) [27]. An analysis of the regulatory ratios of quick and current liquidity (see Table A.8 in the appendix for detailed definitions) confirms that they have remained basically unchanged. Foreign banks and Moscow-based banks exhibit the highest level of liquidity during the whole period under review. One possible explanation for this phenomenon is that Moscow-based banks are on average less engaged in traditional banking operations (collecting retail deposits and channeling them into corporate loans) than regional banks. Furthermore, Moscow-based banks tend to be more active in interbank money markets and therefore have a larger proportion of their assets in a highly liquid form. This difference in bank operations is reflected in the increasing gap in the liquidity indicator between Moscow and regional banks. The finding is a corollary to the finding that, on average, the share of loans in total assets is lower for Moscow-based banks than for the other banks. Unlike the divisions by region and ownership, the distribution of banks by size does not indicate any significant differences in liquidity for banks of various sizes.

11) Large credit is the total sum of the bank's risk-weighted claims to one borrower (or a group of related borrowers) on credits.

Moreover, in line with the other credit risk indicators, the banks included in the deposit insurance scheme hold lower levels of liquidity and the gap between them and the other Russian banks has been increasing since 2005.

In general, high liquidity ratios can be interpreted as having a positive influence on stability at certain levels of liquidity. In the case of emerging economies, liquidity ratios may also be higher if the government does not actively intervene to meet funding gaps, financial institutions are risk-averse or if there are not enough opportunities for hedging [38]. In that case excessive liquidity could indicate structural problems. A bank may be highly liquid simply because: 1) it cannot rely on well-functioning interbank markets or other secondary markets such as those for securities; 2) it prefers to distance itself from «traditional» banking operations such as lending in favour of trading in, e.g., government securities; or 3) both.

Despite sufficient liquidity in general, there has been a lack of efficient mechanisms for interbank intermediation of liquidity. The Russian interbank market is relatively small even in comparison to other emerging markets [38]. This is especially the result of high segmentation and low trust on the interbank market [5], even among the big state-controlled banks. Russian banks are highly liquid but the banking system as a whole is not. Due to the lack of trust, the banking system is vulnerable to occasional liquidity shocks as experienced in summer 2004 and autumn 2007. This clearly complicates banks' liquidity management as well as the conduct of monetary policy in Russia.

Market risk

The net interest margin12) as a percentage of loans is often used as a proxy for the efficiency of financial intermediation, thus uncovering the health of the banking sector. Higher margins indicate lower efficiency and lower competition within the sector and thereby possibly also higher risk. Our analysis indicates that foreign banks have significantly lower net interest margins than private banks, even though recent developments suggest that the net interest margins of foreign banks have increased to the level of state-controlled ones (see Table A.7 in the appendix). In this respect, lower margins most probably reflect the greater efficiency of foreign banks which is connected to the support and know-how from their parent companies. Our indicators are thus in line with Karas et al. (2008) [33], who find that Russian state banks are more efficient than domestic private banks. The net interest margin decreases with the bank's size and therefore it is the lowest for the group of the three largest banks. Regional banks used to have significantly higher net interest margins. However, the situation has changed recently and consequently Moscow-based banks have slightly higher margins, which may suggest increasing efficiency and/or competition. After the implementation of the DIS, the net interest margins of the banks included in it decreased and became significantly lower than the margins of the other banks. This development may indicate a positive impact of the DIS introduction on the banking sector's competition and efficiency; however, more investigation is necessary to confirm this result.

12) The net interest margin is calculated as the difference between the interest income from loans to customers and the interest expense paid on customer deposits.

To sum up, the analysis of ratios measuring financial risk confirms significant differences among groups of Russian banks by size, location, ownership and participation in the DIS. Nevertheless, it is only based on the comparisons of unconditional medians. The following regression analysis provides more insight by uncovering also conditional correlations.

4. Measuring risk - bank insolvency risk (z-score)

In addition to the four classes of bank risk ratios, we use a measure for insolvency risk developed by Boyd and Graham (1988)13) [12] that has been increasingly used in the banking literature. Different modifications of z-scores have been applied in the empirical cross-country [13, 20, 21, 29, 36] as well as single-country studies [34, 35].

The insolvency risk measure («z-score» hereafter) is a statistic indicating the probability of bankruptcy (bank failure). The z-score for each bank i at quarter j is calculated as:

(1) Zj = (ROAit + EQTAit) / a(ROA)it,

where ROAit and (ROA)it are sample estimates of the four quarters moving average and the four quarters standard deviation of bank i's returns on assets at quarters t to t - 3 and EQTAit is the four quarters moving average of the equity capital to assets ratio. A bank's return on assets is calculated as its one-quarter profit before taxes on the quarter's average total assets. A bank's equity to assets ratio is calculated as the equity capital on total assets at the end of a given quarter. As we used the four quarters (backward-looking) moving averages in constructing our insolvency measure as well as explanatory variables, the time span of our analysis effectively covers the years 2000-2006.

Statistically speaking, the z-score represents the number of standard deviations returns would have to fall in order to deplete a bank's equity, under the assumption of normality of the bank's returns. Boyd et al. (2006) [13], however, argue that «it (the z-score) does not require that profits be normally distributed to be a valid probability measure; indeed, all it requires is the existence of the first four moments of the return distribution». A higher z-score corresponds to a greater distance to equity depletion and therefore to lower risk and higher bank stability.

The z-score measure inherently depends on the assumption that the ROA, relying on profit and loss data, gives a useful approximation of a bank's financial health. Since our data is based on Russian accounting system standards, which stress formal reporting rather than economic meaning, it may be questioned whether our data fulfils that requirement [5]. Nevertheless, as we only compare Russian banks with each other, possible flaws in the accounting standards should not be over-emphasized. Moreover, we use the z-score indicator to uncover statistically significant conditional correlations, not causality.

13) This measure originated as a predictor of corporate bankruptcy [3].

4.1. Methodology

Our focus is on the effects of a bank's size, ownership, location and inclusion in the deposit insurance scheme on its insolvency risk (z-score). The bank's size is measured by a continuous variable (logarithm of total assets) whereas ownership, location and inclusion in the deposit insurance scheme are proxied by using corresponding dummy variables. The dummy variable for inclusion in the deposit insurance scheme is fully time-invariant whereas the dummy variables for ownership and location exhibit very little if any within variation. Therefore a standard fixed-effects model is likely to lead to inefficient estimates with very large standard errors14).

We remedy the problem by applying the fixed effects vector decomposition (FEVD) approach by Plumper and Troger (2007) [40]. The approach suggests estimating the model in three steps. First, our dependent variable is regressed only on the cross-section fixed effect and the time-varying factors. Second, the estimated fixed effect (unit effect) is decomposed into the part explained by the time-invariant variables and the unexplainable part (error term). Finally, the model including the unexplained part of the fixed effect is re-estimated by pooled OLS. By design, the remaining error term is no longer correlated with time-invariant variables. Plumper and Troger (2007) [40] show that FEVD estimates are superior (in root mean squared errors) to the traditional fixed effects estimation. In running the FEVD estimations, we use STATA's FEVD module.

We estimate the following model:

where

• z is the z-score for bank i at time t calculated as indicated in the equation

• size stands for the logarithm of total assets of bank i at time t;

• bankSpec is a set of bank i's specific ratios at time t including liquidity, credit growth and the share of loans to individuals in total loans;

• IA is a set of interaction dummy variables between a bank's size and bank-specific factors;

• owner is a set of dummy variables distinguishing among foreign, state-controlled and private banks;

• region is a dummy variable indicating Moscow headquarters of bank i at

time t;

• seas stands for seasonal (i.e. quarterly) dummy variables;

• depInsurance is a dummy variable indicating inclusion in the first wave of the deposit insurance system.

All the variables used in the regressions are four-quarter moving averages. Z-score and total asset variables are in natural logarithms. Bank-specific factors include credit

(2)

ln(z) it = a,. + blSi ze it + b2 (BankSpec) it + РЪ(М) it + PA(seas ) +b5(Owner ) t + b6 (Region),. + b7( DepInsurance) t + ett,

(1);

14) For recent discussions on fixed-effect models with time invariant variables, see, e.g., [6, 43]. For a classic textbook approach using Hausman-Taylor procedures, see [44, p. 235-238].

growth, the liquidity ratio and the share of loans to individuals in total loans. A bank's size, ownership, location and inclusion in the first wave of the deposit insurance system are defined as in the analysis of bank risk ratios in the previous section. To remove potential outliers, 0,5% of both tails of each variable in every quarter was removed. Table A.8 in the appendix gives details of the variables used in the regressions.

A priori, the sign of the coefficient on a bank's size is indeterminate because large banks may be either stabilizing or risky for the banking system, as our previous analysis of risk ratios suggests. Bank-specific risks are captured by the measures of credit risk and liquidity risk. Credit risk is proxied by bank-by-bank credit growth as well as the ratio of loans to individuals to total loans. Liquidity risk is controlled for by introducing the liquidity ratio (liquid assets/total assets) to the model. A priori we do not have an expectation of the sign for these variables.

4.2. Estimation results

In order to analyze the relationship between a bank's size, ownership and location and the risk measured by the z-score, we estimate the model of equation (2) employing the fixed effects vector decomposition described above. The main results are shown in Table 3 below.

Several interesting findings emerge. First, the results consistently indicate that larger banks have significantly lower z-scores and thus higher insolvency risk15). Second, somewhat unexpectedly, foreign-owned banks consistently bear higher insolvency risk than domestic private banks. This result is fully in line with some earlier studies on emerging economies using z-scores as the risk measure [36]. The result naturally reflects the limitations of the risk measure used, as it partly originates from the lower capitalization ratios of the foreign banks. Furthermore, it is necessary to bear in mind that due to data limitations, our foreign ownership dummy variable only accounts for banks that are fully foreign-owned. The overall effect of state ownership on a bank's insolvency risk is positive, i.e. state-controlled banks tend to be more stable. To investigate this result more closely, we add the interaction term of size and state control to our model. This interaction is positive and highly significant. At the same time, the estimated coefficient for the state-controlled dummy variable becomes negative. This indicates that only large state-controlled banks are driving our results and they are more stable than other state-controlled banks.

Third, the Moscow-based banks are always more stable than the regional banks. Based on the data available to us we can not determine the ultimate reason for this significant difference. The higher levels of capitalization in Moscow banks certainly play a role. The underlying reasons may include differences in bank operations, differences in banks' clientele and differences in bank supervision and regulation. Answering the highly interesting question on why the regional differences emerge would clearly merit a study of its own. Finally, similar to our expectations, banks that became part of the deposit insurance system in the first wave are more stable.

15) The z-score regressions are based on the full set of commercial banks, including the three large ones. As a robustness check we did run the estimations without the big three banks, but the results stay unchanged.

Finally, we conclude that the bank-specific characteristics do have a significant role in explaining insolvency risk. In line with earlier literature (e.g. [36]), we find that higher liquidity implies higher insolvency risk. We include an interaction variable of bank size and liquidity, which confirms that large liquid banks are more stable. The growth of a bank's loan stock is used to control for the credit risk. In line with Maechler et al. (2007) [36], its impact is positive in our estimations and this indicates higher stability. This result holds true for Moscow-based banks, while for regional banks the estimated coefficient is negative. We also control for the interaction of bank size and credit growth to see if credit growth affects small banks differently. We find that large banks with high credit growth are in fact more stable than the rest of the sector.

Table 3.

Estimation results

Estimated coefficient

Size (total assets) -0,262***

Loans to households (prop. of loans) -0,355***

Liquidity (liquid to total assets) -0,616***

Credit growth 0,015***

OWNERSHIP, LOCATION AND DEPOSIT INSURANCE

Deposit insurance 0,104***

Foreign bank -0,572***

State-controlled bank -0,534***

Moscow-based bank 0,501***

INTERACTIONS

Size and liquidity 0,054***

Size and credit growth 0,003***

Size and state-controlled 0,100***

Number of observations 27353

R2 0,426

Note: The table contains results for the FEVD regression. We report estimated coefficients as well as their significance (***significant at 1%, ** significant at 5% and * significant at 10%). Seasonal and yearly dummy variables as well as a constant term are included but not reported.

We test the robustness of our empirical results using several techniques.

• First, the results are robust to the exclusion of the three largest state-controlled banks (Sberbank, Gazprombank, VTB) from the sample.

• We split the sample into Moscow-based and regional banks. The FEVD regression model is run for the two subgroups separately. Except for the significance of credit

growth, other results for both subgroups are in line with the results of the main model reported above. Nevertheless, the model seems to fit a little bit better the Moscow-based banks, which account for about 85% of the banking sector assets.

• Finally, the results for the subsample of the 300 largest banks also correspond to our main results reported in Table 3. They only differ in the sign of the deposit insurance scheme dummy variable. In this case it is negative, which means that the banks that entered the system in the first wave are more risky. This is in line with the results of univariate analysis of financial ratios performed in the first part of the paper.

4.3. Z-score components

The z-score measure consists of three main components: the return on assets, capitalization and the volatility of the ROA. In order to investigate the contribution of each of them to explaining differences in the banks' stability, we run our basic model using all of these components as a dependent variable. This approach is in line with previous literature [21, 36]. We report the results of the z-score component regressions in the following, Table 4.

The first component of the z-score measure is capitalization16). In this case, the fit measured by R2 is the highest of all the z-score components. The estimated coefficients are larger than for the other z-score components and almost all of them are significant. The estimated coefficients are mostly in line with the results of the main model, which indicates that the majority of the main results are driven by the contribution of the capitalization ratio. Larger banks have lower capitalization and this result undoubtedly drives our final result that banks with a higher amount of total assets are in general less stable. More liquid banks have lower capitalization, which indicates that banks substitute between liquidity and solvency risk. Nevertheless, liquid large banks tend to have higher capitalization. Both state-controlled and foreign ones are in general better capitalized than private ones. The effect of deposit insurance participation on capitalization is significantly negative. Banks in the deposit insurance system do seem to substitute deposit insurance for capital, or put in other words, take more risks for the same level of capital. This result is in line with earlier literature [23].

The second column contains results for the regression with the ROA as the dependent variable. Similar to the capitalization component of the z-score, almost all the estimated coefficients are significant for the ROA. However, the majority of their signs differ from the results in the main z-score regression. Higher credit growth as well as a higher share of loans to individuals in a bank's loans portfolio are positively related to profitability. Higher liquidity positively influences profitability as measured by the ROA. Given the fact that the average real interest rate on corporate loans was close to zero for much of the period, this is not entirely surprising. Many banks make more than half of their revenues from foreign currency operations. When accounting for a bank's ownership, foreign banks and state-controlled banks have a significantly higher ROA than domestic private ones. Large state-controlled banks are, however, less profitable. Banks included in the DIS in the «first wave» have significantly higher pro-

16) Capitalization is, similar to the calculation of the z-score, calculated as the four-quarter moving average. The other z-score components, the ROA and volatility of the ROA, are calculated in the same way.

fitability than the others, which is in line with our previous result indicating that better banks entered the system first. Moscow-based banks are in general less profitable.

Table 4.

Z-score component regressions

Capitalization ROA Volatility of ROA

Estimated coefficient Estimated coefficient Estimated coefficient

Size (total assets) -0,085*** 0,0002*** -0,003***

Loans to households -0,076*** 0,005*** 0,001**

Liquidity -0,181*** 0,003*** 0,0004

Credit growth 0,003*** 0,0002*** -0,0002***

OWNERSHIP, LOCATION AND DEPOSIT INSURANCE

Deposit insurance -0,010*** 0,001*** -0,002***

Foreign bank 0,092*** 0,002*** 0,009***

State-controlled bank 0,082*** 0,004*** 0,006***

Moscow-based bank 0,134*** -0,003*** 0,001***

INTERACTIONS

Size and liquidity 0,002** -2,8E-05 -0,001***

Size and credit growth -3,0E-04*** -2,3E-05*** -1,2E-05

Size and state-controlled 0,001 -0,001*** -0,0005**

Number of observations 27353 27353 27353

R2 0,785 0,356 0,361

Note: The table contains estimation results of the model described above for different z-score components. We report the estimated coefficients as well as their significance (* significant at 10%, ** significant at 5% and *** significant at 1%). Seasonal and yearly dummy variables as well as a constant term are included but not reported.

The last component of our risk measure is the volatility of the ROA as measured by the standard deviation. Most of the estimated coefficients in this regression are significant but have a different sign than the results presented in our main model. They are also lower in absolute values and therefore, unlike the measure of capitalization, they contribute less to the main results. Thus, the analysis of the z-score components indicates that the differences in the risk profiles of banks are mostly driven by the differences in capitalization.

5. Conclusion

Favourable macroeconomic conditions and important regulatory reforms have backed the rapid growth of Russia's banking sector during this decade. As the economy is increasingly monetized, the role of banks and other financial intermediaries in supporting the continuous growth of investments and private consumption is gaining more importance. Therefore the stability of the banking sector is even more crucial. Compared to most European countries the Russian banking sector is still rightfully characterized as small, regionally fragmented and dominated by a few large state-controlled entities.

On average, the Russian banking sector is believed to be in good financial shape as evidenced also by the Banking Supervision Reports of the CBR. For this paper we use a bank-level dataset on all Russian banks to examine how various measures of risk vary with a bank's size, ownership, location and inclusion in the deposit insurance system. The main objective is the detailed examination of how these various groups of banks differ in their attitudes to risk. We employ two approaches; group-wise comparisons of financial ratios and regression analysis using a z-score measure of bank insolvency risk. The analysis of financial ratios reveals that even though the ratios point to increasing risk over time, they are still on average well on the safe side within all groups of banks. The average levels are all above the regulatory minima set by the Russian Central Bank. Moreover, they are comparable to other transition economies. The rapid growth of the banking sector has not led to excessive risk-taking on average.

The regression analysis of the bank insolvency measure (z-score) proved to be a useful means of deepening the results of group-wise comparisons. Controlling for bank characteristics, large banks in Russia have higher insolvency risk than small ones. Second, in line with the previous literature on emerging economies, foreign-owned banks exhibit higher insolvency risk than domestic banks. Even though the foreign bank presence may in general greatly increase banking sector efficiency and widen the range of banking services available, foreign-owned banks in Russia seem to bear higher risks. The same holds true for the state-controlled banks; however, the large state-controlled banks are more stable than the others. Third, we find that the regional banks are significantly more prone to risk-taking than their counterparts in Moscow. Regional banks only account for a small fraction of the total banking sector assets, thus this finding should not be alarming for the banking sector as a whole.

All in all, we find that risk-taking by Russian banks is approaching levels comparable to other emerging economies. Further, factors similar to those in emerging European economies seem to explain levels of insolvency risk in Russia. We also briefly examined if inclusion in the Russian deposit insurance scheme has influenced a bank's insolvency risk. The results are mixed and further research on this topic is clearly needed.

* * *

References

1. Алескеров Ф.Т., Мартынова Ю.И., Солодков В.М. Анализ и оценка эффективности функционирования банков и банковских систем / / Сборник докладов VIII Международной научной конференции «Модернизация экономики и общественное развитие». М.: ГУ ВШЭ, 2008.

2. Головань С.В., Карминский А.М., Пересецкий А.А Непараметрические оценки эффективности российских банков. Презентация доклада на IX Международной научной конференции «Модернизация экономики и глобализация». М.: ГУ ВШЭ, 2008.

3. Altman E. Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy // The Journal of Finance 1968. № 23(4). P. 589-609.

4. Bank for International Settlements Foreign Direct Investment in the Financial Sector - Experiences in Asia, Central and Eastern Europe, and Latin America. BIS, Committee on the Global Financial System. 2005. June.

5. Barisitz S. Banking in Central and Eastern Europe 1980-2006. L.: Routledge, 2008.

6. Beck N. From Statistical Nuisances to Serious Modelling: Changing How We Think about the Analysis of Time-series-cross-section Data // Political Analysis. 2007. № 15. P. 97-100.

7. Bhaumik S.K.., Piesse J. Does Lending Behaviour of Banks in Emerging Economies Vary by Ownership? Evidence from the Indian Banking Sector // Economic Systems. 2008. № 32(2). P. 119-216.

8. Bonin J., Hasan I., Wachtel P. Bank Performance, Efficiency and Ownership in Transition Countries // Journal of Banking and Finance. 2005. № 29. P. 31-53.

9. Bonin J., Hasan I., Wachtel P. Privatization Matters: Bank Efficiency in Transition Countries // Journal of Banking and Finance 2005. № 29. P. 1927-1952.

10. Bonin J., Hasan I., Wachtel P. Financial Intermediation in Transition Countries: BOFIT Discussion Paper № 12. 2008.

11. Bonin J., Wachtel P. Financial Sector Development in Transition Economies // Financial Markets, Institutions and Instruments. 2003. № 12. P. 1-66.

12. Boyd J., Graham G. The Profitability and Risk Effects of Allowing Bank Holding Companies to Merge with Other Financial Firms: a Simulation Study. Federal Reserve Bank of Minneapolis. Quaterly Review. 1988. Spring.

13. Boyd J.H., De Nicolm G., Jalal A.M. Bank Risk Taking and Competition Revisited: New Theory and New Evidence: IMF Working Paper 06/297. 2006.

14. Camara M.K., Montes-Negret F. Deposit Insurance and Banking Reform in Russia: World Bank Policy Research Working Paper 4056. 2006.

15. Central Bank of Russia. Banking Supervision Report. 2006.

16. Central Bank of Russia. Bulletin of Banking Statistics. 2007, 2008. Various issues.

17. Clarke G., Cull R., Martinez M., Sanchez S. Bank Lending to Small Business in Latin America: Does Bank Origin Matter? // Journal of Money, Credit and Banking. 2005. № 37. P. 83-118.

18. Claeys S., Schoors K. Bank Supervision Russian Style: Evidence of Conflicts between Micro- and Macro-prudential Concerns // Journal of Comparative Economics. 2007. № 35. P. 630-657.

19. Caner S., Kontorovich V Efficiency of the Banking Sector in the Russian Federation with International Comparision // HSE Economic Journal. 2004. Vol. 8. № 3. P. 357-375.

20. De Nicolo G. Size, Charter Value and Risk in Banking: An International Perspective Board of Governors of the Federal Reserve System. International Finance Discussion Paper. № 689. 2001.

21. De Nicolo G., Loukoianova E. Bank Ownership, Market Structure and Risk: IMF Working Paper 07/215. 2007.

22. Dell'Ariccia G., Marquez R. Information and Bank Credit Allocation // Journal of Financial Economics. 2004. № 72. P. 185-214.

23. Demirguc-Kunt A., Kane E.J. Deposit Insurance Around the Globe: Where does it Work?: NBER Working Paper 8493. 2001.

24. Demsetz R., Strahan P. Diversification, Size, and Risk at Bank Holding Companies // Journal of Money, Credit and Banking. 1997. № 29. P. 300-313.

25. Detragiache E., Tressel T., Gupta P. Foreign Banks in Poor Countries: Theory and Evidence // Journal of Finance. 2008. № 55. P. 1133-1161.

26. Giannetti M., Ongena S. Lending by Example: Direct and Indirect Effects of Foreign Banks in Emerging Markets. 2008. Mimeo, available at ssrn.com/abstract=1085310.

27. Grier W. Credit Analysis of Financial Institutions. L.: Euromoney, 2001.

28. Haselmann R, Wachtel P. Risk Taking by Banks in the Transition Countries // Comparative Economic Studies. 2007. № 49. P. 411-429.

29. Iannotta G., Nocera G., Sironi A Ownership Structure, Risk and Performance in the European Banking Industry // Journal of Banking & Finance. 2007. № 31. P. 2127-2149.

30. Jokipii T., Milne A. The Cyclical Behaviour of European Bank Capital Buffers // Journal of Banking & Finance 2008. № 32(8). P. 1440-1451.

31. Karas A., Pyle W., Schoors K Sophisticated Discipline in a Nascent Deposit Market: Evidence from Post-Communist Russia // Oxford Economic Papers. 2009. Forthcoming.

32. Karas A., Schoors K. Heracles or Sisyphus? Finding, Cleaning and Reconstructing a Database of Russian Banks: Ghent University Working Paper 05/327. 2005

33. Karas A., Schoors K., Weill L. Are Private Banks More Efficient than Public Banks? Evidence from Russia: BOFIT Discussion Paper № 3. 2008.

34. Konishi M., Yasuda Y. Factors Affecting Bank Risk Taking: Evidence from Japan // Journal of Banking and Finance 2004. № 28. P. 215-232.

35. Lin S.L., Penm J.H.W., Gong S., Chang Ch Risk-based Capital Adequacy in Assessing on Insolvency-risk and Financial Performances in Taiwan's Banking Industry // Research in Business and Finance. 2005. № 19. P. 111-153.

36. Maechler A. M., Mitra S., Worrell D. Decomposing Financial Risks and Vulnerabilities in Eastern Europe: IMF Working Paper 07/248. 2007.

37. McAllister P., McManus D. Resolving the Scale Efficiency Puzzle in Banking // Journal of Banking and Finance. 1993. № 17. P. 389-405.

38. Moreno R. The Changing Nature of Risks Facing Banks // The Banking System in Emerging Economies: How Much Progress Has Been Made?: BIS Papers № 28. 2006.

39. Peresetsky A.A., Karminsky A.M., Golovan S.V. Russian Banks' Private Deposit Interest Rates and Market Discipline: BOFIT Discussion Paper 2/2007. 2007.

40. Plumper T., Troger V. Efficient Estimation of Time-Invariant and Rarely Changing Variables in Finite Sample Panel Analyses with Unit Fixed Effects // Political Analysis. 2007. № 15(2). P. 124-139.

41. Sapienza P. The Effects of Government Ownership on Bank Lending // Journal of Financial Economics. 2004. № 72. P. 357-384.

42. Vernikov A Russia's Banking Sector Transition: Where to?: BOFIT Discussion Paper 5/2007. 2007.

43. Wilson S., Butler D. A Lot More to Do: The Sensitivity of Time-Series Cross-Section Analysis to Simple Alternative Specifications // Political Analysis. 2007. № 15(2). P. 101-123.

44. Wooldridge J. Econometric Analysis of Cross Section and Panel Data. Cambridge: MIT Press, 2002.

Summary statistics of the main variables

Appendix

Table À.1.

Variable Obs Mean Median Std. dev.

Z-score (ln) 34700 4,25 4,20 1,24

Total assets 41382 4105 307 52706

Liquidity ratio 41380 0,33 0,28 0,22

Loan loss provisions 40130 0,07 0,03 0,12

Credit growth 33969 4,64 0,39 209,05

GDP growth 40971 0,02 0,06 0,10

Note: Summary statistics for the observations that are actually used in the z-score regression are not significantly different from these figures.

Table A.2.

Capitalization ratio of banks by ownership, region, size and inclusion in DIS

CAPITALIZATION 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1469 0,362 1322 0,333 1312 0,318 1237 0,322 1327 0,303 1323 0,278 1238 0,243 856 0,187 1015 0,190

OWNERSHIP GROUPS

Private obs. 1420 1271 1258 1182 1265 1258 1170 795 946

med 0,366 0,337 0,323 0,329 0,306 0,281 0,246 0,191 0,190

State-controlled obs. med 30 0,287 30 0,287 32 0,250 30 0,250 33 0,232 32 0,222 32 0,177 29 0,138 32 0,150

Foreign obs. med 19 0,111 21 0,175 22 0,236 25 0,258 29 0,239 33 0,236 36 0,206 32 0,177 37 0,160

medians significantly different yes yes yes yes yes yes yes yes yes

REGION

Moscow-based banks obs. med 567 0,378 570 0,359 586 0,350 614 0,354 643 0,328 661 0,308 620 0,275 357 0,195 469 0,190

Regional banks obs. 588 591 595 598 684 662 618 499 546

med 0,359 0,315 0,297 0,298 0,284 0,251 0,213 0,182 0,178

medians significantly different no yes yes yes yes yes yes no no

SIZE CATEGORIES

Small obs. 489 440 436 411 439 439 412 285 338

med 0,539 0,454 0,434 0,439 0,407 0,381 0,330 0,269 0,280

Medium-sized obs. med 490 0,387 441 0,349 438 0,306 413 0,307 444 0,301 442 0,281 413 0,237 285 0,180 338 0,190

Continued

CAPITALIZATION 1999 2000 2001 2002 2003 2004 2005 2006 2007

Large obs. 487 438 435 410 441 439 410 283 336

med 0,235 0,227 0,243 0,259 0,240 0,217 0,182 0,142 0,130

The Big 3 obs. med 3 0,112 3 0,244 3 0,248 3 0,254 3 0,183 3 0,180 3 0,128 3 0,128 3 0,160

medians significantly

different yes yes yes yes yes yes yes yes yes

DEPOSIT INSURANCE

SCHEME (DIS)

Included in DIS obs. med 801 0,284 801 0,255 802 0,213 649 0,172 632 0,162

Not included in DIS obs. med 419 0,367 522 0,338 436 0,312 207 0,258 172 0,251

medians significantly

different yes yes yes yes yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Loans to assets ratio by bank ownership, location, size and in the deposit insurance scheme

Table A.3. participation

LOANS TO ASSETS RATIO 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1469 0,481 1326 0,428 1313 0,485 1238 0,521 1331 0,535 1326 0,555 1238 0,582 856 0,614 1015 0,627

OWNERSHIP GROUPS

Private obs. 1420 1275 1259 1183 1269 1261 1170 795 946

med 0,481 0,431 0,491 0,524 0,538 0,556 0,584 0,616 0,628

State-controlled obs. med 30 0,431 30 0,418 32 0,474 30 0,520 33 0,531 32 0,591 32 0,594 29 0,633 32 0,669

Foreign obs. med 19 0,428 21 0,276 22 0,257 25 0,294 29 0,414 33 0,309 36 0,368 32 0,500 37 0,495

medians significantly

different no yes yes yes yes yes no no yes

REGION

Moscow-based banks

Regional banks

medians significantly different

obs. 567 571 586 615 646 663

med 0,425 0,401 0,451 0,493 0,496 0,506

obs. 588 593 595 598 685 663

med 0,462 0,437 0,505 0,541 0,564 0,596

620 357 469

0,515 0,550 0,561

618 499 546

0,635 0,651 0,659

yes yes yes yes yes yes yes yes yes

Continued

LOANS TO ASSETS RATIO 1999 2000 2001 2002 2003 2004 2005 2006 2007

SIZE CATEGORIES

Small obs. med 489 0,503 442 0,436 437 0,499 412 0,496 443 0,487 442 0,516 412 0,554 285 0,598 338 0,552

Medium-sized obs. med 490 0,486 442 0,459 438 0,479 413 0,522 444 0,555 442 0,578 413 0,585 285 0,62 338 0,631

Large obs. 487 439 435 410 441 439 410 283 336

med 0,443 0,395 0,478 0,538 0,545 0,568 0,596 0,622 0,671

The Big 3 obs. med 3 0,332 3 0,363 3 0,472 3 0,530 3 0,437 3 0,577 3 0,590 3 0,495 3 0,486

medians significantly

different yes yes no no yes yes no no yes

DEPOSIT INSU-

RANCE SCHEME

Included in DIS obs. med 801 0,556 801 0,583 802 0,610 649 0,631 632 0,654

Not included in DIS obs. med 419 0,490 525 0,497 436 0,503 207 0,516 172 0,595

medians significantly

different yes yes yes yes yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Table A.4.

Nonperforming loans to total loans by bank ownership, location, size and the deposit insurance scheme

NONPERFORMING LOANS 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1423 0,019 1275 0,008 1265 0,004 1181 0,003 1280 0,003 1277 0,003 1226 0,005 853 0,009 1009 0,007

OWNERSHIP

GROUPS

Private obs. 1374 1226 1214 1128 1220 1214 1159 792 940

med 0,019 0,008 0,004 0,003 0,003 0,003 0,005 0,009 0,007

State-controlled obs. med 30 0,022 30 0,014 31 0,005 30 0,014 33 0,009 32 0,008 32 0,008 29 0,010 32 0,008

Foreign obs. med 19 0,000 19 0,000 20 0,000 23 0,002 27 0,000 31 0,000 35 0,000 32 0,003 37 0,001

medians significantly

different no yes no yes yes yes yes no yes

REGION

Moscow-based banks obs. med 537 0,001 541 0,000 559 0,000 575 0,000 612 0,001 630 0,001 608 0,002 356 0,009 464 0,006

Regional banks obs. 575 574 578 582 668 647 618 497 545

med 0,040 0,018 0,009 0,006 0,006 0,006 0,008 0,009 0,008

medians significantly

different yes yes yes yes yes yes yes no yes

SIZE CATEGORIES

Small obs. 454 408 403 367 406 406 403 282 333

med 0,036 0,012 0,008 0,000 0,002 0,001 0,003 0,008 0,005

Medium-sized obs. med 482 0,011 432 0,008 428 0,003 404 0,003 436 0,004 433 0,003 410 0,004 285 0,007 337 0,005

Large obs. 484 432 431 407 435 435 410 283 336

med 0,020 0,007 0,003 0,004 0,005 0,005 0,007 0,010 0,009

The Big 3 obs. med 3 0,149 3 0,046 3 0,023 3 0,027 3 0,019 3 0,017 3 0,015 3 0,012 3 0,012

medians significantly

different yes no no yes yes yes yes yes yes

DEPOSIT INSURANCE

Included in DI obs. 797 798 802 647 630

med 0,005 0,005 0,007 0,008 0,009

Not included in DI obs. 403 419 424 205 172

med 0,001 0,001 0,002 0,010 0,005

medians significantly

different yes yes yes no yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Table A.5.

Loan loss provisions by bank ownership, location, size and participation in the deposit insurance scheme

LOAN LOSS PROVISIONS 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1423 0,054 1275 0,043 1264 0,030 1181 0,025 1280 0,024 1277 0,025 1226 0,033 853 0,036 1009 0,038

OWNERSHIP GROUPS

Private obs. 1374 1226 1213 1128 1220 1214 1159 792 940

med 0,055 0,043 0,030 0,025 0,025 0,025 0,035 0,038 0,039

State-controlled obs. med 30 0,061 30 0,042 31 0,025 30 0,031 33 0,027 32 0,022 32 0,025 29 0,029 32 0,032

Foreign obs. med 19 0,018 19 0,037 20 0,022 23 0,013 27 0,015 31 0,011 35 0,005 32 0,011 37 0,012

medians significantly

different yes no no yes no no yes yes yes

REGION

Moscow-based banks obs. med 537 0,025 541 0,022 559 0,016 575 0,018 612 0,024 630 0,022 608 0,039 356 0,053 464 0,051

Regional banks obs. 575 574 578 582 668 647 618 497 545

med 0,081 0,063 0,038 0,030 0,025 0,026 0,030 0,030 0,032

medians significantly

different yes yes yes yes no no yes yes yes

SIZE CATEGORIES

Small obs. 454 408 403 367 406 406 403 282 333

med 0,068 0,056 0,032 0,018 0,017 0,019 0,028 0,030 0,039

Medium-sized obs. med 482 0,038 432 0,037 428 0,027 404 0,025 436 0,023 433 0,021 410 0,030 285 0,036 337 0,037

Large obs. 484 432 430 407 435 435 410 283 336

med 0,057 0,043 0,030 0,030 0,031 0,032 0,042 0,042 0,039

The Big 3 obs. med 3 0,199 3 0,090 3 0,067 3 0,060 3 0,054 3 0,061 3 0,037 3 0,037 3 0,036

medians significantly

different yes yes no yes yes yes yes yes no

DEPOSIT INSU-

RANCE SCHEME

Included in DIS obs. med 797 0,026 797 0,027 802 0,031 647 0,032 630 0,036

Not included in DIS obs. med 403 0,021 480 0,021 424 0,042 206 0,066 172 0,059

medians significantly

different yes yes yes yes yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Table À.6.

Liquidity ratio by bank ownership, location, size and participation in the deposit insurance scheme

LIQUIDITY RATIO 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1469 0,236 1326 0,301 1311 0,291 1238 0,283 1331 0,284 1326 0,281 1238 0,256 856 0,222 1015 0,220

OWNERSHIP GROUPS

Private obs. 1420 1275 1257 1183 1269 1261 1170 795 946

med 0,231 0,299 0,287 0,279 0,276 0,278 0,255 0,221 0,220

State-controlled obs. med 30 0,334 30 0,328 32 0,315 30 0,325 33 0,296 32 0,269 32 0,224 29 0,195 32 0,180

Foreign obs. med 19 0,420 21 0,590 22 0,521 25 0,518 29 0,429 33 0,405 36 0,334 32 0,230 37 0,260

medians significantly

different yes yes yes yes yes yes yes no no

REGION

Moscow-based banks obs. med 567 0,279 571 0,344 586 0,338 615 0,321 646 0,334 663 0,335 620 0,322 357 0,278 469 0,280

Regional banks obs. 588 593 595 598 685 663 618 499 546

med 0,259 0,296 0,271 0,258 0,247 0,240 0,201 0,187 0,180

medians significantly

different no yes yes yes yes yes yes yes yes

SIZE CATEGORIES

Small obs. 489 442 437 412 443 442 412 285 338

med 0,184 0,249 0,253 0,274 0,281 0,277 0,253 0,234 0,290

Medium-sized obs. med 490 0,218 442 0,295 437 0,289 413 0,284 444 0,277 442 0,291 413 0,263 285 0,230 338 0,220

Large obs. 487 439 434 410 441 439 410 283 336

med 0,298 0,370 0,323 0,288 0,288 0,279 0,254 0,200 0,180

The Big 3 obs. med 3 0,406 3 0,283 3 0,304 3 0,261 3 0,354 3 0,273 3 0,265 3 0,230 3 0,230

medians significantly

different yes yes yes no no no no no yes

DEPOSIT INSURANCE

(DI)

Included in DI obs. 801 802 802 649 632

med 0,265 0,268 0,226 0,199 0,185

Not included in DI obs. 419 434 436 206 172

med 0,316 0,329 0,336 0,315 0,290

medians significantly

different yes yes yes yes yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Table A.7.

Net interest margin to total loans by bank ownership, location, size and participation in the deposit insurance scheme

NET INTEREST MARGIN 1999 2000 2001 2002 2003 2004 2005 2006 2007

obs. 1423 1277 1262 1181 1280 1277 1229 761 942

TOTAL SAMPLE

med 0,023 0,035 0,029 0,036 0,033 0,029 0,028 0,025 0,024

OWNERSHIP GROUPS

Private

State-controlled

Foreign medians significantly different

obs. 1374 1229 1211 1129 1221 1214 1161 709 878

med 0,023 0,036 0,030 0,036 0,033 0,029 0,028 0,025 0,024

obs. 30 29 31 30 33 32 32 25 32

med 0,042 0,046 0,034 0,041 0,030 0,024 0,023 0,018 0,020

obs. 19 19 20 22 26 31 36 27 32

med 0,016 0,015 0,016 0,013 0,014 0,012 0,013 0,017 0,020

yes yes yes yes yes yes yes yes yes

REGION

Moscow-based banks

Regional banks medians significantly different

obs. 537 544 560 575 612 630 611 311 434

med 0,014 0,020 0,018 0,028 0,026 0,025 0,028 0,025 0,026

obs. 575 574 578 583 668 647 618 450 508

med 0,046 0,053 0,040 0,044 0,038 0,032 0,027 0,024 0,023

yes yes yes yes yes yes no no yes

SIZE CATEGORIES

Small

Medium-sized

Large The Big 3

medians significantly different

obs. 454 411 405 368 406 406

med 0,040 0,053 0,040 0,048 0,047 0,039

obs. 482 431 425 403 436 433

med 0,030 0,039 0,031 0,036 0,032 0,029

obs. 484 433 429 407 435 435

med 0,016 0,024 0,024 0,029 0,027 0,024

obs. 3 2 3 3 3 3

med 0,006 0,006 0,008 0,021 0,015 0,014

yes yes yes yes yes yes

406 246 286

0,036 0,032 0,032

410 263 327

0,028 0,023 0,024

410 249 326

0,023 0,021 0,019

3 3 3

0,017 0,018 0,014

yes yes yes

DEPOSIT INSURANCE SCHEME

Included in DIS

Not included in DIS medians significantly different

obs. 777 778 785 733 797 799 802 587 694

med 0,033 0,040 0,035 0,038 0,033 0,029 0,026 0,024 0,022

obs. 349 347 356 355 403 418 424 173 217

med 0,023 0,028 0,024 0,033 0,031 0,029 0,031 0,028 0,029

yes yes yes yes yes no yes yes yes

Note: In order to utilize all the the first quarter of each year.

available data, all the indicators are calculated at the end of

Table A.8.

Variable description

VARIABLE DESCRIPTION

Size Capitalization Loans to assets Nonperforming loans Loan loss provisions Liquidity ratio Loans to individuals Net interest margin Credit growth Oil price GDP growth total assets, mln.RUB ratio of equity to total assets ratio of total loans (to nonfinancial clients) to total assets ratio of nonperforming loans to total loans ratio of loan loss provisions to total loans ratio of liquid assets to total assets ratio of loans to individuals to total loans the difference between interest income from loans to customers and interest expense paid on customer deposits as a proportion of total loans annual change in loans to nonfinancial clients average export price for crude oil for preceding quarter ($ per ton), Rosstat quarterly growth of real GDP, Rosstat

DUMMY VARIABLES

Foreign bank State-controlled bank Moscow bank Big 3 Deposit insurance system 100% foreign owned bank as reported quarterly by the CBR bank included in the list of state banks by Vernikov (2007) bank's headquarters are located in Moscow three largest banks by assets: Sberbank, VTB and Gazprombank bank entered DIS before the end of the first quarter of 2005

REGULATION RATIOS

N1 - capital adequacy ratio N2 - quick liquidity ratio N3 - current liquidity ratio N7 - maximum large credit risk bank's equity capital to the overall risk-weighted assets minus the sum of the reserves created for the depreciation of securities and possible losses sum of the bank's highly liquid assets to the sum of the bank's liabilities on demand accounts sum of the bank's liquid assets to the sum of the bank's liabilities on demand account and accounts up to 30 days percentage of the total amount of large credit risks (which is the sum of the bank's risk-weighted claims to one borrower) in the bank's equity capital

Risk-taking by Russian Banks: Do Location, Ownership and Size Matter?

Fungacova Z., Solanko L.

The Russian banking sector has experienced enormous growth rates during the last 6-7 years. The rapid growth of assets has, however, contributed to a decrease in the capital adequacy ratio, thus influencing the ability of banks to cope with risk. Using quarterly data spanning from 1999 to 2007 on all Russian banks, we investigate the relationship between bank characteristics and risk-taking by Russian banks. The analysis of financial ratios reveals that, on average, the risk levels are still below those observed in Central and Eastern Europe. Combining the group-wise comparisons of financial ratios and the results of insolvency risk analysis based on fixed effects vector decomposition, three main conclusions emerge. First, controlling for bank characteristics, large banks have higher insolvency risk than small ones. Second, foreign-owned banks exhibit higher insolvency risk than domestic banks and large state-controlled banks are, unlike other state-controlled banks, more stable. Third, we find that the regional banks engage in significantly more risk-taking than their counterparts in Moscow.

Keywords: bank risk-taking, banks in transition, Russia.

1. Introduction

Banking sectors in most countries of the Commonwealth of the Independent states (CIS), Russia included, have experienced nearly phenomenal growth rates during recent years. As a consequence of the dramatically improved macroeconomic situation

We are grateful for the valuable comments and suggestions we have received from Stephan Barisitz, Randall Filer, Michael Funke, Iftekhar Hasan, Esa Jokivuolle, Iikka Korhonen, Aaron Mehrotra, Tuomas Takalo, Laurent Weill, participants of the BOFIT seminar in Helsinki (March 2008), the IX International Academic Conference in Moscow (April 2008), the Sixth ESCB Workshop on Emerging Markets in Helsinki (May 2008) and the 10th EACES conference in Moscow (August 2008).

All opinions expressed are those of the authors and do not necessarily reflect the views of the Bank of Finland.

Fungacova Z. - Bank of Finland Institute for Economies in Transition (BOFIT). E-mail: zuzana.fungacova@bof.fi

Solanko L. - Bank of Finland Institute for Economies in Transition (BOFIT). E-mail: laura.solanko@bof.fi E-mail: laura.solanko@bof.fi

Статья поступила в Редакцию в январе 2009 г.

and important legislative changes, the ratio of banking sector assets in Russian GDP grew annually by more than 2 percentage points between 2001 and 2007. This ratio exceeded 60 percent by the end of 2007. Simultaneously, bank credit to the private sector has more than doubled to 30 percent of GDP.

With the rapid growth of total assets, deposits and loan stocks, Russian banks are increasingly assuming their role as financial intermediaries channeling household deposits and foreign borrowing into domestic corporate credits. This necessarily causes changes in the banks' assets and liability structures, attitudes towards risk-taking and risk management. Rapid credit growth is likely to increase (potential) banking sector risks. On the other hand, the ongoing financial deepening also indicates that the Russian banking sector is beginning to have an impact on private sector (both corporate and individual) behaviour and investments. That is, banks in Russia as well as in most other transition economies, are starting to look like banks elsewhere. They are by no means problem-free, but the challenges they need to tackle are similar to what banks in other emerging economies face. Given their growing role in economic development, surprisingly little is known about these banks' risk-taking behaviour.

The development of the banking sector in transition economies, as well as the financial sector in general, have been studied extensively. Barisitz (2008) and Bonin and Wachtel (2003) [5, 11] provide excellent recent overviews. Many studies focus on the effects of bank privatization on their performance in transition countries [8, 9], but until recently risk-taking by banks in transition has been a largely neglected area of research. Recent literature on the Russian banking sector has focused on bank supervision and the introduction of the deposit insurance system [14, 18, 42], market discipline and deposit interest rates [31, 39] and the efficiency of banks [1, 2, 19, 33] .

A handful of recent papers provide cross-country evidence on bank risk-taking in emerging economies. Haselmann and Wachtel (2007) [28] use several accounting measures of bank risk to examine the risk-taking behaviour of banks in 20 transition countries including Russia. They analyze differences in risk measures by bank ownership, size and market share. Using survey data from the EBRD, they complement the analysis with various measures of institutional quality. The results suggest that there is no group of banks with excessive risk-taking and that an unsound institutional environment leads to higher capital holdings and less credit risk-taking by banks. Maechler et al. (2007) [36] examine the effect of various types of financial risks on the bank stability in 18 Central and Eastern European economies. Their results indicate that foreign banks tend to have a higher risk profile than domestic ones but there is no significant difference between the risk profiles of larger and smaller banks. Furthermore, credit growth relates to greater bank stability and only the acceleration of growth seems to add vulnerability.

To the best of our knowledge, no study on bank risk-taking has focused on Russia or any other CIS country. However, with its 1100 banking institutions, Russia in particular provides an extremely rich test case for analyzing risk-taking. Additionally, the large number of bank failures (more than 300 since year 20001)) highlights the fact that banking in Russia is still riskier than in most developed countries. Therefore examining the determinants of risk-taking is crucial for understanding the prospects for future economic growth. Furthermore, if Russia is to become a global financial centre,

1) For more details see: www.banki.ru.

a goal clearly stated by, e.g., President Medvedev in spring 2008, we need to know much more about the behaviour of Russian banking institutions.

Currently the Russian banking sector is extremely fragmented, with a few large banks and a great number of very small ones. Especially in comparison with Central European transition economies, the state has retained a large share of control whereas the role of foreign banks has been very limited. These two structural features have often been mentioned as the main hindrances to further banking sector reform and growth. In this paper we discuss the extent to which the characteristic features of the sector determine the risk-taking behaviour of Russian banks.

We use a large panel of practically all Russian commercial banks covering the post-1998-crisis period, from April 1999 to April 2007. The large, Moscow-based and state-controlled banks form the backbone of the Russian banking sector. In line with previous literature, we therefore focus on the effects of bank size and ownership structure on bank risk-taking. Furthermore, we control for the location of the banks to see if Moscow-based banksdiffer in their risk-taking habits. Additionally, we are able to examine the influence of what probably was the most important institutional change during the period, the introduction of a deposit insurance scheme, on the risk-taking of Russian banks.

In measuring risk-taking, we use two approaches. First, we conduct a univariate analysis of traditional financial risk ratios based on accounting data. Second, we run a regression analysis of bank insolvency risk measured by the z-score indicator. The two approaches produce similar results. First, risk-taking increases with size. Second, controlling for other bank characteristics, banking institutions located outside Moscow tend to bear higher risks. And finally, ownership does matter for risk taking. Surprisingly, foreign-owned banks are found to be more risk-taking than other banks.

The next section provides a brief overview of the Russian banking sector. Section three describes the data and provides group-wise comparisons of financial risk measures by size and ownership categories and by location, as well as by inclusion in the deposit insurance scheme. Section four complements the previous results with a z-score analysis and section five concludes the analysis.

2. Banking industry in Russia

After the crisis-ridden 1990's, especially the deep recession and financial collapse of 1998, the Russian economy has grown annually by more than six percent since 2000. The banking system has experienced rapid growth since 2001, when the sector recovered from the insolvencies and the complete lack of trust created by the 1998 turmoil. Trust in counterparties is still fairly low especially at the interbank markets and the sector is prone to rumors. This was exemplified in the summer 2004 when rumors and tight liquidity created a «mini-crisis» in the banking industry. The effects were, however, not long-lasting. Bank credit to the private sector as a ratio to GDP has more than doubled during the last decade. This is very rapid growth even compared to the fast-growing emerging economies of Central and Southeastern Europe. The resulting financial deepening has been supported by a stable macroeconomic environment, increasing incomes and institutional reforms.

Continuous economic growth, rising real incomes, declining inflation and public sector surpluses have enabled fast increases in the private sector credit share. The majority of credits are financed by private sector deposits, which have increased by

10 per cent annually during the last six years [16]. Also net foreign borrowing has increased, even though the level of total foreign liabilities in Russian banks is still relatively modest at on average below 20% of total liabilities.

Table 1.

Banking system indicators , % of GDP

2004 2005 2006 2007 2008

Total assets 42,1 41,7 44,8 51,9 61,0

net foreign asset position -1,4 -1,9 -2,7 -5,9 -9,0

credit to the private sector 20,2 22,8 25,2 29,9 37,2

o/w enterprises 18,3 19,6 20,3 22,9 28,2

o/w households 1,9 3,2 4,9 7,0 9,0

deposits by the public 23,6 24,4 27,3 32,0 37,0

o/w households 11,5 11,6 12,8 14,2 15,6

Note: Data concerns beginning of each period.

Source: Central Bank of Russia.

Furthermore, a number of important institutional reforms have undoubtedly helped fuel banking sector growth. The most important one was the introduction of the deposit insurance system (DIS). The federal law on compulsory deposit insurance was adopted in December 2003. The law made the formerly implicit guarantee of state-controlled banks explicit and outlined clear rules for banks entering the system. The Deposit Insurance Authority began its operations in 2004, and by the end of March

2005 the first 824 banks were admitted into the system. Most of the rejected banks were small, as the banks already admitted accounted for 98 percent of household deposits. This did raise some concerns on the entry requirements not being interpreted rigorously enough.

By the end of September 2005, when the deadline for joining the system expired, 927 banks out of the 1150 applicants were admitted [14]2). During 2006-2007 Central Bank of Russia (CBR) gradually revoked the licenses to attract household deposits from banks not included in the system. Initially private deposits up to RUR 100000 were covered in full. Later the coverage limit was raised to RUR 190000 in August

2006 and to RUR 400000 in March 20073). During 2003-2005 also several other important laws, e.g., clarifying the rules for mortgage lending and mortgage-backed securities, were enacted. The law from 2005 gave the framework for the operations of private credit bureaux.

2) In order to pacify depositors during the mini-banking crisis of summer 2004, the government enacted a law granting temporary deposit insurance to all banks. Therefore, irrespective of possible inclusion in the deposit insurance system, all Russian banks were guaranteed blanket deposit insurance for deposits up to RUR 100000 from July 2004 until the end of 2006.

3) The limit was further raised to RUR 700000 in October 2008. See: http://www.asv.org.ru/in-surance/.

During the last few years Russian banks have intensively diversified into household lending, especially mortgages, as well as lending to SMEs. Credit maturities have also increased and maturities of over three years are not uncommon. The volumes of mortgage lending are, however, still tiny as less than 10% of homes in Russia are bought using a mortgage (Interfax, 2008). Another remarkable recent trend is the continuing de-dollarization of banking assets and liabilities. Like many transition countries, Russia was heavily dollarised and immediately after the 1998 crisis the use of dollars was very widespread. The share of foreign currency loans has now stabilized at below 25% of corporate loans. Corporate borrowers typically have a significant portion of their earnings in foreign currencies, so currency mismatches should not pose a systemic risk.

In light of all these changes, the structure of the Russian banking sector has remained surprisingly unchanged. The large, state-controlled banks still dominate the market. Even though the number of banks has decreased from 2084 at the end of 2000 to a mere 1243 by the end of 2007, the great majority of the banks are still tiny and can hardly be called banks. At the end of 2007 some 900 banks had the right to attract household deposits and only 300 banks had a general banking license. The foreign ownership share remained fairly limited as evidenced by the Table 2 below. There were 202 banks with a foreign ownership at the end of 2007, 62 of them fully foreign-owned.

Table 2.

Bank ownership in selected countries in 2005

Number of banks Number of foreign-owned banks, % of total Asset share of foreign-owned banks, % of total Domestic credit to private sector (% of GDP)

Estonia 13 77 99,4 57

Slovak Republic 23 70 97,3 34,7

Czech Republic 36 75 84,4 35,7

Lithuania 12 50 91,7 41,3

Hungary 38 71 82,6 49,8

Poland 61 82 74,2 29,2

Latvia 23 43 57,9 59

Slovenia 25 36 22,6 56,4

Russia 1253 4 8,3 26,1

Source: EBRD Transition Report 2006.

Our dataset ends in April 2007, just before the the global credit crunch caused by the subprime market problems in the US started to evolve. Initially the Russian banking industry was only mildly affected, in large part thanks to increasing crude oil prices that provided ample liquidity in the domestic market. Along with falling crude oil prices and drastically deteriorating situation at the international financial markets also the Russian banking sector began to face serious problems by the end of 2008.

3. Measuring risk - financial and regulation ratios

3.1. Data

Our dataset covers most of the banks operating in Russia over the period of April 1999 - April 2007. It consists of banks' quarterly balance sheets and profit and loss accounts. Regulatory ratios calculated by the Central Bank of Russia (CBR) are also partially included in our data and we use them in the analysis to support our main results. The data are provided by the financial information agency Interfax and originated in the Central Bank of Russia. For a more detailed description of the dataset used, see Karas and Schoors (2005) [32]. As the sample period starts in 1999, our results are not directly influenced by the financial crises of August 1998. The data constitutes an unbalanced panel, because there were banks entering and leaving the market due to mergers or failures. A brief overview of the main variables based on summary statistics is provided in Table A.1 in the appendix.

The banks are divided into different subgroups by size, ownership and location as well as inclusion in the deposit insurance system. We use the book value of total bank assets as a measure of size4). Bank size is especially important in Russia, where a handful of the largest banks account for most of the banking sector assets. At the end of 2006, large state-controlled banks accounted for about 40% of the sector assets [15]. Taking into account the overly concentrated nature of the Russian banking sector, we separate for the three largest banks (Sberbank, VTB and Gazprombank). In general, due to more possibilities for diversification and better access to financial markets, large banks are supposed to be less risky. Nevertheless, as Demsetz and Strahan (1997) [24] point out, large banks offset their potential benefits from diversification with lower capital ratios and more risky loan portfolios. Also empirical evidence on the relationship between size and risk has produced slightly mixed results [28, 29].

As for ownership, we distinguish among three ownership groups to determine majority ownership: state-controlled, foreign and domestic private banks. The foreign ownership dummy variable is based on the CBR data on 100% foreign-owned banks published quarterly. State-controlled banks are defined using the list provided in Ver-nikov (2007)5). Due to its special role as a state development bank, we do not include Vneshekonombank (VEB).

Ownership may be important for risk-taking behaviour for various reasons. State-owned banks are often assumed to take higher risks than the private ones. The underlying reasons differ according to one's view on the character of state-owned banks. Sapienza (2004) [41] distinguishes three alternative views. The social view suggests that state banks intervene to correct for the market failure caused by private

4) We first separate the three largest banks as a group of their own. The rest of the banking sector is divided into three groups. Small banks are those with total assets below 33rd per-centile, medium banks have assets between 33rd and 66th percentiles and the large ones have total assets above the 66th percentile in every time period. Alternative measures of size based on the market share of the aggregate domestic credit as well as participation in the interbank market provide us with a very similar distribution of banks into subgroups and therefore we only use total assets as a proxy for bank size.

5) This list largely overlaps with the other lists of state-controlled banks used by Karas et al. (2008) [33]. Moreover, our number also corresponds to the number of government-controlled banks in the Bank Supervision Report (2006).

banks, which «cherry-pick» the best customers and would leave the not very profitable ones without financial services. This view implies that state banks are engaged in more risky and less profitable operations but possibly enjoy soft budget constraints. The political view sees state banks as well as state enterprises more as a mechanism for pursuing politicians' private interests, such as maximizing employment or delivering favours for political protégées. This view implies that state banks may be forced to lend on a non-commercial basis i.e. due to political or other reasons. The agency view sees state banks as basically benevolent maximizers of social welfare but plagued by corruption and misallocation. Recent evidence from industrialized countries [20, 29] suggests that state-owned banks typically exhibit higher risk than other types of banks.

Studies on transition economies have, however, produced mixed results [21, 36]. In transition economies state-owned banks may be less efficient and more risk-prone due to Soviet legacies, unrestructured management or soft budget constraints. These findings, usually based on Central European countries (see e.g. [8]), are challenged by Karas et al. (2008) [33], who show that in Russia state-owned banks are not less efficient than domestic private banks.

Foreign-owned banks may have a different risk profile due to less local expertise and fewer local connections compared to the domestically owned banks. Their operations may also be less risky since they might often be able to cherry pick the most creditworthy borrowers in an emerging market [7]. Additionally, these banks can often rely on strong parent companies to provide them with access to better risk management techniques and possible diversification of country risk. On the other hand, foreign ownership may aggravate risks if parent banks tend to stress rapid credit growth in order to relieve tightening interest margins at home. Moreover, integration into the global financial system has also highlighted new issues related to risk management and financial vulnerability.

Foreign bank entry has been one of the decisive factors shaping banking sector development in Central and Eastern European transition countries. The available empirical evidence supports the common view that foreign-owned banks are more efficient than other types of banks in these countries ([5, 8, 9] and references therein). Furthermore, there is a growing literature exploring the effects of the presence of foreign-owned banks on domestic credit markets in emerging economies6). The role of foreign-owned banks in Russia has been dramatically different from those in the Central European banking sector. The share of foreign capital in the Russian banking sector was tiny up until spring 2007 as no major privatizations had taken place. The Russian banking sector is clearly more distant (both geographically and culturally) and therefore less attractive than the new and prospective EU member countries. Moreover, acquiring a large market share is not as easy as it was in Central Europe. Nevertheless, the foreign-owned banks operating in Russia may be extremely important as a benchmark for domestic ones and it is therefore most interesting to examine if they differ in their risk-taking.

6) Mostly the results on the benefits of the foreign bank presence are mixed. Detragiache et al. (2008) [25] show that banks give fewer loans after being acquired by a foreign investor. Clarke et al. (2005) [17] find that foreign banks make more loans to SMEs than domestic ones. Foreign banks may be reluctant to lend to opaque borrowers, but induce domestic banks to lend to them [22]. Giannetti and Ongena (2008) [26] suggest that foreign banks enhance access to credit, especially where financial development is low.

The division by ownership and size is rather standard. A bank's location within a single country and its inclusion in the deposit insurance scheme are more specific to Russia. Economic developments in different parts of Russia vary a lot. About half of the Russian banks are located in Moscow. The other half, located in the other regions of the Russian Federation, are mainly small banks constituting only 15% of the total banking sector assets. It has been occasionally argued that regional banks are more inclined to lend to local enterprises and to small and medium-sized businesses, thereby promoting growth more than Moscow-based banks. Moscow-based banks, on the other hand, are more active in interbank money markets. If true, this should also be reflected in differences in risk measures. Therefore we split the sample into two depending on the location of the bank's headquarters in Moscow or elsewhere in the Russian Federation. The division into regional and non-regional banks is unavoidably somewhat arbitrary as a large number of banks headquartered both in and outside Moscow have wide networks outside their home region. But the division used is the best available approximation for Moscow and non-Moscow banks. If the banks do not differ in their risk-taking based on the location of their headquarters, the division should not be significant in our analysis. But, as will be seen, the statistically significant result survives all our robustness checks.

Russia adopted a deposit insurance system in 2004 with the majority of banks screened and admitted into the system by end-March 2005. The deposit insurance system was expected to increase the confidence in and stability of the banking sector, as well as to level the playing field between large and small banks. The academic literature on deposit insurance increasingly emphasizes that explicit deposit insurance has the potential to affect bank risk-taking. Since it reduces depositors' incentives to monitor banks, it may encourage risk-taking and imprudent banking practices. The Russian data offers us a unique opportunity to test whether the introduction of a deposit insurance system affects bank risk-taking in the short run. We consider two groups of banks based on the point at which they entered the system. We create a dummy variable indicating if the bank was included into the system in the «first wave», by end - March 2005. Inclusion of the banks in the deposit insurance system is defined using the information from the Russian Deposit Insurance Agency.

3.2. Risks faced by banks and corresponding financial ratios

Banking is by nature a business of balancing risks. There is, however, no single, universal measure that could be used to assess risk-taking behaviour by banks. Thus, we rely on two different approaches. The first one is based on a univariate analysis of financial risk ratios, which are either calculated using the accounting data or belong to the regulatory ratios used by the central bank. We analyze different categories of financial risk separately by employing the relevant financial ratios as well as regulation ratios used by the CBR (for definitions, see Table A.8 with a description of variables in the appendix). Furthermore, we also test the significance of the differences in financial risk ratios among different subgroups of banks7). The second approach, discussed in section four, relies on the regression analysis of bank insolvency risk as measured by the z-score indicator.

7) We use a nonparametric K-sample test on the equality of medians.

Capitalization

Capitalization is calculated as a ratio of equity to total assets and it serves to measure leverage risk. Due to rapid asset growth, the level of capitalization declines during the period analyzed (see Table A.2 in the appendix). Capitalization is, however, still higher than in most other transition countries as reported in Haselmann and Wachtel (2007) [28]. On average, capitalization decreases with size and thus small banks tend to have higher capital ratios than larger banks. This is in line with the «too big to fail» hypothesis as well as with the perceived difficulties smaller banks face in accessing interbank markets in Russia. Larger banks in general have better opportunities for risk diversification and thus also benefit from lower costs of funding [37].

The capitalization of private banks is significantly higher than that of state and foreign banks during the whole period under review. These banks, unlike state-controlled or foreign banks, usually do not have a kind of «backup» in the form of the state or a strong parent company abroad. That is most probably the reason why they hold a higher proportion of equity capital. Foreign banks are slightly better capitalized than state banks, which is consistent with the results for the CIS in [21]. Banks located outside Moscow tend to maintain lower equity, but the gap between regional and Moscow banks has decreased since 2006 and thus the difference between these two groups of banks is no longer significant. Banks included in the DIS maintain a significantly lower equity than the other banks. There are two possible explanations for this. The first one concerns moral hazard issues connected with the participation in the deposit insurance scheme. The other is selection bias. It indicates that the banks entering the system were the better ones, which, based on their results, were obvious candidates for inclusion immediately when the system was introduced.

The CBR regulation ratio N1 used to assess capital adequacy8 confirms these trends as well. Even though the capital adequacy ratio has declined in recent years, its average value of 14,5% for November 2006 [15] still clearly exceeds the minimal requirements set by the central bank9). This indicates that Russian banks on average tend to keep slightly higher capital buffers than banks in the EU-25 countries as Jokipii and Milne (2008) report [30]. It is, however, clear that relatively large capital buffers at the beginning of our sample period are a natural reaction to the uncertainty following the crisis of 1998. The gradual decrease of capital buffers is then to a certain extent the result of the improvements in the macroeconomic environment. Nevertheless, it may also indicate that the operations of Russian banks are becoming more efficient or that the institutional environment is improving [10, 28]. The unfavourable global development resulting from the sub-prime crisis and liquidity problems in the second half of 2007 made banks more cautious again and the majority of banks increased their capital adequacy ratios towards the end of 2007 [16].

8) Unlike the indicator of capitalization, the N1 ratio is for most of the banks available only until 2005.

9) The Financial Stability Report 2006 issued by the central bank reports that according to Bank of Russia Instruction № 110_I, dated January 16, 2004, the minimum capital adequacy ratio for a bank (N1) is 10% if the bank has a capital of at least 5 million euros and 11% if the bank has a capital of less than 5 million euros. Only 11 credit institutions violated the capital adequacy ratio in 2006 and 19 in 2005 (Bank of Russia Financial Stability Report, 2006).

Credit risk

Analyzing credit risks is becoming increasingly important in Russia due to its rapid credit growth. The increase in the loans to total assets ratio (see Table A.3 in the appendix) suggests that the growth of lending has been higher than the growth in total assets, implying a gradual shift towards riskier operations of banks. Domestic banks have significantly higher lending ratios than foreign banks, whereas regional banks tend to lend more than Moscow-based ones10). On average, however, the total loans to total assets ratio in our sample is comparable with the sample of transition economies as reported in Haselmann and Wachtel (2007) [28]. Similar to our expectations, banks that belong to the deposit insurance system lend more. There are again two possible explanations for this. The first one suggests that banks in the DIS may take more risks as they are backed up by the system. The latter indicates that insured banks are on average better and more efficient and therefore they are able to bear higher risks.

One of the most commonly used indicators of credit risk is the ratio of nonper-forming loans (NPL) to total loans. The share of NPLs in Russia has indeed increased during the last years, but the levels are not yet anywhere close to becoming alarming. The median levels based on our calculations (see Table A.4 in the appendix) are still below the quality level of 1,5 per cent recommended by Grier (2001) [27]. It is, however, necessary to bear in mind that this is an ex post measure of the risks assumed by banks. When considering banks by ownership, state-controlled banks exhibit a significantly higher ratio of nonperforming loans than others. One might take this as indirect evidence of state-controlled banks' lending, willingly or unwillingly, to any customer, also to the uncreditworthy one. It is, however, interesting to note that the share of NPLs among the state-controlled banks has stayed basically unchanged in recent years. The recent increase in the NPL share has been caused mainly by private domestic banks. On the other hand, foreign banks have the lowest level of NPLs, which may reflect their relatively short period of operation on the Russian market, better credit risk management, or both.

The ratio of NPLs is increasing with the bank's size, which suggests that larger banks are able to sustain a larger proportion of NPLs. The difference between small and large banks is, however, gradually decreasing. The shrinking of this gap is the result of both an increase in the NPL ratio of small banks and a decrease among the large ones. Despite this development, the variation between banks of different sizes still remains significant. There are significant differences in the proportion of NPLs by location as well. Even though regional banks still tend to have a larger ratio of NPLs, similar to when we account for size, the gap between Moscow and regional banks has decreased recently. There are also differences between banks that are part of the deposit insurance system and the ones that are not. The ones included in the scheme have in general higher nonperforming loan ratios, which can be a natural consequence of higher lending by these banks.

Since banks with nonperforming loans are obliged to make loan loss provisions, a comparable measure of credit risk is the ratio of loan loss reserves to total loans. Its development basically corresponds to changes in the proportion of nonperforming

10) The underlying reasons for the different asset structure of regional and Moscow-based banks may include variations in fixed assets like buildings and branch-office networks. This issue would clearly merit a study of its own.

loans (see Table A.4 in the appendix). The proportion of loan loss reserves in total loans is the lowest for the foreign-owned banks. Even though the proportion of loan loss reserves was the highest for the three largest banks in 1999, nowadays this ratio is basically the same for banks of all sizes. This seems to serve as evidence for the special position of these state-controlled banks. The loan loss indicator further suggests that the deposit insurance scheme implementation contributed to changes in loan loss reserves. Before the deposit insurance scheme was implemented, loan loss reserves were significantly higher for the banks that later entered the scheme. However, with the implementation of the scheme, reserves in the banks not included in the system increased and they are higher compared to the banks that are part of the DIS.

Maximum large credit risk is a regulation ratio that measures the proportion of the total amount of large credit risks11) in a bank's equity capital. It increases over time and tends to be higher for the state-controlled banks and for the regional banks. This could indicate that these banks have close connections with large state-controlled or regional companies. The maximum large credit risk ratio is also higher for larger banks with the exception of the three largest ones. Moreover, it is significantly lower for the banks outside the deposit insurance system, which once again indicates that banks that are part of the system are able to engage in relatively more risky activities.

Even though our analysis of credit risk measures suggests that the operations of state-controlled banks tend to be relatively riskier than the others, the comparison of the credit risk indicators to the corresponding figures in other countries as well as to the critical values indicated in the literature suggest no excessive risk-taking. Our results are thus in line with the CBR [15] in that, on average, the credit risk of Russian banks remains moderate.

Liquidity risk

The Russian banking sector's liquidity as measured by the ratio of liquid to total assets has decreased slightly in recent years, but its level, reported in Table A.6 in the appendix, is still comparable to the other transition countries as well as to the quality level recommended by Grier (2001) [27]. An analysis of the regulatory ratios of quick and current liquidity (see Table A.8 in the appendix for detailed definitions) confirms that they have remained basically unchanged. Foreign banks and Moscow-based banks exhibit the highest level of liquidity during the whole period under review. One possible explanation for this phenomenon is that Moscow-based banks are on average less engaged in traditional banking operations (collecting retail deposits and channeling them into corporate loans) than regional banks. Furthermore, Moscow-based banks tend to be more active in interbank money markets and therefore have a larger proportion of their assets in a highly liquid form. This difference in bank operations is reflected in the increasing gap in the liquidity indicator between Moscow and regional banks. The finding is a corollary to the finding that, on average, the share of loans in total assets is lower for Moscow-based banks than for the other banks. Unlike the divisions by region and ownership, the distribution of banks by size does not indicate any significant differences in liquidity for banks of various sizes.

11) Large credit is the total sum of the bank's risk-weighted claims to one borrower (or a group of related borrowers) on credits.

Moreover, in line with the other credit risk indicators, the banks included in the deposit insurance scheme hold lower levels of liquidity and the gap between them and the other Russian banks has been increasing since 2005.

In general, high liquidity ratios can be interpreted as having a positive influence on stability at certain levels of liquidity. In the case of emerging economies, liquidity ratios may also be higher if the government does not actively intervene to meet funding gaps, financial institutions are risk-averse or if there are not enough opportunities for hedging [38]. In that case excessive liquidity could indicate structural problems. A bank may be highly liquid simply because: 1) it cannot rely on well-functioning interbank markets or other secondary markets such as those for securities; 2) it prefers to distance itself from «traditional» banking operations such as lending in favour of trading in, e.g., government securities; or 3) both.

Despite sufficient liquidity in general, there has been a lack of efficient mechanisms for interbank intermediation of liquidity. The Russian interbank market is relatively small even in comparison to other emerging markets [38]. This is especially the result of high segmentation and low trust on the interbank market [5], even among the big state-controlled banks. Russian banks are highly liquid but the banking system as a whole is not. Due to the lack of trust, the banking system is vulnerable to occasional liquidity shocks as experienced in summer 2004 and autumn 2007. This clearly complicates banks' liquidity management as well as the conduct of monetary policy in Russia.

Market risk

The net interest margin12) as a percentage of loans is often used as a proxy for the efficiency of financial intermediation, thus uncovering the health of the banking sector. Higher margins indicate lower efficiency and lower competition within the sector and thereby possibly also higher risk. Our analysis indicates that foreign banks have significantly lower net interest margins than private banks, even though recent developments suggest that the net interest margins of foreign banks have increased to the level of state-controlled ones (see Table A.7 in the appendix). In this respect, lower margins most probably reflect the greater efficiency of foreign banks which is connected to the support and know-how from their parent companies. Our indicators are thus in line with Karas et al. (2008) [33], who find that Russian state banks are more efficient than domestic private banks. The net interest margin decreases with the bank's size and therefore it is the lowest for the group of the three largest banks. Regional banks used to have significantly higher net interest margins. However, the situation has changed recently and consequently Moscow-based banks have slightly higher margins, which may suggest increasing efficiency and/or competition. After the implementation of the DIS, the net interest margins of the banks included in it decreased and became significantly lower than the margins of the other banks. This development may indicate a positive impact of the DIS introduction on the banking sector's competition and efficiency; however, more investigation is necessary to confirm this result.

12) The net interest margin is calculated as the difference between the interest income from loans to customers and the interest expense paid on customer deposits.

To sum up, the analysis of ratios measuring financial risk confirms significant differences among groups of Russian banks by size, location, ownership and participation in the DIS. Nevertheless, it is only based on the comparisons of unconditional medians. The following regression analysis provides more insight by uncovering also conditional correlations.

4. Measuring risk - bank insolvency risk (z-score)

In addition to the four classes of bank risk ratios, we use a measure for insolvency risk developed by Boyd and Graham (1988)13) [12] that has been increasingly used in the banking literature. Different modifications of z-scores have been applied in the empirical cross-country [13, 20, 21, 29, 36] as well as single-country studies [34, 35].

The insolvency risk measure («z-score» hereafter) is a statistic indicating the probability of bankruptcy (bank failure). The z-score for each bank i at quarter j is calculated as:

(1) Zj = (ROAit + EQTAit) / a(ROA)it,

where ROAit and (ROA)it are sample estimates of the four quarters moving average and the four quarters standard deviation of bank i's returns on assets at quarters t to t - 3 and EQTAit is the four quarters moving average of the equity capital to assets ratio. A bank's return on assets is calculated as its one-quarter profit before taxes on the quarter's average total assets. A bank's equity to assets ratio is calculated as the equity capital on total assets at the end of a given quarter. As we used the four quarters (backward-looking) moving averages in constructing our insolvency measure as well as explanatory variables, the time span of our analysis effectively covers the years 2000-2006.

Statistically speaking, the z-score represents the number of standard deviations returns would have to fall in order to deplete a bank's equity, under the assumption of normality of the bank's returns. Boyd et al. (2006) [13], however, argue that «it (the z-score) does not require that profits be normally distributed to be a valid probability measure; indeed, all it requires is the existence of the first four moments of the return distribution». A higher z-score corresponds to a greater distance to equity depletion and therefore to lower risk and higher bank stability.

The z-score measure inherently depends on the assumption that the ROA, relying on profit and loss data, gives a useful approximation of a bank's financial health. Since our data is based on Russian accounting system standards, which stress formal reporting rather than economic meaning, it may be questioned whether our data fulfils that requirement [5]. Nevertheless, as we only compare Russian banks with each other, possible flaws in the accounting standards should not be over-emphasized. Moreover, we use the z-score indicator to uncover statistically significant conditional correlations, not causality.

13) This measure originated as a predictor of corporate bankruptcy [3].

4.1. Methodology

Our focus is on the effects of a bank's size, ownership, location and inclusion in the deposit insurance scheme on its insolvency risk (z-score). The bank's size is measured by a continuous variable (logarithm of total assets) whereas ownership, location and inclusion in the deposit insurance scheme are proxied by using corresponding dummy variables. The dummy variable for inclusion in the deposit insurance scheme is fully time-invariant whereas the dummy variables for ownership and location exhibit very little if any within variation. Therefore a standard fixed-effects model is likely to lead to inefficient estimates with very large standard errors14).

We remedy the problem by applying the fixed effects vector decomposition (FEVD) approach by Plumper and Troger (2007) [40]. The approach suggests estimating the model in three steps. First, our dependent variable is regressed only on the cross-section fixed effect and the time-varying factors. Second, the estimated fixed effect (unit effect) is decomposed into the part explained by the time-invariant variables and the unexplainable part (error term). Finally, the model including the unexplained part of the fixed effect is re-estimated by pooled OLS. By design, the remaining error term is no longer correlated with time-invariant variables. Plumper and Troger (2007) [40] show that FEVD estimates are superior (in root mean squared errors) to the traditional fixed effects estimation. In running the FEVD estimations, we use STATA's FEVD module.

We estimate the following model:

where

• z is the z-score for bank i at time t calculated as indicated in the equation

• size stands for the logarithm of total assets of bank i at time t;

• bankSpec is a set of bank i's specific ratios at time t including liquidity, credit growth and the share of loans to individuals in total loans;

• IA is a set of interaction dummy variables between a bank's size and bank-specific factors;

• owner is a set of dummy variables distinguishing among foreign, state-controlled and private banks;

• region is a dummy variable indicating Moscow headquarters of bank i at

time t;

• seas stands for seasonal (i.e. quarterly) dummy variables;

• depInsurance is a dummy variable indicating inclusion in the first wave of the deposit insurance system.

All the variables used in the regressions are four-quarter moving averages. Z-score and total asset variables are in natural logarithms. Bank-specific factors include credit

(2)

ln(z) it = a,. + blSi ze it + b2 (BankSpec) it + РЪ(Щ it + PA(seas ) +b5(Owner ) t + b6 (Region),. + b7( DepInsurance) t + ett,

(1);

14) For recent discussions on fixed-effect models with time invariant variables, see, e.g., [6, 43]. For a classic textbook approach using Hausman-Taylor procedures, see [44, p. 235-238].

growth, the liquidity ratio and the share of loans to individuals in total loans. A bank's size, ownership, location and inclusion in the first wave of the deposit insurance system are defined as in the analysis of bank risk ratios in the previous section. To remove potential outliers, 0,5% of both tails of each variable in every quarter was removed. Table A.8 in the appendix gives details of the variables used in the regressions.

A priori, the sign of the coefficient on a bank's size is indeterminate because large banks may be either stabilizing or risky for the banking system, as our previous analysis of risk ratios suggests. Bank-specific risks are captured by the measures of credit risk and liquidity risk. Credit risk is proxied by bank-by-bank credit growth as well as the ratio of loans to individuals to total loans. Liquidity risk is controlled for by introducing the liquidity ratio (liquid assets/total assets) to the model. A priori we do not have an expectation of the sign for these variables.

4.2. Estimation results

In order to analyze the relationship between a bank's size, ownership and location and the risk measured by the z-score, we estimate the model of equation (2) employing the fixed effects vector decomposition described above. The main results are shown in Table 3 below.

Several interesting findings emerge. First, the results consistently indicate that larger banks have significantly lower z-scores and thus higher insolvency risk15). Second, somewhat unexpectedly, foreign-owned banks consistently bear higher insolvency risk than domestic private banks. This result is fully in line with some earlier studies on emerging economies using z-scores as the risk measure [36]. The result naturally reflects the limitations of the risk measure used, as it partly originates from the lower capitalization ratios of the foreign banks. Furthermore, it is necessary to bear in mind that due to data limitations, our foreign ownership dummy variable only accounts for banks that are fully foreign-owned. The overall effect of state ownership on a bank's insolvency risk is positive, i.e. state-controlled banks tend to be more stable. To investigate this result more closely, we add the interaction term of size and state control to our model. This interaction is positive and highly significant. At the same time, the estimated coefficient for the state-controlled dummy variable becomes negative. This indicates that only large state-controlled banks are driving our results and they are more stable than other state-controlled banks.

Third, the Moscow-based banks are always more stable than the regional banks. Based on the data available to us we can not determine the ultimate reason for this significant difference. The higher levels of capitalization in Moscow banks certainly play a role. The underlying reasons may include differences in bank operations, differences in banks' clientele and differences in bank supervision and regulation. Answering the highly interesting question on why the regional differences emerge would clearly merit a study of its own. Finally, similar to our expectations, banks that became part of the deposit insurance system in the first wave are more stable.

15) The z-score regressions are based on the full set of commercial banks, including the three large ones. As a robustness check we did run the estimations without the big three banks, but the results stay unchanged.

Finally, we conclude that the bank-specific characteristics do have a significant role in explaining insolvency risk. In line with earlier literature (e.g. [36]), we find that higher liquidity implies higher insolvency risk. We include an interaction variable of bank size and liquidity, which confirms that large liquid banks are more stable. The growth of a bank's loan stock is used to control for the credit risk. In line with Maechler et al. (2007) [36], its impact is positive in our estimations and this indicates higher stability. This result holds true for Moscow-based banks, while for regional banks the estimated coefficient is negative. We also control for the interaction of bank size and credit growth to see if credit growth affects small banks differently. We find that large banks with high credit growth are in fact more stable than the rest of the sector.

Table 3.

Estimation results

Estimated coefficient

Size (total assets) -0,262***

Loans to households (prop. of loans) -0,355***

Liquidity (liquid to total assets) -0,616***

Credit growth 0,015***

OWNERSHIP, LOCATION AND DEPOSIT INSURANCE

Deposit insurance 0,104***

Foreign bank -0,572***

State-controlled bank -0,534***

Moscow-based bank 0,501***

INTERACTIONS

Size and liquidity 0,054***

Size and credit growth 0,003***

Size and state-controlled 0,100***

Number of observations 27353

R2 0,426

Note: The table contains results for the FEVD regression. We report estimated coefficients as well as their significance (***significant at 1%, ** significant at 5% and * significant at 10%). Seasonal and yearly dummy variables as well as a constant term are included but not reported.

We test the robustness of our empirical results using several techniques.

• First, the results are robust to the exclusion of the three largest state-controlled banks (Sberbank, Gazprombank, VTB) from the sample.

• We split the sample into Moscow-based and regional banks. The FEVD regression model is run for the two subgroups separately. Except for the significance of credit

growth, other results for both subgroups are in line with the results of the main model reported above. Nevertheless, the model seems to fit a little bit better the Moscow-based banks, which account for about 85% of the banking sector assets.

• Finally, the results for the subsample of the 300 largest banks also correspond to our main results reported in Table 3. They only differ in the sign of the deposit insurance scheme dummy variable. In this case it is negative, which means that the banks that entered the system in the first wave are more risky. This is in line with the results of univariate analysis of financial ratios performed in the first part of the paper.

4.3. Z-score components

The z-score measure consists of three main components: the return on assets, capitalization and the volatility of the ROA. In order to investigate the contribution of each of them to explaining differences in the banks' stability, we run our basic model using all of these components as a dependent variable. This approach is in line with previous literature [21, 36]. We report the results of the z-score component regressions in the following, Table 4.

The first component of the z-score measure is capitalization16^ In this case, the fit measured by R2 is the highest of all the z-score components. The estimated coefficients are larger than for the other z-score components and almost all of them are significant. The estimated coefficients are mostly in line with the results of the main model, which indicates that the majority of the main results are driven by the contribution of the capitalization ratio. Larger banks have lower capitalization and this result undoubtedly drives our final result that banks with a higher amount of total assets are in general less stable. More liquid banks have lower capitalization, which indicates that banks substitute between liquidity and solvency risk. Nevertheless, liquid large banks tend to have higher capitalization. Both state-controlled and foreign ones are in general better capitalized than private ones. The effect of deposit insurance participation on capitalization is significantly negative. Banks in the deposit insurance system do seem to substitute deposit insurance for capital, or put in other words, take more risks for the same level of capital. This result is in line with earlier literature [23].

The second column contains results for the regression with the ROA as the dependent variable. Similar to the capitalization component of the z-score, almost all the estimated coefficients are significant for the ROA. However, the majority of their signs differ from the results in the main z-score regression. Higher credit growth as well as a higher share of loans to individuals in a bank's loans portfolio are positively related to profitability. Higher liquidity positively influences profitability as measured by the ROA. Given the fact that the average real interest rate on corporate loans was close to zero for much of the period, this is not entirely surprising. Many banks make more than half of their revenues from foreign currency operations. When accounting for a bank's ownership, foreign banks and state-controlled banks have a significantly higher ROA than domestic private ones. Large state-controlled banks are, however, less profitable. Banks included in the DIS in the «first wave» have significantly higher pro-

16) Capitalization is, similar to the calculation of the z-score, calculated as the four-quarter moving average. The other z-score components, the ROA and volatility of the ROA, are calculated in the same way.

fitability than the others, which is in line with our previous result indicating that better banks entered the system first. Moscow-based banks are in general less profitable.

Table 4.

Z-score component regressions

Capitalization ROA Volatility of ROA

Estimated coefficient Estimated coefficient Estimated coefficient

Size (total assets) -0,085*** 0,0002*** -0,003***

Loans to households -0,076*** 0,005*** 0,001**

Liquidity -0,181*** 0,003*** 0,0004

Credit growth 0,003*** 0,0002*** -0,0002***

OWNERSHIP, LOCATION AND DEPOSIT INSURANCE

Deposit insurance -0,010*** 0,001*** -0,002***

Foreign bank 0,092*** 0,002*** 0,009***

State-controlled bank 0,082*** 0,004*** 0,006***

Moscow-based bank 0,134*** -0,003*** 0,001***

INTERACTIONS

Size and liquidity 0,002** -2,8E-05 -0,001***

Size and credit growth -3,0E-04*** -2,3E-05*** -1,2E-05

Size and state-controlled 0,001 -0,001*** -0,0005**

Number of observations 27353 27353 27353

R2 0,785 0,356 0,361

Note: The table contains estimation results of the model described above for different z-score components. We report the estimated coefficients as well as their significance (* significant at 10%, ** significant at 5% and *** significant at 1%). Seasonal and yearly dummy variables as well as a constant term are included but not reported.

The last component of our risk measure is the volatility of the ROA as measured by the standard deviation. Most of the estimated coefficients in this regression are significant but have a different sign than the results presented in our main model. They are also lower in absolute values and therefore, unlike the measure of capitalization, they contribute less to the main results. Thus, the analysis of the z-score components indicates that the differences in the risk profiles of banks are mostly driven by the differences in capitalization.

5. Conclusion

Favourable macroeconomic conditions and important regulatory reforms have backed the rapid growth of Russia's banking sector during this decade. As the economy is increasingly monetized, the role of banks and other financial intermediaries in supporting the continuous growth of investments and private consumption is gaining more importance. Therefore the stability of the banking sector is even more crucial. Compared to most European countries the Russian banking sector is still rightfully characterized as small, regionally fragmented and dominated by a few large state-controlled entities.

On average, the Russian banking sector is believed to be in good financial shape as evidenced also by the Banking Supervision Reports of the CBR. For this paper we use a bank-level dataset on all Russian banks to examine how various measures of risk vary with a bank's size, ownership, location and inclusion in the deposit insurance system. The main objective is the detailed examination of how these various groups of banks differ in their attitudes to risk. We employ two approaches; group-wise comparisons of financial ratios and regression analysis using a z-score measure of bank insolvency risk. The analysis of financial ratios reveals that even though the ratios point to increasing risk over time, they are still on average well on the safe side within all groups of banks. The average levels are all above the regulatory minima set by the Russian Central Bank. Moreover, they are comparable to other transition economies. The rapid growth of the banking sector has not led to excessive risk-taking on average.

The regression analysis of the bank insolvency measure (z-score) proved to be a useful means of deepening the results of group-wise comparisons. Controlling for bank characteristics, large banks in Russia have higher insolvency risk than small ones. Second, in line with the previous literature on emerging economies, foreign-owned banks exhibit higher insolvency risk than domestic banks. Even though the foreign bank presence may in general greatly increase banking sector efficiency and widen the range of banking services available, foreign-owned banks in Russia seem to bear higher risks. The same holds true for the state-controlled banks; however, the large state-controlled banks are more stable than the others. Third, we find that the regional banks are significantly more prone to risk-taking than their counterparts in Moscow. Regional banks only account for a small fraction of the total banking sector assets, thus this finding should not be alarming for the banking sector as a whole.

All in all, we find that risk-taking by Russian banks is approaching levels comparable to other emerging economies. Further, factors similar to those in emerging European economies seem to explain levels of insolvency risk in Russia. We also briefly examined if inclusion in the Russian deposit insurance scheme has influenced a bank's insolvency risk. The results are mixed and further research on this topic is clearly needed.

* * *

References

1. Алескеров Ф.Т., Мартынова Ю.И., Солодков В.М. Анализ и оценка эффективности функционирования банков и банковских систем / / Сборник докладов VIII Международной научной конференции «Модернизация экономики и общественное развитие». М.: ГУ ВШЭ, 2008.

2. Головань С.В., Карминский А.М., Пересецкий А.А Непараметрические оценки эффективности российских банков. Презентация доклада на IX Международной научной конференции «Модернизация экономики и глобализация». М.: ГУ ВШЭ, 2008.

3. Altman E. Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy // The Journal of Finance 1968. № 23(4). P. 589-609.

4. Bank for International Settlements Foreign Direct Investment in the Financial Sector - Experiences in Asia, Central and Eastern Europe, and Latin America. BIS, Committee on the Global Financial System. 2005. June.

5. Barisitz S. Banking in Central and Eastern Europe 1980-2006. L.: Routledge, 2008.

6. Beck N. From Statistical Nuisances to Serious Modelling: Changing How We Think about the Analysis of Time-series-cross-section Data // Political Analysis. 2007. № 15. P. 97-100.

7. Bhaumik S.K.., Piesse J. Does Lending Behaviour of Banks in Emerging Economies Vary by Ownership? Evidence from the Indian Banking Sector // Economic Systems. 2008. № 32(2). P. 119-216.

8. Bonin J., Hasan I., Wachtel P. Bank Performance, Efficiency and Ownership in Transition Countries // Journal of Banking and Finance. 2005. № 29. P. 31-53.

9. Bonin J., Hasan I., Wachtel P. Privatization Matters: Bank Efficiency in Transition Countries // Journal of Banking and Finance 2005. № 29. P. 1927-1952.

10. Bonin J., Hasan I., Wachtel P. Financial Intermediation in Transition Countries: BOFIT Discussion Paper № 12. 2008.

11. Bonin J., Wachtel P. Financial Sector Development in Transition Economies // Financial Markets, Institutions and Instruments. 2003. № 12. P. 1-66.

12. Boyd J., Graham G. The Profitability and Risk Effects of Allowing Bank Holding Companies to Merge with Other Financial Firms: a Simulation Study. Federal Reserve Bank of Minneapolis. Quaterly Review. 1988. Spring.

13. Boyd J.H., De Nicolm G., Jalal A.M. Bank Risk Taking and Competition Revisited: New Theory and New Evidence: IMF Working Paper 06/297. 2006.

14. Camara M.K., Montes-Negret F. Deposit Insurance and Banking Reform in Russia: World Bank Policy Research Working Paper 4056. 2006.

15. Central Bank of Russia. Banking Supervision Report. 2006.

16. Central Bank of Russia. Bulletin of Banking Statistics. 2007, 2008. Various issues.

17. Clarke G., Cull R., Martinez M., Sanchez S. Bank Lending to Small Business in Latin America: Does Bank Origin Matter? // Journal of Money, Credit and Banking. 2005. № 37. P. 83-118.

18. Claeys S., Schoors K. Bank Supervision Russian Style: Evidence of Conflicts between Micro- and Macro-prudential Concerns // Journal of Comparative Economics. 2007. № 35. P. 630-657.

19. Caner S., Kontorovich V Efficiency of the Banking Sector in the Russian Federation with International Comparision // HSE Economic Journal. 2004. Vol. 8. № 3. P. 357-375.

20. De Nicolo G. Size, Charter Value and Risk in Banking: An International Perspective Board of Governors of the Federal Reserve System. International Finance Discussion Paper. № 689. 2001.

21. De Nicolo G., Loukoianova E. Bank Ownership, Market Structure and Risk: IMF Working Paper 07/215. 2007.

22. Dell'Ariccia G., Marquez R. Information and Bank Credit Allocation // Journal of Financial Economics. 2004. № 72. P. 185-214.

23. Demirguc-Kunt A., Kane E.J. Deposit Insurance Around the Globe: Where does it Work?: NBER Working Paper 8493. 2001.

24. Demsetz R., Strahan P. Diversification, Size, and Risk at Bank Holding Companies // Journal of Money, Credit and Banking. 1997. № 29. P. 300-313.

25. Detragiache E., Tressel T., Gupta P. Foreign Banks in Poor Countries: Theory and Evidence // Journal of Finance. 2008. № 55. P. 1133-1161.

26. Giannetti M., Ongena S. Lending by Example: Direct and Indirect Effects of Foreign Banks in Emerging Markets. 2008. Mimeo, available at ssrn.com/abstract=1085310.

27. Grier W. Credit Analysis of Financial Institutions. L.: Euromoney, 2001.

28. Haselmann R, Wachtel P. Risk Taking by Banks in the Transition Countries // Comparative Economic Studies. 2007. № 49. P. 411-429.

29. Iannotta G., Nocera G., Sironi A Ownership Structure, Risk and Performance in the European Banking Industry // Journal of Banking & Finance. 2007. № 31. P. 2127-2149.

30. Jokipii T., Milne A. The Cyclical Behaviour of European Bank Capital Buffers // Journal of Banking & Finance. 2008. № 32(8). P. 1440-1451.

31. Karas A., Pyle W., Schoors K Sophisticated Discipline in a Nascent Deposit Market: Evidence from Post-Communist Russia // Oxford Economic Papers. 2009. Forthcoming.

32. Karas A., Schoors K. Heracles or Sisyphus? Finding, Cleaning and Reconstructing a Database of Russian Banks: Ghent University Working Paper 05/327. 2005

33. Karas A., Schoors K., Weill L. Are Private Banks More Efficient than Public Banks? Evidence from Russia: BOFIT Discussion Paper № 3. 2008.

34. Konishi M., Yasuda Y. Factors Affecting Bank Risk Taking: Evidence from Japan // Journal of Banking and Finance 2004. № 28. P. 215-232.

35. Lin S.L., Penm J.H.W., Gong S., Chang Ch Risk-based Capital Adequacy in Assessing on Insolvency-risk and Financial Performances in Taiwan's Banking Industry // Research in Business and Finance. 2005. № 19. P. 111-153.

36. Maechler A. M., Mitra S., Worrell D. Decomposing Financial Risks and Vulnerabilities in Eastern Europe: IMF Working Paper 07/248. 2007.

37. McAllister P., McManus D. Resolving the Scale Efficiency Puzzle in Banking // Journal of Banking and Finance. 1993. № 17. P. 389-405.

38. Moreno R. The Changing Nature of Risks Facing Banks // The Banking System in Emerging Economies: How Much Progress Has Been Made?: BIS Papers № 28. 2006.

39. Peresetsky A.A., Karminsky A.M., Golovan S.V. Russian Banks' Private Deposit Interest Rates and Market Discipline: BOFIT Discussion Paper 2/2007. 2007.

40. Plumper T., Troger V. Efficient Estimation of Time-Invariant and Rarely Changing Variables in Finite Sample Panel Analyses with Unit Fixed Effects // Political Analysis. 2007. № 15(2). P. 124-139.

41. Sapienza P. The Effects of Government Ownership on Bank Lending // Journal of Financial Economics. 2004. № 72. P. 357-384.

42. Vernikov A. Russia's Banking Sector Transition: Where to?: BOFIT Discussion Paper 5/2007. 2007.

43. Wilson S., Butler D. A Lot More to Do: The Sensitivity of Time-Series Cross-Section Analysis to Simple Alternative Specifications // Political Analysis. 2007. № 15(2). P. 101-123.

44. Wooldridge J. Econometric Analysis of Cross Section and Panel Data. Cambridge: MIT Press, 2002.

Summary statistics of the main variables

Appendix

Table À.1.

Variable Obs Mean Median Std. dev.

Z-score (ln) 34700 4,25 4,20 1,24

Total assets 41382 4105 307 52706

Liquidity ratio 41380 0,33 0,28 0,22

Loan loss provisions 40130 0,07 0,03 0,12

Credit growth 33969 4,64 0,39 209,05

GDP growth 40971 0,02 0,06 0,10

Note: Summary statistics for the observations that are actually used in the z-score regression are not significantly different from these figures.

Table A.2.

Capitalization ratio of banks by ownership, region, size and inclusion in DIS

CAPITALIZATION 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1469 0,362 1322 0,333 1312 0,318 1237 0,322 1327 0,303 1323 0,278 1238 0,243 856 0,187 1015 0,190

OWNERSHIP GROUPS

Private obs. 1420 1271 1258 1182 1265 1258 1170 795 946

med 0,366 0,337 0,323 0,329 0,306 0,281 0,246 0,191 0,190

State-controlled obs. med 30 0,287 30 0,287 32 0,250 30 0,250 33 0,232 32 0,222 32 0,177 29 0,138 32 0,150

Foreign obs. med 19 0,111 21 0,175 22 0,236 25 0,258 29 0,239 33 0,236 36 0,206 32 0,177 37 0,160

medians significantly different yes yes yes yes yes yes yes yes yes

REGION

Moscow-based banks obs. med 567 0,378 570 0,359 586 0,350 614 0,354 643 0,328 661 0,308 620 0,275 357 0,195 469 0,190

Regional banks obs. 588 591 595 598 684 662 618 499 546

med 0,359 0,315 0,297 0,298 0,284 0,251 0,213 0,182 0,178

medians significantly different no yes yes yes yes yes yes no no

SIZE CATEGORIES

Small obs. 489 440 436 411 439 439 412 285 338

med 0,539 0,454 0,434 0,439 0,407 0,381 0,330 0,269 0,280

Medium-sized obs. med 490 0,387 441 0,349 438 0,306 413 0,307 444 0,301 442 0,281 413 0,237 285 0,180 338 0,190

Continued

CAPITALIZATION 1999 2000 2001 2002 2003 2004 2005 2006 2007

Large obs. 487 438 435 410 441 439 410 283 336

med 0,235 0,227 0,243 0,259 0,240 0,217 0,182 0,142 0,130

The Big 3 obs. med 3 0,112 3 0,244 3 0,248 3 0,254 3 0,183 3 0,180 3 0,128 3 0,128 3 0,160

medians significantly

different yes yes yes yes yes yes yes yes yes

DEPOSIT INSURANCE

SCHEME (DIS)

Included in DIS obs. med 801 0,284 801 0,255 802 0,213 649 0,172 632 0,162

Not included in DIS obs. med 419 0,367 522 0,338 436 0,312 207 0,258 172 0,251

medians significantly

different yes yes yes yes yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Loans to assets ratio by bank ownership, location, size and in the deposit insurance scheme

Table A.3. participation

LOANS TO ASSETS RATIO 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1469 0,481 1326 0,428 1313 0,485 1238 0,521 1331 0,535 1326 0,555 1238 0,582 856 0,614 1015 0,627

OWNERSHIP GROUPS

Private obs. 1420 1275 1259 1183 1269 1261 1170 795 946

med 0,481 0,431 0,491 0,524 0,538 0,556 0,584 0,616 0,628

State-controlled obs. med 30 0,431 30 0,418 32 0,474 30 0,520 33 0,531 32 0,591 32 0,594 29 0,633 32 0,669

Foreign obs. med 19 0,428 21 0,276 22 0,257 25 0,294 29 0,414 33 0,309 36 0,368 32 0,500 37 0,495

medians significantly

different no yes yes yes yes yes no no yes

REGION

Moscow-based banks

Regional banks

medians significantly different

obs. 567 571 586 615 646 663

med 0,425 0,401 0,451 0,493 0,496 0,506

obs. 588 593 595 598 685 663

med 0,462 0,437 0,505 0,541 0,564 0,596

620 357 469

0,515 0,550 0,561

618 499 546

0,635 0,651 0,659

yes yes yes yes yes yes yes yes yes

Continued

LOANS TO ASSETS RATIO 1999 2000 2001 2002 2003 2004 2005 2006 2007

SIZE CATEGORIES

Small obs. med 489 0,503 442 0,436 437 0,499 412 0,496 443 0,487 442 0,516 412 0,554 285 0,598 338 0,552

Medium-sized obs. med 490 0,486 442 0,459 438 0,479 413 0,522 444 0,555 442 0,578 413 0,585 285 0,62 338 0,631

Large obs. 487 439 435 410 441 439 410 283 336

med 0,443 0,395 0,478 0,538 0,545 0,568 0,596 0,622 0,671

The Big 3 obs. med 3 0,332 3 0,363 3 0,472 3 0,530 3 0,437 3 0,577 3 0,590 3 0,495 3 0,486

medians significantly

different yes yes no no yes yes no no yes

DEPOSIT INSU-

RANCE SCHEME

Included in DIS obs. med 801 0,556 801 0,583 802 0,610 649 0,631 632 0,654

Not included in DIS obs. med 419 0,490 525 0,497 436 0,503 207 0,516 172 0,595

medians significantly

different yes yes yes yes yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Table A.4.

Nonperforming loans to total loans by bank ownership, location, size and the deposit insurance scheme

NONPERFORMING LOANS 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1423 0,019 1275 0,008 1265 0,004 1181 0,003 1280 0,003 1277 0,003 1226 0,005 853 0,009 1009 0,007

OWNERSHIP

GROUPS

Private obs. 1374 1226 1214 1128 1220 1214 1159 792 940

med 0,019 0,008 0,004 0,003 0,003 0,003 0,005 0,009 0,007

State-controlled obs. med 30 0,022 30 0,014 31 0,005 30 0,014 33 0,009 32 0,008 32 0,008 29 0,010 32 0,008

Foreign obs. med 19 0,000 19 0,000 20 0,000 23 0,002 27 0,000 31 0,000 35 0,000 32 0,003 37 0,001

medians significantly

different no yes no yes yes yes yes no yes

REGION

Moscow-based banks obs. med 537 0,001 541 0,000 559 0,000 575 0,000 612 0,001 630 0,001 608 0,002 356 0,009 464 0,006

Regional banks obs. 575 574 578 582 668 647 618 497 545

med 0,040 0,018 0,009 0,006 0,006 0,006 0,008 0,009 0,008

medians significantly

different yes yes yes yes yes yes yes no yes

SIZE CATEGORIES

Small obs. 454 408 403 367 406 406 403 282 333

med 0,036 0,012 0,008 0,000 0,002 0,001 0,003 0,008 0,005

Medium-sized obs. med 482 0,011 432 0,008 428 0,003 404 0,003 436 0,004 433 0,003 410 0,004 285 0,007 337 0,005

Large obs. 484 432 431 407 435 435 410 283 336

med 0,020 0,007 0,003 0,004 0,005 0,005 0,007 0,010 0,009

The Big 3 obs. med 3 0,149 3 0,046 3 0,023 3 0,027 3 0,019 3 0,017 3 0,015 3 0,012 3 0,012

medians significantly

different yes no no yes yes yes yes yes yes

DEPOSIT INSURANCE

Included in DI obs. 797 798 802 647 630

med 0,005 0,005 0,007 0,008 0,009

Not included in DI obs. 403 419 424 205 172

med 0,001 0,001 0,002 0,010 0,005

medians significantly

different yes yes yes no yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Table A.5.

Loan loss provisions by bank ownership, location, size and participation in the deposit insurance scheme

LOAN LOSS PROVISIONS 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1423 0,054 1275 0,043 1264 0,030 1181 0,025 1280 0,024 1277 0,025 1226 0,033 853 0,036 1009 0,038

OWNERSHIP GROUPS

Private obs. 1374 1226 1213 1128 1220 1214 1159 792 940

med 0,055 0,043 0,030 0,025 0,025 0,025 0,035 0,038 0,039

State-controlled obs. med 30 0,061 30 0,042 31 0,025 30 0,031 33 0,027 32 0,022 32 0,025 29 0,029 32 0,032

Foreign obs. med 19 0,018 19 0,037 20 0,022 23 0,013 27 0,015 31 0,011 35 0,005 32 0,011 37 0,012

medians significantly

different yes no no yes no no yes yes yes

REGION

Moscow-based banks obs. med 537 0,025 541 0,022 559 0,016 575 0,018 612 0,024 630 0,022 608 0,039 356 0,053 464 0,051

Regional banks obs. 575 574 578 582 668 647 618 497 545

med 0,081 0,063 0,038 0,030 0,025 0,026 0,030 0,030 0,032

medians significantly

different yes yes yes yes no no yes yes yes

SIZE CATEGORIES

Small obs. 454 408 403 367 406 406 403 282 333

med 0,068 0,056 0,032 0,018 0,017 0,019 0,028 0,030 0,039

Medium-sized obs. med 482 0,038 432 0,037 428 0,027 404 0,025 436 0,023 433 0,021 410 0,030 285 0,036 337 0,037

Large obs. 484 432 430 407 435 435 410 283 336

med 0,057 0,043 0,030 0,030 0,031 0,032 0,042 0,042 0,039

The Big 3 obs. med 3 0,199 3 0,090 3 0,067 3 0,060 3 0,054 3 0,061 3 0,037 3 0,037 3 0,036

medians significantly

different yes yes no yes yes yes yes yes no

DEPOSIT INSU-

RANCE SCHEME

Included in DIS obs. med 797 0,026 797 0,027 802 0,031 647 0,032 630 0,036

Not included in DIS obs. med 403 0,021 480 0,021 424 0,042 206 0,066 172 0,059

medians significantly

different yes yes yes yes yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Table À.6.

Liquidity ratio by bank ownership, location, size and participation in the deposit insurance scheme

LIQUIDITY RATIO 1999 2000 2001 2002 2003 2004 2005 2006 2007

TOTAL SAMPLE obs. med 1469 0,236 1326 0,301 1311 0,291 1238 0,283 1331 0,284 1326 0,281 1238 0,256 856 0,222 1015 0,220

OWNERSHIP GROUPS

Private obs. 1420 1275 1257 1183 1269 1261 1170 795 946

med 0,231 0,299 0,287 0,279 0,276 0,278 0,255 0,221 0,220

State-controlled obs. med 30 0,334 30 0,328 32 0,315 30 0,325 33 0,296 32 0,269 32 0,224 29 0,195 32 0,180

Foreign obs. med 19 0,420 21 0,590 22 0,521 25 0,518 29 0,429 33 0,405 36 0,334 32 0,230 37 0,260

medians significantly

different yes yes yes yes yes yes yes no no

REGION

Moscow-based banks obs. med 567 0,279 571 0,344 586 0,338 615 0,321 646 0,334 663 0,335 620 0,322 357 0,278 469 0,280

Regional banks obs. 588 593 595 598 685 663 618 499 546

med 0,259 0,296 0,271 0,258 0,247 0,240 0,201 0,187 0,180

medians significantly

different no yes yes yes yes yes yes yes yes

SIZE CATEGORIES

Small obs. 489 442 437 412 443 442 412 285 338

med 0,184 0,249 0,253 0,274 0,281 0,277 0,253 0,234 0,290

Medium-sized obs. med 490 0,218 442 0,295 437 0,289 413 0,284 444 0,277 442 0,291 413 0,263 285 0,230 338 0,220

Large obs. 487 439 434 410 441 439 410 283 336

med 0,298 0,370 0,323 0,288 0,288 0,279 0,254 0,200 0,180

The Big 3 obs. med 3 0,406 3 0,283 3 0,304 3 0,261 3 0,354 3 0,273 3 0,265 3 0,230 3 0,230

medians significantly

different yes yes yes no no no no no yes

DEPOSIT INSURANCE

(DI)

Included in DI obs. 801 802 802 649 632

med 0,265 0,268 0,226 0,199 0,185

Not included in DI obs. 419 434 436 206 172

med 0,316 0,329 0,336 0,315 0,290

medians significantly

different yes yes yes yes yes

Note: In order to utilize all the available data, all the indicators are calculated at the end of the first quarter of each year.

Table A.7.

Net interest margin to total loans by bank ownership, location, size and participation in the deposit insurance scheme

NET INTEREST MARGIN 1999 2000 2001 2002 2003 2004 2005 2006 2007

obs. 1423 1277 1262 1181 1280 1277 1229 761 942

TOTAL SAMPLE

med 0,023 0,035 0,029 0,036 0,033 0,029 0,028 0,025 0,024

OWNERSHIP GROUPS

Private

State-controlled

Foreign medians significantly different

obs. 1374 1229 1211 1129 1221 1214 1161 709 878

med 0,023 0,036 0,030 0,036 0,033 0,029 0,028 0,025 0,024

obs. 30 29 31 30 33 32 32 25 32

med 0,042 0,046 0,034 0,041 0,030 0,024 0,023 0,018 0,020

obs. 19 19 20 22 26 31 36 27 32

med 0,016 0,015 0,016 0,013 0,014 0,012 0,013 0,017 0,020

yes yes yes yes yes yes yes yes yes

REGION

Moscow-based banks

Regional banks medians significantly different

obs. 537 544 560 575 612 630 611 311 434

med 0,014 0,020 0,018 0,028 0,026 0,025 0,028 0,025 0,026

obs. 575 574 578 583 668 647 618 450 508

med 0,046 0,053 0,040 0,044 0,038 0,032 0,027 0,024 0,023

yes yes yes yes yes yes no no yes

SIZE CATEGORIES

Small

Medium-sized

Large The Big 3

medians significantly different

obs. 454 411 405 368 406 406

med 0,040 0,053 0,040 0,048 0,047 0,039

obs. 482 431 425 403 436 433

med 0,030 0,039 0,031 0,036 0,032 0,029

obs. 484 433 429 407 435 435

med 0,016 0,024 0,024 0,029 0,027 0,024

obs. 3 2 3 3 3 3

med 0,006 0,006 0,008 0,021 0,015 0,014

yes yes yes yes yes yes

406 246 286

0,036 0,032 0,032

410 263 327

0,028 0,023 0,024

410 249 326

0,023 0,021 0,019

3 3 3

0,017 0,018 0,014

yes yes yes

DEPOSIT INSURANCE SCHEME

Included in DIS

Not included in DIS medians significantly different

obs. 777 778 785 733 797 799 802 587 694

med 0,033 0,040 0,035 0,038 0,033 0,029 0,026 0,024 0,022

obs. 349 347 356 355 403 418 424 173 217

med 0,023 0,028 0,024 0,033 0,031 0,029 0,031 0,028 0,029

yes yes yes yes yes no yes yes yes

Note: In order to utilize all the the first quarter of each year.

available data, all the indicators are calculated at the end of

Table A.8.

Variable description

VARIABLE DESCRIPTION

Size Capitalization Loans to assets Nonperforming loans Loan loss provisions Liquidity ratio Loans to individuals Net interest margin Credit growth Oil price GDP growth total assets, mln.RUB ratio of equity to total assets ratio of total loans (to nonfinancial clients) to total assets ratio of nonperforming loans to total loans ratio of loan loss provisions to total loans ratio of liquid assets to total assets ratio of loans to individuals to total loans the difference between interest income from loans to customers and interest expense paid on customer deposits as a proportion of total loans annual change in loans to nonfinancial clients average export price for crude oil for preceding quarter ($ per ton), Rosstat quarterly growth of real GDP, Rosstat

DUMMY VARIABLES

Foreign bank State-controlled bank Moscow bank Big 3 Deposit insurance system 100% foreign owned bank as reported quarterly by the CBR bank included in the list of state banks by Vernikov (2007) bank's headquarters are located in Moscow three largest banks by assets: Sberbank, VTB and Gazprombank bank entered DIS before the end of the first quarter of 2005

REGULATION RATIOS

N1 - capital adequacy ratio N2 - quick liquidity ratio N3 - current liquidity ratio N7 - maximum large credit risk bank's equity capital to the overall risk-weighted assets minus the sum of the reserves created for the depreciation of securities and possible losses sum of the bank's highly liquid assets to the sum of the bank's liabilities on demand accounts sum of the bank's liquid assets to the sum of the bank's liabilities on demand account and accounts up to 30 days percentage of the total amount of large credit risks (which is the sum of the bank's risk-weighted claims to one borrower) in the bank's equity capital