The effect of ownership on the determinants of profitability in the Russian banking sector

Table of Contents

1. Introduction

2. Literature Review
2.1 Performance
2.2 Corporate Governance
2.3 Ownership

3. Data Sources and Variables
3.1 Data Sources and Ownership Classification
3.2 Dependent Variables
3.3 Explanatory Variables
3.3.1 Numeric Explanatory Variables
3.3.1 Dummy Explanatory Variables
3.4 Other Control Variables
3.5 Descriptive Statistics

4. Model Specification
4.1 The model
4.2 Econometric Methodology

5. Empirical Results
5.1 Summary Statistics and Correlations
5.2 Regression Results and Discussion
5.3 Robustness Checks

6. Conclusions

7. Limitations and Directions for Further Research

References

Appendices
Appendix A: Notes on the Empirical Research
Appendix B: Summary Statistics and Regression Results
Appendix C: List of Banks

Abstract

This paper examines variations in the return on assets (ROA), return on equity (ROE), net interest margin (NIM), and operating cost ratio (OCR) determinants across ownership structures in 350 Russian banking institutions over the years 2002 to 2012. It is innovative in three ways: firstly, it brings together financial and corporate governance indicators as explanatory variables; secondly, it uses an extensive number of dependent and explanatory variables compared with previous studies on Russia; thirdly, it classifies the ownership of banking institutions using the definition of majority ownership at 20% level. The findings suggest that there are some differences in the determinants of bank ’ s profitability by ownership type: personnel cost ratio and bank portfolio composition ratio dominate in state-influenced banks; capitalisation, reserves and default risk ratios dominate in domestic private-owned banks; and capitalisation ratios dominate in foreign-owned banks. Operating efficiency and credit risk are important factors influencing performance across all ownership types.

Furthermore, control variables such as inflation, GDP growth, the size of the banking sector, and the bank asset share ratio are significant, suggesting that they play an important role in the profitability of a bank. Corporate governance variables such as Bureau van Dijk (BVD) independence ranking and listing on an exchange are found to be insignificant. The location of a bank (central or regional) is an important factor influencing its operating cost ratio. The conclusion is that bank ownership plays an important role in Russia and should not be ignored when analysing performance determinants.

Tables and Figures

Table 1: Summary statistics

Table 2: BVD statistics

Table 3: Location statistics

Table 4: Listed on exchange statistics

Table 5: Correlation matrix

Table 6: Main regressions

Table 7: Fixed-Effects regressions

Table 8: Averaged regressions

Table 9: Lagged regressions

Figure 1: Classification of state-related banks as suggested by Vernikov

Figure 2: The evolution of banks by ownership type in the sample of 350 banks

Figure 3: Bank profitability (ROA and ROE are shown in percentages) over the years investigated

Figure 4: Bank profitability (NIM and OCR are shown in percentages) over the years investigated

1. Introduction

“Russia’s banks have long been the Achilles’ heel of the country’s economy”, wrote a journalist for The Wall Street Journal in an article on Central Bank of Russia (CBR) in March 2013. The author of the article also suggested that there are three main points Russia should improve: (1) get the state out of banking, (2) keep banks out of geopolitics, and (3) enforce existing banking regulations. The importance that the banking sector has on the economy and development of a country is decisive¹, therefore investigating this sector is crucial for Russia’s future policy developments. It is true that state-influenced banks dominate Russian banking, and that sometimes they act as de facto branches of the foreign ministry. Nevertheless, despite the historical prevalence of state ownership in firms and banks in Russia, some studies² have found that the government has made an effort to create market competition and improve corporate governance measures. Firms are keen to see the political control be reduced to a minimum so that they can attract foreign direct investment and effectively compete internationally. A recent example of ‘depoliticization’³ is Putin’s move to liberate the oil and gas sector from Gazprom’s monopolistic power. The introduction of competition to Russia’s oil and gas sector means that Gazprom will no longer be the only giant to control market prices in Russia as well as export prices internationally. Rosneft and other petroleum companies have already signed contracts with European and Chinese governments on oil and gas exports⁴. However, the same ‘depoliticization’ does not seem to apply to the banking sector. Indeed, others studies⁵ have found that the market share of state-controlled banks has been growing since the 1998 and 2004 crises, and has reached 54% of all banking assets in 2011, with the top five banks⁶ accounting for 49.6% of the banking total assets (Vernikov, 2011). Furthermore, in Russia ownership boundaries are blurred, and corporate pyramidal control is widely spread. The importance that the banking sector has on an economy and development of a country is decisive as pointed out by La Porta, therefore investigating this sector is crucial for Russia’s future policy developments.

Capital reserves are a bank’s first line of defence against unexpected losses, and profits are a crucial source of capital. Traditional financial performance measures are return on assets (ROA), return on equity (ROE) and net interest margin (NIM). In addition, I use the operating cost ratio (OCR) as a measure of cost efficiency.

I identify a lack of researches investigating the determinants of profitability in the Russian banking sector by ownership type and in aggregation, and I am interested in attempting to fill this gap. This study mainly builds on the work of Fungacova and Poghosyan (2009) who have investigated the determinants of bank interest margins in Russia. I extend their research by using three additional performance variables: ROA, ROE, and OCR.

In my research I attempt to answer the following questions:

- Do the determinants of profitability in Russian banks vary by ownership type?
- Does corporate governance have a significant impact on the performance of banks?
- Does the location of a bank matter for its performance?

This study is innovative in three ways:

- It cumulates financial and corporate governance indicators as explanatory variables;
- It uses an extensive number of dependent and explanatory variables compared with previous studies on Russia;
- It classifies the ownership of a bank using the definition of majority ownership at 20% level⁷ ;

In this paper the banking sample of 350 top banks by total assets (as in 2012) is subdivided into three groups by ownership type: state-influenced, private domestic-owned and foreign-owned banks in order to examine variations in the impact of return on assets (ROA), return on equity (ROE), net interest margin (NIM) and operating cost ratio (OCR) determinants across ownership structures. My findings suggest that there are some differences in the determinants of bank profitability by ownership type: personnel cost and bank portfolio composition ratios dominate in state-influenced banks; capitalisation, reserves and default risk ratios in domestic private-owned banks; capitalisation ratios in foreign-owned banks. Operating efficiency and credit risk ratios are important factors across all ownership types. Furthermore, control variables such as inflation, GDP growth, the size of the banking sector, and the bank’s asset share ratio are significant suggesting that they play an important role in the profitability of a bank. Corporate governance variables such as Bureau Van Dijk (BVD) independence ranking and listing on an exchange are found to be insignificant. The location dummy reduces the operating cost ratio of a bank if this bank is not centrally located (outside Moscow or St Petersburg regions). My conclusion is in line with Fungacova and Poghosyan’s findings (2009): bank ownership needs to be considered when analysing the performance determinants in Russia. This paper is structured as follows: in section two I review the relevant literature on performance, corporate governance and ownership; in section three I describe the data sources and the dataset, as well as all dependent, independent and control variables; in section four I outline the model specifications and the econometric methodology used; in section five I discuss my findings and the robustness checks performed; section six is the conclusion, and finally in section seven I outline the main limitations of this paper as well as give directions for future research.

2. Literature Review

2.1 Performance

The performance of banks has been explored in many cross-country studies. In particular, Demirguc-Kunt and Huizinga (1999) used bank-level data for 80 countries over the 1988-1995 period, and found that in developing countries foreign banks have higher profit margins than domestic banks, while the opposite was true for the developed countries. Similarly, Micco et al (2004), in assessing the relationship between bank performance and ownership in a sample of 119 countries over the years 1995 to 2002, found that ownership and performance are strongly correlated for the developing countries in particular, and that state-owned banks in developing countries tend to have lower profitability and higher costs than their foreign peers.

Claeys and Vennet (2004) investigated the determinants of bank interest margins in Central and Eastern European countries. They found that concentration, operational efficiency, capital adequacy and risk behaviour are important determinants of margins in banks. Also, Cornett et al (2009) examined how state ownership and its involvement in a country’s banking system affects bank performance between 1989 and 2004. They found that state-owned banks were less profitable, held less core capital, and had greater credit risk than their privately owned peers. In addition, performance differences were more significant in countries with greater government involvement and corruption in the banking sector.

In country-specific studies on the performance of banks, Athanasoglou et al (2004) examined the effects of bank-specific, industry-specific and macroeconomic determinants of bank profitability in a sample of Greek banks. The authors found that all bank-specific determinants⁸ - with the exception of the size - affect bank profitability significantly in the anticipated way. A year later, Choi and Hasan (2005) investigated the effect of ownership and governance on the performance of commercial banks in Korea. In particular, the authors studied the effects of foreign investors, and the presence of outside directors in the corporate board structure. The evidence indicated that the extent of foreign ownership has a positive effect on the performance, while outside directors do not seem to have any significant impact on the performance of a bank.

Regarding the performance studies on Russian banking system in particular, there is a limited number of them.

Love and Rachinsky (2007) investigated the ownership, corporate governance and operating performance of banks in Russia and Ukraine, using survey data provided by the International Finance Corporation (IFC) in 2003 and 2006. The authors found that there is an economically large but insignificant relationship between governance and performance, and concluded that corporate governance has a second-order effect on the operating performance of banks in Russia and Ukraine. They also found that banks with more concentrated ownership have lower rankings on corporate governance.

A few years later, Fungacova and Poghosyan (2009) analysed the interest margin determinants in the Russian banking sector with a particular emphasis on the ownership structure. Their results suggest that bank ownership needs to be considered when analysing interest margin determinants. Indeed, the authors found that most of the traditional determinants of bank interest margins differ in their impact when considering ownership categories of banks: the impact of credit risk on net interest margins is only significant for domestic private banks; the estimated coefficient for the size of operations is significant for private domestic and foreign-owned banks; the impact of the market concentration is insignificant. The authors also showed that there are certain similarities across banks with different ownership structures: there is a large significant impact of operational costs and risk aversion across all ownership subgroups.

Interestingly, in a research conducted a year later, Karas et al (2010) found that Russian state-owned banks are not less efficient than private domestic banks. The authors concluded that the Russian banking system might benefit from increased levels of competition and greater involvement of foreign banks. On the opposite, Mahlberg et al (2011) analysed the development of the Russian banking sector between 2001 and 2007, and assessed the impact of ownership on the operations of commercial banks. Using the stochastic frontier approach to determine cost efficiency differences, the authors found that foreign-owned banks are more efficient than state-owned banks, and that state-owned banks have higher efficiency levels than private domestic banks. Private domestic banks are less efficient mainly due to the higher interest rates they have to pay for their deposits.

To the best of my knowledge, the only study that deals with corporate governance indicators (other than ownership) and banks’ valuation in the Russian banking sector was conducted by Vernikov and Bokov in 2008. A sample of 10 acquisitions and IPOs has been taken over the five years preceding the study, and authors used price-to-book value of equity as the dependent variable. They employed a series of proxy variables for the quality of corporate governance as independent variables. They found that asset size, shareholder concentration, quality of auditors, stability of the management board and the strategic nature of the transaction dummy variable have a significant impact on investors' decisions.

2.2 Corporate Governance

Shleifer and Vishny in “A Survey of Corporate Governance” (1997) wrote one of the most famous surveys on corporate governance up to date. They followed the widely accepted definition as in the UK Cadbury Code: “Corporate governance is the system by which companies are directed and controlled”. The principal question that corporate governance poses is how to assure that financiers get a return on their investments. The agency problem (separation of ownership and control) is central to this issue. The authors provide Russia as an example of poor management and managerial asset stripping⁹. In addition, they refer to the pyramidal structure of Russian banks: “a Western investor can control a Russian company with 75% ownership, whereas a Russian investor can do so with only 25% ownership” (p.755, Shleifer, Vishny, 1997). The Russian management can use a variety of techniques against foreign investors such as declaring some shares illegal, losing voting records etc. These corrupt practices constitute one of the main reasons why Western investors are reluctant to invest in Russian companies. Other reasons cited by the authors are the virtual absence of minority shareholder protection and the incompetence of managers to restructure the companies so that shareholders can benefit.

A few years later, La Porta et al (2002) wrote one of the most extensive studies on governmental ownership of banks around the world, which encompasses a sample of 92 countries. The authors explored the development and the political views¹⁰ on the government’s participation in financial markets, and found evidence that the political view is prevailing. Indeed, their results confirmed that governmental ownership causes an inefficient allocation process, slower financial development, as well as lower efficiency and slower economic development.

In studies on Russia in particular, Boycko et al (1993) investigated the privatisation process in Russia. Their conclusions at that time, which are now twenty years old, still apply and describe fairly well why the Russian banking sector did not develop into a well-oiled financial intermediation machine. They wrote that potential investors fear that firms controlled by politicians will not maximize profits, and even if value is created, shareholders will not capture it. They came to the conclusion that privatisation in Russia does not necessarily mean ‘depoliticization’ of institutions. Recent studies on Russian banks in particular show the tight grip that the state still holds over many institutions, primarily through indirect ownership and pyramidal schemes. McGee and Preobragenskaya (2004) described the efforts Russian corporations have made since the 1998 Russian banking crisis. Indeed, after the Russian default, organisations have been formed with the aim of protecting minority shareholders and promoting good corporate governance practices. Their study includes an extensive literature review on corporate governance in Russia, along with an analysis of practices and changes incurred. Their methodology consisted of interviewing managers from the six¹¹ biggest auditing companies in Russia. They found that Russian companies have improved their corporate governance practices and are on the right track, but that issues of transparency are still prevalent, which they explain by cultural factors¹². Similarly, McCarthy and Puffer (2003) emphasised the cultural importance in Russia and constructed a framework for corporate governance analysis for Russian companies specifically, building on “a cultural embeddedness model as well as agency and stakeholder theories” (p. 397, McCarthy, Puffer, 2003).

Judge et al (2003) used survey data and found that effective corporate governance is essential for firm performance in Russia. In particular, they found a negative relationship between CEO duality¹³ and firm performance, as well as between the proportion of inside directors and firm performance. A recent survey (2007) on corporate governance in the Russian banking sector specifically has been conducted by the IFC¹⁴. It revealed that the Russian banking sector has adopted a good corporate governance framework, which they can develop. By investigating the five key areas of corporate governance (internal control and risk management, commitment, board practices, transparency, and shareholder rights) the authors concluded that even though corporate governance in Russian banks is acceptable, it is far from being perfect.

2.3 Ownership

Vernikov is the best-known contributor of corporate governance research in the Russian Banking sector. He is especially known to have concentrated his investigations on one particular aspect of corporate governance - the ownership structure. In his papers¹⁵ the author uses the banking industry case to show that public ownership in Russia is not easy to discern. Vernikov (2010) in his research found that the government controls directly only four institutions, but if pyramidal structures and indirect ownership are investigated, the author found that federal and regional authorities control 50 financial institutions. Vernikov classified the degree of control of public authorities on banks into three broad categories - state-influenced, state-controlled and state-owned banks.

He explored these categories in more detail in his work with Glushkova - “Size of Public Sector in the Russian Banking Industry” (2010), where both authors examined the channels of state participation. They found that ownership, governance and other forms of control constitute the three main channels of governmental participation.

They firstly explored the ownership factor and found that state-ownership can be separated into direct and indirect participation. Directly state-owned banks “consist of banks with federal, regional government, local administration, the Central Bank of Russia, and federal or regional property funds acting as stockholders” (p.339, Vernikov, Glushkova, 2010). Indirectly state-owned banks are those “whose capital is owned either in full or partly by non-financial enterprises, non-bank financial institutions or by banks that are in turn state- owned” (p.339, Vernikov, Glushkova, 2010).

Secondly, the authors found that banks can be influenced through governance directly - through representatives of legislative and executive bodies on board, and indirectly - banks with executives of state-owned or state-governed firms acting as members of the board.

Finally, there are other channels of control than governance or ownership - these are banks that are ‘politically connected’ - where “top officials or politicians act as shareholders or members of the board” (p.340, Vernikov, Glushkova, 2010). In Russia, for example, the Central Bank of Russia controls placements of certain banks.

An interesting note is the following: while Vernikov in his works uses 50% of equity as majority ownership threshold, La Porta et al. (2002) defined a bank as state-owned if the government holds at least 20% of equity and acts as the single largest majority shareholder.

The broad classification of banks with state participation as outlined above, can be rearranged into three categories - state-owned, state-controlled and state- influenced banks. State-owned banks constitute the narrowest category and include banks that are majority state-owned in terms of equity¹⁶. The state- controlled banks category, in addition to state-owned banks, also includes banks with minority governmental equity participation, but with an identified direct or indirect state governance participation, in the absence of majority equity ownership. State-influenced banks include both state-owned and state-controlled banks, as well as banks with an identified “political connection”.

illustration not visible in this excerpt

Figure 1: Classification of state-related banks as suggested by Vernikov

3. Data Sources and Variables

3.1 Data Sources and Ownership Classification

The data is gathered from the Bankscope, CBR and World Bank databases. I take a sample of the top 350 banks¹⁷ from 2002 to 2012. The sample is an unbalanced panel as there are banks entering and leaving during the eleven-year period investigated, in addition to numerous missing data points. The data is missing randomly¹⁸. I generate ownership dummies and create sub-samples so I can run separate regressions on each ownership type¹⁹.

In this research, I follow the definition proposed by La Porta et al. (2002) and classify a bank as state-owned if the government holds at least 20% of equity and acts as the single largest major shareholder. For state-controlled and state- influenced banks, Vernikov’s definitions are followed as described in part 2.3 of this paper, with the exception that majority stake is defined at 20% or more. All the banks identified as belonging to any of these three categories are put together as state-influenced. Similarly, I classify a bank as being foreign owned where the foreign ownership is at least 20% of equity and where the foreign owner is the single largest major shareholder²⁰. Finally, the rest of the sample constitutes private domestic banks where neither a major foreign shareholder nor state influence have been detected.

3.2 Dependent Variables

- Return on Assets (ROA) is a measure of efficiency, which summarizes the ability of the management to produce net earnings from the assets of the bank. It is calculated as the net profit of a bank as a percentage of total assets. It is by far the most popular measure of performance as it assesses directly the financial return of a shareholder’s investment, and also includes assets financed by borrowing. So it is of interest to both shareholders and lenders. In addition, ROA is usually available from public sources and it allows the comparison between organisations of different sizes from the same industry.
- Return on Equity (ROE) is another common measure of profitability, calculated as net income as a percentage of total equity. ROE is considered to be the most important ratio for investors as it measures the return on the money investors have put into the company and the retained profit. Nevertheless, ROE doesn’t account for the greater risks associated with leverage. This risk is captured by the ROA ratio.
- Net Interest Margin (NIM) is a common measure of the efficiency of portfolio management of banks, and is calculated as the ratio of net interest income (difference between interest earned and expenses associates with liabilities) to total assets. It is a commonly used proxy for the income generation capacity of the intermediation functions of banks. - Operating Expenses/Total Assets (OCR) is the operating cost ratio. It indicates the total amount of the expenses needed to manage the assets of the bank²¹. When this ratio is high compared with peer banks, investors usually see it as a red flag.

3.3 Explanatory Variables

3.3.1 Numeric Explanatory Variables

- Equity/Lag of Total Assets (CAP1) is a capitalisation measure where the total assets are lagged by one period²².
- Equity/Assets (CAP2) is the widely used measure of capitalisation, also called leverage. It captures the quality of a bank’s management and risk preferences. The higher the ratio, the higher is a bank’s risk aversion. - Total Loans/Total Earning Assets (CR1) is a credit risk ratio that captures the potential variation of powers between banks²³. The ratio provides a general measure of the financial position of a bank, including its ability to meet financial requirements for outstanding loans. - Non-performing Loans/ Total Loans (CR2) measures the credit risk faced by a bank²⁴. Bank non-performing loans to total gross loans are the value of non-performing loans²⁵ divided by the total value of the loan portfolio (including non-performing loans before the deduction of specific loan-loss provisions). The loan amount recorded as nonperforming should be the gross value of the loan as recorded on the balance sheet, not just the amount that is overdue.
- Non-Interest Revenues/Total Assets (IM) is a measure of revenue other
than interest from loans, calculated as non-interest revenue divided by total assets. This figure tends to be higher for banks that derive their income from commission related services provided to its customers, therefore differentiating large retail commercial banks from investment banking institutions²⁶.
- Personnel Expenses/Total Loans (OPEF) is a measure of a bank’s operating efficiency²⁷.
- Overhead Costs/Total Assets (OHTA) is a measure of efficiency and of the way the management uses the funds. The ratio is calculated as overhead costs divided by total assets²⁸.
- Total Deposits/Total Assets (PF) is a measure of a bank’s funding, and is calculated as total deposits to total assets²⁹.
- Total Loans/Total Deposits (LTD) is a measure of a bank’s exposure, calculated by dividing the bank's total loans by its deposits³⁰. - Cash and Liquid Assets/Total Liabilities (CAR) is called the cash asset ratio, and is a measure of liquidity (how easily a bank can service debt and cover short-term liabilities if the need arises). Potential creditors use this ratio in determining whether to make short-term loans. - Loan Loss Provisions/Total Loans (DF) is a measure of default risk, calculated as loan loss provisions (an expense set aside as an allowance for bad loans such as customer defaults) to loans³¹.

3.3.1 Dummy Explanatory Variables

- BvD Independence Index (BVD) is the independence indicator constructed by Bureau van Dijk and available from the Bankscope website. This indicator categorizes the degree of independence of a bank. The indicator ranks are as following:

A+, A, A- means there is no recorded shareholder with more than 25% direct or total ownership;

B+, B, B- means there is no recorded shareholder with more than 50% direct or total ownership and one or more shareholders are recorded with more than 25% direct or total ownership;

C+, C is when there is one recorded shareholder with more than 50% total ownership;

D is when there is one recorded shareholder with more than 50% of direct ownership;

U is when the status of independence is unknown³²

- Location (CITY) is a dummy variable taking the value of 0 if the bank is located in Moscow or St Petersburg, and the value of 1 otherwise.

- Quotation dummy (QUOT) is a dummy variable taking the value of 0 if a bank is listed on a stock exchange, and the value of 1 otherwise³³.

3.4 Other Control Variables

A set of control variables is used to isolate the effects of bank characteristics on performance from other factors influencing profitability. - Inflation rate (INF) is a common macroeconomic control variable used in studies such as Bashir (2001), Micco et al (2004), Claeys, Vennet (2004), Athanasoglou (2004).

- GDP growth rate (GDPGR) is a common macroeconomic control variable used in studies such as Micco et al (2004), Claeys, Vennet (2004). It is a proxy for the change in economic conditions. I use nominal GDP here, as per past studies.
- Total assets (LOG_LAGTA) is the log of lagged total assets, and is used as the size measure in this research, following Naceur, Goaied (2001), Berger et al (2004), Micco et al (2004) studies.
- Bank asset share (LOG_LAGSHARE) - log of lagged values of the share of a bank’s assets over total banking assets in the country, is used following Micco et al (2004) study.

3.5 Descriptive Statistics

Figure two below shows the evolution of banks by ownership type in the sample of 350 banks used here. It can be seen that there is an exponential increase in the number of state-influenced banks starting from 2011, which means that the state might be making a come-back, as pointed out by Vernikov (2009). In addition, the number of foreign-owned banks has been growing steadily since 2006. The number of private banks has been growing until 2011, but has reduced significantly between 2011 and 2012 years. This might mean that their ownership has changed, and because it can be seen that the number of state-influenced and foreign banks is growing, I suspect that domestic banks have become either state- influenced or foreign. Finally, the number of banks with unknown ownership has decayed exponentially since 2010, meaning that Russian banks are becoming more transparent and their ownership structure is becoming publically available. An alternative explanation to the graph below might be that a large number of banks have been discovered to be state-owned, as most of the time the foreign and private banks are open about their ownership structure. Therefore, this might mean that there is a trend of improved transparency rather than a major trend towards state ownership. It can be seen that the number of domestic and foreign banks has gone down, which might mean that domestic banks have gone bankrupt and that foreign banks have decided to get out of Russia.

illustration not visible in this excerpt

Figure 2: The evolution of banks by ownership type in the sample of 350 banks

Figures three and four below show how ROA, ROE, NIM and OCR evolve over time for the top 350 banking institutions in Russia over 2002-2012 years. As expected, ROA and ROE have evolved simultaneously over the time span investigated, with a peak in 2007. The financial crisis years can be seen for both indicators with lowest values in 2009, and slow recovery starting at the end of 2009, beginning of 2010. Interestingly, NIM does not show any sign of crisis, and presents a slowly falling pattern. NIM is mainly affected by the level of competition and improved efficiency. It is steadily going down, which means that some inefficient banks have closed down because they could not compete. On the opposite, OCR peaks during the crisis because the volume of lending had dropped sharply, but banks couldn't reduce their costs immediately, partly because they had lost some economies of scale.

illustration not visible in this excerpt

Figure 3: Bank profitability (ROA and ROE are shown in percentages) over the years investigated

illustration not visible in this excerpt

Figure 4: Bank profitability (NIM and OCR are shown in percentages) over the years investigated

4. Model Specification

4.1 The model

The general model to be estimated is of the following linear form³⁴:

illustration not visible in this excerpt

where:

[illustration not visible in this excerpt]is the profitability of a bank i at time t with i= 1,..N and t= 1,..T

[illustration not visible in this excerpt]takes the value of the four dependent variables accordingly - ROA, ROE, NIM and OCR

[illustration not visible in this excerpt]is the sum of the explanatory variables X, and β is the coefficient of each explanatory variable X - c is a constant term

- The error term is of the following form:

illustration not visible in this excerpt

where v i is the unobserved bank-specific effect, and u it is the idiosyncratic error

The explanatory variables X it are grouped together in the equation (1). They can be divided into numerical, dummy and control variables, as described in section three. Therefore, the equation can be re-arranged as follows:

illustration not visible in this excerpt³⁵

where the X it s with superscripts j, l and m denote numerical, dummy and control variables respectively.

4.2 Econometric Methodology

I use an unbalanced panel of Russian banking institutions in the period 2002- 2012. The model (3) is the basis of the estimation. In statistic relationships the literature usually applies the least squares methods on Fixed (FE) and Random Effects (RE) models (Athanasoglou et al, 2004).

The econometric analysis of the model (3) poses the following issues: first, the individual effects are examined to see if they are fixed or random; second, it is possible that due to the time frame used (during which changes due to post-1998 and 2008 crises occurred) time effects are present in the error component of the model; third, there can be stationarity issues in the panel; and finally, some potential endogeneity issues might arise.

To address the FE versus RE issue, I run both models. Given that there are three dummy variables that are nearly invariant in time, I run RE models with dummies and FE models excluding dummy variables. I use the Hausman test to see if the difference in coefficients between FE (without dummies) and RE is systematic or not, and if the FE model is appropriate to run (again, excluding dummies). The null hypothesis of the Hausman test is that the preferred model is RE versus the alternative FE model. It tests whether the unique errors (ui) are correlated with the regressors, and the null hypothesis is they are not. It is

important to acknowledge the different nature of the estimator: while FE controls for unobserved banks heterogeneity, RE does not³⁶.

Furthermore, it is possible that given the instabilities and developments that took place during the years 2002 to 2012, time effects are present in the error component of the FE model. The error term would then be:

illustration not visible in this excerpt

where λ t is the unobserved time effect. The relevant test for time effects³⁷ is run for each FE model to see if time fixed effects are needed.

Finally, in order to check for stationarity³⁸ and endogeneity problems³⁹, I run robustness checks. To address the stationarity issue I run model (3) with averaged values over three-year non-overlapping periods. To control for possible endogeneity problems I use lagged explanatory variables as a robustness check⁴⁰. The results are presented in section five below.

5. Empirical Results

5.1 Summary Statistics and Correlations

The summary statistics for the aggregate sample of all banks, and for each ownership type separately are presented in Table 1 in Appendix B. I analyse the means for each variable for the aggregate sample and by ownership type below.

As expected, the ROA and ROE levels are highest in foreign-owned banks (1.91% and 12.50% respectively), followed by private domestic banks for the ROA (1.61%) and by state-owned banks for the ROE (11.39%). As demonstrated in past studies outlined in section two here, in general foreign-owned banks were found to be more efficient, so produced higher net earnings from their assets. Foreign-owned banks also achieve higher returns on the money investors have put into the bank.

The NIM is highest for the private domestic banks (6.59%) and lowest for the state-influenced banks (5.23%). This means that private-owned banks, both domestic and foreign, are better at managing their portfolios. There are several possible reasons for this, such as lower funding costs, different loan portfolios (for example, loans to businesses or individuals or secured/non-secured loans), and better salesmanship.

The OCR is lowest for the state-influenced banks (6.86%), and is the double for the private and foreign banks (14.46% and 14.07% respectively). This shows that state-influenced banks incur lower operating costs than their domestic private and foreign-owned peers, probably due to lower funding costs. This is unexpected, because it means that it is particularly hard for private and domestic banks to survive given such conditions.

The capitalisation ratio, with lagged total assets as denominator, is lowest for foreign owned banks (0.01) and highest for state-influenced banks (0.06). Capitalisation with total assets as denominator is distributed more uniformly amongst bank ownership categories: 0.17 for state-influenced, 0.16 for private domestic and 0.15 for foreign, which shows that all three ownership categories have similar quality of bank management and risk preferences. OPEF is lowest for private domestic banks (0.06%), and highest for foreign banks (0.75%). This means that private domestic banks are better at managing their operating efficiency when compared with their peers.

OHTA is lowest for the state-influenced banks (0.09%) and nearly double for the private and foreign banks (0.17%). This means that state-influenced banks are better at achieving the lowest operating expenses possible without sacrificing their good services or competitiveness.

LTD ratio is highest for state-influenced banks (7.83), followed by foreign- owned (1.85) and private (0.93). LTD ratio also depends on where a bank’s funding is coming from: (1) the money markets (which is short-term funding, and in this case banks may well have a problem given these LTD ratios), or (2) the bond markets or the state equity (long-term funding, and in this case banks should do fine with such LTD ratios).

There is little difference between the banks by ownership type for the rest of the variables such as CR1, CR2, IM, PF, CAR and DF.

For the corporate governance variables, in the sample used here 38.56% are classified as D, followed by 23.14% classified as A+, 20% classified as unknown and 13.43% classified as B+ (Table 2 in Appendix B). This means that most of the banks in the sample have one recorded shareholder with more than 50% of direct ownership, followed by no recorded shareholder with more than 25% direct or total ownership. It is to be noted that in the Bankscope database the banks investigated do not change their BVD ranking during the years observed, which suggests that there might be a need for a more precise and sensitive to evolutions type of corporate governance ranking.

In addition, 34% of all banks are situated outside Moscow or St Petersburg regions, which shows that the banking sample chosen here has a good coverage of non-central regions of Russia as well (Table 3 in Appendix B). It is important to emphasise that only 13.14% of banks in the sample are listed on an exchange, showing that in Russia capital markets are not important for a bank to grow in terms of assets (Table 4 in Appendix B).

The correlation matrix as presented in Table 5 in Appendix B shows two high correlations that could pose multicollinearity problems: LOG_OHTA and OCR (0.76) and DF and LOG_CR2 (0.73). I exclude LOG_OHTA in models three, 3.1, 3.2, and 3.3 to avoid multicollinearity issues. In what concerns DF and LOG_CR2, which are both independent variables, I exclude LOG_CR2 from all the regression models. LOG_CR2 is a credit risk ratio calculated as non- performing loans to total loans, but there exists already the CR1 variable that is a proxy for the credit risk too (calculated as loans to total earning assets). DF is the default risk calculated as loan loss reserve to gross loans. Therefore, I decide to exclude LOG_CR2 to avoid multicollinearity problems.

It is interesting to see that there are low correlations between the dependent variables. ROA and ROE are correlated at 0.41, while the rest of the dependent variables are correlated at less than 0.20. There are also some very low negative correlations between the dependent variables: ROA and OCR (-0.05) and ROE and OCR (-0.02). This is intuitive given that usually ROA and ROE ratios move in the opposite direction with OCR.

5.2 Regression Results and Discussion

For the entire banking sample I run RE regressions with corporate governance dummy variables, and FE regressions without dummy variables. As reported in Tables 6 and 7 in Appendix B, there are some minor differences in significance and the scale of the impact of various explanatory variables on the dependent variables. Nevertheless, most of the results from FE and RE regressions are in line with each other.

The FE model controls for all time-invariant differences between the banks, so the estimated coefficients of the FE models cannot be biased because of omitted time-invariant characteristics. The fact that I obtain results which are in line with each other for the numerical explanatory variables for FE and RE regressions is reassuring given that I cannot run a FE regression with time-invariant dummy variables.

For state-influenced banks the Breusch-Pagan Lagrange Multiplier (LM) test is insignificant at 5% level for all models, therefore I run a simple OLS regression as there is no evidence of significant difference across state-influenced banks. It is interesting to see that for state-influenced banks there is no significant difference across banks for all the profitability indicators. This means that state-influenced banks are homogeneous amongst themselves.

Domestic private owned banks and foreign banks are homogeneous only in terms of ROA and ROE performance indictors, which signifies that there is less homogeneity compared with their state-influenced peers.

For foreign-owned banks the LM test is insignificant at 5% level for models 1.3 (ROA) and 2.3 (ROE), on which I run simple OLS regressions. For NIM and OCR I run RE regressions, as the LM test is significant at 1% level. For private-owned domestic banks the LM test is insignificant at 5% level for models 2.2 (ROE), on which I run a simple OLS regression. For ROA, NIM and OCR I run RE regressions, as the LM test is significant at 1% level.

For the aggregate banking sample, I found that capitalisation variables (both CAP1 and CAP2) have a significant positive effect on the dependent variables. Indeed, an increase of CAP1 by 1% leads to an increase of 0.004 units of ROA, 0.0324 units of ROE, and 0.0002 units of OCR. CAP2 has a significant positive effect on ROA (0.06) and NIM (0.07). For state-influenced banks, there is only one significant positive relationship between CAP2 and OCR (0.01). For private domestic banks I found that capitalisation proxies have a positive effect on dependent variables, and that there is a significant positive effect of CAP2 on ROA (0.05) and on NIM (0.07). Similarly, for foreign-owned banks I found that there is a significant positive effect of CAP1 on ROA (an increase of CAP1 by 1% leads to an increase of ROA by 0.0128 units) and on ROE (an increase of CAP1 by 1% leads to an increase of 0.1704 units in ROE), as well as of CAP2 on ROA (0.10).

The positive impact of capitalisation variables on the performance of a bank is as expected. Indeed, capitalisation refers to the amount of funds available to maintain a bank’s business, and is a safety net used in crisis periods. In addition, during the years investigated there has a been a lot of M&A activity going on since the Russian default crisis and during the pre-2009 boom years. This strengthens the assumption that the relation between the capitalisation ratios and the performance ratios should be positive. A bank with a good capital position is able to pursue business opportunities more effectively and has more time and flexibility to deal with problems arising from unexpected losses, therefore achieving higher profits. The positive impact of capitalisation ratios on the operating cost ratio also suggest that higher leverage induces higher expenses on a bank’s accounting statement, in addition to higher funding costs. Credit risk (CR1) has a positive effect on performance variables for the aggregate sample, and its impact is significant for NIM (0.08) and OCR (0.01). For state-influenced bank the credit risk ratio has a significant positive effect on all performance variables (0.03 on ROA, 0.25 on ROE, 0.07 on NIM), except for OCR where the effect is positive but insignificant. For private-owned banks the credit risk ratio has a significant positive effect on ROA (0.02) and NIM (0.13). For foreign owned banks the credit risk ratio has a significant positive effect on the performance proxies (0.19 on ROE, 0.13 on NIM and 0.01 on OCR). It is considered that an increased exposure to credit risk is associated with decreased profitability. The higher the credit risk ratio, the lower the bank’s liquidity, therefore the bank is more exposed to defaults. Banks normally would improve performance and profitability by improving monitoring and credit risk management. The fact that this rule does not seem to apply to the Russian banking sector might reflect the dominance of state-influenced banks in the country’s banking industry. Indeed, the state guarantees and directs the loans issued, so the bank does not control the provisions held for loan losses. The considerable asset growth of Russian banks (which can be seen in the last CBR report) is composed 70% of loans, meaning that loans play a crucial role in the performance of a bank.

The loans issued in Russia have grown by 8.4% since 2011⁴¹ - this data might help to explain the positive effect in the foreign-owned banks⁴². The reserves ratio (IM) produces mixed results. It has a significant positive effect in model three (an increase by 1% in IM increases NIM by 0.0064 units), and a significant negative effect in model four (an increase by 1% in IM decreases OCR by 0.0003 units) in the aggregate sample. This means that the reserves ratio impacts positively the net interest margin, but negatively the operating cost ratio. For state-influenced and foreign-owned banks the effects are insignificant across all dependent variables. For private domestic banks IM has a positive significant effect on NIM (an increase by 1% in IM increases NIM by 0.0094 units). Reserves are calculated as non-earning assets to total assets, therefore the high cost of holding large reserves is usually compensated by a higher NIM. Theoretically, there should be no significant effect of IM on the operating cost ratio, because non-earning assets do not incur higher operating costs. The slight negative effect of IM on OCR (an increase by 1% in IM decreases OCR by 0.0003 units) for the aggregate model might be explained by data deficiencies. The Operating Efficiency ratio (OPEF) has a positive and significant effect on the performance variables in the aggregate sample: an increase by 1% in OPEF increases ROA by 0.0043 units, increases ROE by 0.0095 units, and increases OCR by 0.0003 units. For state-influenced banks OPEF has a positive significant effect on ROA (an increase by 1% in OPEF increases ROA by 0.0106 units) and ROE (an increase by 1% in OPEF increases ROE by 0.0757 units). For private domestic banks OPEF has a positive significant effect on ROA (an increase by 1% in OPEF increases ROA by 0.0047 units), ROE (an increase by 1% in OPEF increases ROE by 0.0344 units) and NIM (an increase by 1% in OPEF increases NIM by 0.0189 units). For foreign banks OPEF has a positive significant effect on all dependent variables: an increase by 1% in OPEF increases ROA by 0.0051 units, increases ROE by 0.0305, increases NIM by 0.0097 units and increases OCR by 0.0004 units.

The ratio of a bank’s personnel expenses to total loans is a measure of the bank’s operating efficiency. The positive impact of OPEF on performance measures can be explained as follows: higher personnel expenses produce a better quality of service and therefore generate more revenues for the bank. Furthermore, efficient expenses management is a prerequisite for improved profitability in the still evolving Russian banking sector. Nevertheless, another way to improve efficiency which is extensively used in developed countries, but less so in emerging ones, is to reduce personnel costs either by improving processes or automation. In addition, the positive impact of OPEF on the operating cost ratio means that higher personnel expenses induce higher operating costs.

OHTA has a positive significant effect on ROA (an increase by 1% in OHTA increases ROA by 0.0051 units) and on NIM (an increase by 1% in OHTA increases NIM by 0.0036 units) for state-influenced banks. The ratio has an insignificant effect across the rest of the sub-samples, and across the aggregate sample.

The overhead ratio is a metric that allows companies to evaluate expenses as a percentage of income. The total cost of a bank (net of interest payments) can be separated into operating cost and other expenses (including taxes, depreciation etc.). Only operating expenses can be viewed as the outcome of the bank’s management. In general, a company strives to achieve the lowest operating expenses level possible, but must balance these cuts with maintaining the quality of service. OHTA has a significant positive effect in state-influenced banks only and this might reflect the fact that state-influenced banks do not pay as much attention to their expenses management compared with their peers, because of the governmental support they receive.

The bank portfolio composition (PF) has a significant negative effect on OCR (-0.20) for the aggregate sample. PF has a significant positive effect on ROA (4.38), ROE (20.53), NIM (4.39) and a significant negative effect on OCR (-0.19) for state-influenced banks. For private banks PF has a significant negative effect on OCR (-0.17), and for foreign-owned banks a significant positive effect on NIM (2.79).

The negative effect on OCR in domestic banks (private and state-influenced) means that higher deposit levels reduce the banks' cost of capital. In addition, deposits are generally seen as a cheap and stable form of funding. The positive effect of PF on ROA, ROE, and NIM means that higher deposit levels produce lower cost of capital for banks, reduce the need for equity and improve the profitability levels.

For foreign-owned banks LTD has a significant positive effect on OCR (an increase by 1% in LTD increases OCR by 0.0003 units), and CAR has a significant positive effect on NIM (an increase by 1% in CAR increases NIM by 0.0066 units). Both variables have insignificant effects across the rest of sub-samples and in the aggregate sample.

LTD is a liquidity measure (loans to customer deposits). High LTD means the bank is highly leveraged and therefore risky, but it all depends what other sources of funding it has available. Issuing loans increases the operating cost ratio of a bank as it usually means that the expenses of a bank are growing. CAR is an alternative measure of liquidity (liquid assets to total liabilities) and has a positive effect on the performance of a bank, as expected. Indeed, high levels of liquidity reduce the cost of debt capital and thus improve NIM.

The default risk proxy (DF) has a significant negative effect on ROE (-0.33) for the aggregate sample. For private banks DF has a significant positive effect on NIM (0.11) and OCR (0.01). The variable has insignificant effects across the rest of sub-samples and the aggregate sample.

This ratio is part of the asset quality ratios of the bank and determines the quality of loans. The higher the ratio, the more problematic are the loans. Indeed, a high DF ratio means that the bank’s borrowers cannot meet their obligations.

The positive effects on NIM in the domestic sample probably reflect high corruption (lending to dubious friend and associates) and the lack of good credit assessment and credit reference agencies. The positive effects on OCR means that higher default rates cause higher operating costs.

For the corporate governance proxies, city dummy (CITY) has a significant and negative effect on the OCR (-0.05) for the aggregate sample, as well as for private-domestic (-0.05) and foreign-owned banks (-0.04).

This can be explained by the fact that when the bank is not centrally located (outside Moscow or St Petersburg) its operating expenses are reduced. This finding is intuitive, as Moscow or St Petersburg are the most financially active areas of Russia, and being located outside these regions can be advantageous in terms of the expenses incurred by a bank - presumably salaries, rents and other expenses are all higher in Moscow and St Petersburg.

For the control variables, inflation (INF) has a significant positive effect on NIM (0.18) and OCR (0.02) for the aggregate sample. INF has a significant positive effect on OCR for state-influenced (0.05), and on NIM (0.18) and OCR (0.02) for private banks. For foreign-owned banks inflation (INF) has a significant positive effect on ROA (0.18), ROE (2.46) and OCR (0.02). This might reflect the fact that foreign-owned banks are better at predicting inflation levels and adjusting their operations accordingly⁴³. In addition, if inflation is high it is easier for a bank to widen its margins. In addition, the positive effect of INF on OCR means that higher levels of inflation induce higher operating costs.

GDPGR has a significant positive effect on ROA (0.07), ROE (0.54), NIM (0.04) and a significant negative effect on OCR (-0.02) for the aggregate sample and for private-domestic sub-sample (0.05 for ROA, 0.54 for ROE, 0.04 for NIM and -0.02 for OCR). GDPGR has a significant positive effect on ROA (0.17) and ROE (1.05), and a significant negative effect on OCR (-0.01) for foreign-owned banks.

The issue of GDP growth is similar to that of inflation, as it is linked to salaries and prices in a country. Usually a higher nominal GDP growth is associated with a booming economy and advantageous financial activities, but might also mean higher consumption levels because of higher prices due to inflation. The negative effect of GDPGR on the OCR variable means that higher levels of GDP growth are beneficial for the operating costs of a bank. LOG_LAGTA has a significant positive effect on ROA and ROE, and a significant negative effect on OCR for the aggregate sample. For foreign-owned banks LOG_LAGTA has a significant positive effect on ROA and ROE. LOG_LAGTA has a significant positive effect on ROE for state-influenced banks.

The LOG_LAGTA is a proxy variable for the size of the banking sector. Its positive effect on ROA (0.29) and ROE (2.56) in the aggregate sample reflects the fact that a growing banking sector has a positive effect on profitability ratios. Similarly, for state-influenced banks LOG_LAGTA has a significant positive effect on ROE (7.44), and for foreign-owned banks on ROA (1.11) and on ROE (14.51). LOG_LAGTA has a negative effect on the OCR variable, suggesting that a bigger size of the banking sector is beneficial, as it reduces operating costs, and induces better economies of scale.

For state-influenced banks LOG_LAGSHARE has a significant negative effect on ROE (-5.76), and a significant negative effect on ROA (-0.98) and ROE (-15.17) for domestic banks. For the aggregate sample LOG_LAGSHARE has a significant negative effect on OCR (-0.02).

This means that banks in Russia face some issues dealing with their growing size, as reflected by the negative influence of LOG_LAGSHARE on ROA, but see their operating costs reduced, as shown by the negative impact on OCR. An alternative interpretation may be that banks are lending a lot of money, giving them economies of scale, but not lending wisely enough, therefore getting some defaults. ROE measures the amount of net income returned as a percentage of shareholders equity. Return on equity measures a corporation's profitability by revealing how much profit a company generates with the money shareholders have invested. The negative impact of LOG_LAGSHARE on ROE means that growth in size does not necessary imply bigger shareholder investments.

It is worth noting that BVD has an insignificant but positive effect on all the depend variables and across all ownership types, except for the OCR (the effect is insignificant, and negative). This means that corporate governance has a positive effect on the performance ratios, even if it is insignificant and small in scale. CITY negatively and significantly influences OCR meaning that regional location decreases operating costs and increases operational efficiency in a bank. Interestingly, being listed on an exchange (QUOT) has a negative effect on all models in the foreign banking sample, which means that domestic banks suffer less in terms of their performance indicators when they are not listed on an exchange. Again, this phenomenon reflects the state dominance in the ownership of Russian banks, as performance doesn’t seem to be related to international markets. In addition, being listed on an exchange gives the bank an increased compliance overhead, but ought to give better access to funding.

5.3 Robustness Checks

I run two regressions for each model as a robustness check for stationarity and endogeneity potential problems: random effects regressions with averaged variables over three year non-overlapping periods (I call them ‘averaged regressions’), and random effects regressions with lagged explanatory variables (I call them ‘lagged regressions’). The results are reported in Tables 8 and 9 in Appendix B.

The averaged regressions account for the missing data. They produce similar results to initial RE regressions, and none of the significant coefficients is contradicted. This is reassuring as these robustness checks confirm the exactness of the RE results which have been discussed earlier in this paper. In what concerns the lagged results, I obtain some contradictory results when compared with initial RE regressions. There is one significant contradiction in model one, one in model two, and one in model three. For models one and two I obtain a significant positive effect of DF on the ROE (0.04 and 0.18 respectively), and for model three I obtain a significant negative effect of OHTA on NIM (an increase of 1% in OHTA decreases NIM by 0.0027 units) The negative effect of OHTA on NIM is as expected, and it can be considered that the lagged model might have produced a more accurate result in the case of model three. The differences in models one and two might be due to poorer fit of the lagged models, which can be seen by lower R-Squared values (4% for model one and 5% for model two). Nevertheless, most of the results are in line with the initial RE regressions, emphasising the credibility of the RE results discussed earlier.

6. Conclusions

In this paper I have investigated the bank ownership structure to examine variations in the impact of return on assets, return on equity, net interest margin and operating cost ratio determinants across ownership structures. The study is innovative in three ways: firstly, it cumulates financial and corporate governance indicators as explanatory variables; secondly, it uses an extensive number of explanatory and dependent variables compared with previous studies on Russia; thirdly, it classifies banking institutions by the definition of majority ownership at 20% level.

My findings are in line with previous studies⁴⁴, and suggest that there are some differences in determinants of bank profitability by ownership type: personnel cost and bank portfolio composition ratios dominate in state-influenced banks; capitalisation, reserves and default risk ratios in domestic private-owned banks; capitalisation ratios in foreign-owned banks. In addition, operating efficiency and credit risk ratios are important determinants of profitability across all ownership types. Furthermore, control variables such as inflation, GDP growth, the size of the banking system, and the bank asset share ratio are significant suggesting that they play an important role in the profitability of a bank. Corporate governance variables such as the BVD independence ranking and listing on an exchange are found to be insignificant. The location variable reduces the expenses of a bank if it is not centrally situated (outside Moscow or St Petersburg regions).

Overall, these empirical results provide evidence that the profitability of stateinfluenced, private domestic and foreign banks is shaped by different determinants. Indeed, this study shows that most of the traditional determinants of bank profitability ratios differ in their impact when considering ownership categories of banks in Russia. Nevertheless, there are also similarities across different ownership structures. For example, operating efficiency and credit risk ratios are important factors across all ownership types. Corporate governance dummies used here are found to be insignificant, but BVD has a positive effect on performance, while being listed on an exchange seems to be more important for foreign-owned banks present in Russia. Being located outside central regions of Moscow and St Petersburg reduces the expenses of a bank.

Similarly to Fungacova and Poghosyan (2009) I conclude that bank ownership plays an important role in emerging markets and should not be ignored when analysing performance determinants. The approach followed here can be a basis for further investigations of bank profitability determinants in emerging markets, and can be used to suggest optimal policies to implement to the bank’s management.

7. Limitations and Directions for Further Research

The main limitations and direction for further research are the following: - Due to lack of public information on corporate governance in Russian banks only a limited number of proxies have been used, available from Bankscope and CBR websites. It would be interesting to investigate the effects of further corporate governance indicators such as frequency of the executive body sessions, ownership concentration, stakeholder engagement and corruption for example⁴⁵.

- While the ownership of each bank has been reviewed according to Bankscope’s historical data, around 20%-50% of ownership structure is missing for the sample used here, depending on the year. In future research, it would be interesting to gain access to all historical data available for Russian banks and to re-classify them.

- Furthermore, the BVD indicator composed by Bureau Van Dijk does not change over time for the banks investigated. It would be interesting to come up with a more precise and sensitive ranking, especially in what concerns the classification of state-influenced banks in Russia. - There is a lack of publicly available market-based indicators of performance such as Price-to-Earnings ratio, credit default swaps, senior debt spread, etc. This type of data can be accessed from payable sources such as Interfax. Using these indicators as explanatory variables could be a good way to investigate further the hypothesis that Russian banks' performance is not linked to markets, as found in this study.

- To control more accurately for potential endogeneity problems instrumental variables or Arellano-Bond techniques could be used in future researches.

References

Athanasoglou, P., Brissimis, S., Delis, M. “ Bank-specific, Industry-specific and Macroeconomic determinants of bank profitability ”, Bank of Greece, 2004

Bashir, A. “ Assessing the Performance of Islamic Banks: Some Evidence from Middle East ”, Topics in Middle Eastern and North African Economies, 2001

Bebczuk, R. “ Corporate governance and Ownership: measurement and impact on corporate performance and dividend policies in Argentina ”, Inter- American Development Bank, 2005

Beck, T., Cull, R., Jerome, A., “ Bank privatisation and performance: Empirical; evidence from Nigeria ”, Journal of Banking and Finance, Vol. 29, pp.2355-2379, 2005

Berger, A., Humphrey, D. “ Efficiency of financial institutions: International survey and directions for future research ” , European Journal of Operational Research, Vol.98, pp.175-212, 1997

Berger, A., Mester, L. “ Inside the Black Box: What Explains Differences in the Efficiencies of Financial Institutions? ” , Financial Institutions Centre, 1997

Bokov, V., Vernikov, A. “ Quality of Governance and Bank Valuation in Russia: An empirical study ” , State University - Higher School of Economics, 2008

Bonin, J., Hasan, I., Wachtel, P. “ Bank performance, efficiency and ownership in transition countries ” , Journal of Banking and Finance, Vol.29, pp.31-53, 2005

Boyco, M., Shleifer, A., Vishny, R. “ Privatisation Russia ” , Brookings Papers on Economic Activity, Vol.2, 1993

Choi, S., Hasan, I. “ Ownership, Governance, and Bank Performance: Korean Experience ”, Financial Markets, Institutions and Instruments, Vol.14, No.4, 2005

Claeys, S., Vennet, V. “ Determinants of Bank Interest Margins in Central and Eastern Europe: A comparison with the West ” , Ghent University, Belgium, 2004

Cornett, M., Guo, L., Khaksari, S., Tehranian, H. “ The impact of state ownership on performance differences in privately-owned versus state-owned banks: an international comparison ” , Journal of Financial Intermediation, 2009

De, B. “ Ownership Effects on Bank Performance: A Panel study of Indian Banks ” , ICICI research centre, 2003

Demirgus-Kunt, A., Huizinga, H. “ Determinants of Commercial Bank Interest Margins and Profitability: Some International Evidence ” , The World Banj Economic Review, Vol.13, No.2, pp.379-408, 1999

Drakos, K. “ Assessing the success if reform in transition banking 10 years later: an interest margins analysis ” , Journal of Policy Modelling, Vol.25, pp.309- 317, 2003

ECB “ How to measure bank performance ” , Chapter 2, 2010

Fungacova, Z., Poghosyan, T. “ Determinants of bank interest margins in Russia: Does bank ownership matter? ” , BOFIT, 2009

Glushkova, E., Vernikov, A. “ Size of public sector in the Russian banking industry ” , State University - Higher School of Economics, pp.337-344, 2010

Goddard, J., Molyneux, P., Wilson, J. “ The profitability of European banks: a cross-sectional and dynamic panel analysis ” , The Manchester School, Vol. 72, No.3, pp.363-381, 2004

Gorshkov, V. “ Foreign banks ’ entry and the development of the Russian banking sector ” , conference of Economic and financial system development in the Pacific-Rim region, 2012

IFC “ Russian Banking sector corporate governance survey, a snapshot on improvements made ” , World Bank Group, 2007

Judge, W., Naoumova, I., Koutzevol, N. “ Corporate governance and firm performance in Russia: am empirical study ”, Journal of World Business, Vol.38, pp.385-396, 2003

Karas, A., Schoors, K., Weill, L. “ Are private banks more efficient than public banks? Evidence from Russia ” , BOFIT, 2008

Kwan, S. “ Operating performance of banks among Asian economies: An international and rime series comparison ” , Journal of Banking and Finance, Vol.27, pp.471-489, 2003

La Porta, R., Lopez-de-Silanes, F., Shleifer, A. “ Government Ownership of Banks ”, The journal of finance, Vol.12, No.1, 2002

Lloyd-William, D., Molyneux, P., Thornton, J. “ Market structure and performance in Spanish banking ”, Journal of Banking and Finance, Vol.18, pp.433-443, 1994

Love, I., Rachinsky, A. “ Corporate Governance, Ownership and Bank Performance in Emerging Markets: Evidence from Russia and Ukraine ”, Worldbank, 2007

Mahlberg, B., Bernhard, U., Haiss, P. “ Russian Banking: The impact of ownership on efficiency and performance ”, Vienna University of Economics and Business, Institute for International Business, 2011

McCarthy, D., Puffer, S. “ Corporate governance in Russia: a framework for analysis ” , Journal of World Business Vol.38, pp.397-415, 2003

McGee, R., Preobragenskaya, G. “ Recent developments in corporate governance in Russia ”, IABPAD, Annual Conference, 2004

Micco, A., Panizza, U., Yanez, M. “ Bank Ownership and Performance ” , Inter-American Development Bank, 2004

Mukherjee, A., Nath, P., Pal, M. “ Performance benchmarking and strategic homogeneity of Indian banks ”, International Journal of Bank Marketing, Vol. 20/3, pp.122-139, 2002

Naceur, S., Goaied, M. “ The determinants of the Tunisian deposit banks ’ performance ” , Applied Financial Economics, Vol.11, pp.317-319, 2001

Nikiel, E., Opiela, T. “ Customer type and bank efficiency in Poland, implications for emerging market banking ” , Contemporary Economic Policy, Vol.20, No.3, pp.255-271, 2002

Reuters “Putin signals end to Gazprom's Russian gas export monopoly”, 21st of June 2013 online article, http://uk.reuters.com/article/2013/06/21/putin-gas- exports-idUKL5N0EX1WJ20130621

Schwaider, S.M., Liebeg, D. “ Determinants of Bank Interest Margins in Central and Eastern Europe ”, Financial Stability Report, Issue 14, pp.68-84, 2007

Shleifer, A., Vishny, R. “ A survey of corporate governance ” , The journal of finance, Vol.12, No.2, 1997

Simberova, I., Kocmanova, A., Nemecek, P. “ Corporate Governance performance measurement - key performance indictors ” , Economics and Management, Vol. 17:4, 2012

Spong, K., Sullivan, R., DeYoung, R. “ What makes a bank efficiency? - A look at financial characteristics and bank management and ownership structure ” , Journal of banking and finance, pp.221-249, 1993

The Wall Street Journal “ Putin ’ s New Central Banker ” , Vol.XXXI, No.35, 20 March 2013

Tripe, D., Minh To, H. “ Factors influencing the performance of foreignowned banks in New Zealand ”, Journal of International Financial Markets, Institutions and Money Vol. 12, pp.341-357, 2002

Vernikov, A. “ Corporate Governance and control in Russian banking ” , State University - Higher School of Economics, 2007

Vernikov, A. “ Direct and Indirect state ownership of Banks in Russia ” , Higher School of Economics, Moscow, 2010

Vernikov, A. “ Government Banking in Russia: Magnitude and New Features ”, IWH Discussion Paper, 2011

Vernikov, A. “ Russia ’ s banking sector transition: Where to? ”, BOFIT, 2007

Vernikov, A. “ Russian banking: The state makes a comeback? ” , BOFIT, 2009

Vernikov, A., Glushkova, E. “ How big is the visible hand of state in the

Russian banking industry? ”, Higher school of Economics, Moscow, 2010

Williams, B. “ Factors affecting the performance of foreign-owned banks in Australia: A cross-sectional study ” , Journal of Banking and Finance, Vol. 22, pp.197-219, 1998

World Finance Review “ Russia ’ s Banking Sector Sees Growth in 2012 ” , p.42, September 2012

Appendices

Appendix A: Notes on the Empirical Research

1. Unusual and Influential data

The amount of data missing has been identified using ‘mdesc’ and⁴⁶ ‘codebook’ commands. I have also tested if the data is missing randomly using ‘mvpatterns’. The ‘graph matrix’ command allowed me to see the scatterplots of the variables and to look for any outliers. There seemed to be no data far away from other points that required extra attention. Nevertheless, after the regression analysis, I used the ‘predict’ command to create residuals, and examined the studentized residuals to identifying outliers.

I identified five extreme observations for Model one: bank 23 (r of - 23.25), bank 281 (r of -11.78), bank 304 (r of -10.01), bank 341 (r of 15.17), bank 330 (r of 14.66); two extreme observations for Model two: bank 23 (r of -22.43), bank 281 (r of 24.17); two extreme observations for Model four: bank 89 (r of 13.18), bank 202 (r of 12.47). The rest of residuals are grouped together, with the majority of them between -2 and +2. When running the regressions I have included and excluded these banks in each model respectively to identify any difference in coefficients and standard errors. The results reported in the regression analysis in Appendix B are the ones with the above- mentioned banks excluded, as econometric models fit better without these banks (smaller standard errors and higher R-squares).

2. Normality of Residuals

I used commands such as ‘kdensity’, ‘pnorm’, and ‘qnorm’ to check the normality of the residuals. There are also numerical tests for testing normality such as ‘swilk’ or ‘iqr’, but these tests tend to be very sensitive.

Normality of variables or residuals is not required in order to obtain

unbiased estimates of the regression coefficients. Normality is necessary only for hypothesis tests to be valid, the estimation of the coefficients only requires that the errors be identically and independently distributed.

Furthermore, there is no assumption or requirement that the predictor variables must be normally distributed. If this were the case then it would not be possible to use dummy coded variables.

3. Homoscedasticity

As mentioned above, one of the main assumptions for the ordinary least squares regression is the homogeneity of variance of the residuals. If the model is well fitted, there should be no pattern to the residuals plotted against the fitted values. If the variance of the residuals is non-constant then the residual variance is said to be heteroskedastic. There exist graphical and non-graphical methods for detecting heteroskedasticity. A commonly used graphical method is to plot the residuals versus fitted (predicted) values (‘rvfplot’ command with the ‘yline(0)’ option to put a reference line at y=0). I have also regressed the residuals to see if their mean is close to zero (desired outcome). It is the case for all the regression models used here.

Apart from the visual analysis for heteroskedasticity, Stata 12 has the ‘xttest3’ command that can be installed. The test’s null hypothesis is homoskedasticity. The option ‘robust’ has been used to control for heteroskedasticity problems in all regression models, as well as robust standard errors with the command ‘cluster’. With the robust option the standard errors take into account issues concerning heterogeneity and lack of normality. Therefore, standard errors are valid as they are heteroskedasticity- and autocorrelation-consistent - HAC clustered standard errors.

4. Multicollinearity

The term collinearity implies that two variables are near perfect linear combinations of one another. I have checked for potential multicollinearity issues using correlation matrixes (‘corr’ and ‘pwcorr’), and excluded correlation over 0.70 as per convention rules. I have also checked with the command ‘collin’. For all the regressions used here, when run as simple multivariate regressions, VIF is no bigger than 2.92. This means that no multicollinearity issues have been identified.

5. Linearity

Variables have been explored for normality with graphical methods such as histograms, boxplots, and stem-and-leaf plots. Variables such as CAP1, CR2, IM, OHTA and CAR fit better the normality assumption when transformed with the ‘log’ function. The natural logarithm cannot be used for variables that take negative values (such as ROA, ROE, NIM, OCR, CAP2, DF). To select the appropriate transformation and make variables more normally distributed, Stata has the ‘ladder’ and ‘gladder’ commands.

Nevertheless, some variables cannot be completely normalized.

I used the command ‘acprplot’ to see the linearity of the predictor variables, and there is very little deviation from linearity.

I used the command ‘nlcheck’ which can be installed in Stata 12 to test the linearity of individual regressors. A significant test result indicates that the linearity assumption is violated. I have also tested graphically with scatterplots for linearity.

6. Model Specification

The question if the model has an omitted or an irrelevant variable is more of a theoretical than of a statistical one. All the variables used here are backed up by past research and financial theory.

7. Independence

Independence is necessary so that the errors associated with one observation are not correlated with the errors of any other observation. This dataset contains data on 350 banks from Russia. It is very possible that the scores for each bank may not be independent, and this could lead to residuals that are not independent within the dataset. I used the command ‘xtserial’ to test the autocorrelation of the error term with panel data, and this command performs the Lagram- Multiplier test for serial correlation. The null is no serial correlation. Nevertheless, if autocorrelation is detected the command ‘cluster’ (which is used for heteroskedasticity issues as well) is used to remedy autocorrelation problems.

Appendix B: Summary Statistics and Regression Results

illustration not visible in this excerpt

Appendix C: List of Banks

illustration not visible in this excerpt

[...]

---

¹ La Porta et al (2002)

² McGee, Preobragenskaya (2004); IFC (2007)

³ Refer to Boycko et al (1993)

⁴ Reuters, 21st of June 2013 online article

⁵ Vernikov (2007); Vernikov (2009); Vernikov (2011)

⁶ Sberbank, VTB, Gazprombank, Russian Agricultural Bank (also referred to as Rosselkhozbank), and Bank of Moscow by total assets in 2012 (statistics from the CBR website)

⁷ Previous studies on the Russian Banking sector have been done in their majority by Vernikov, who defines major ownership at 50%. In this study I decide to follow La Porta et al (2002) who define majority ownership at 20%.

⁸ Capital, credit risk, productivity, expenses management

⁹ Refer to p.748 in Shleifer, Vishny, 1997

¹⁰ Both the development and the political views imply that state ownership of banks should be more important in poorer countries, countries with less developed financial markets and less well functioning institutions. The development view also implies that state ownership should benefit financial and economic development and productivity growth. The political view implies that governmental ownership of banks will crowd out the financing of private firms (La Porta et al, 2002).

¹¹ Deloitte & Touche, KPMG, PwC, Ajour, PKF, IDA (p.5, McGee, Preobragenskaya, 2004)

¹² Under the Soviet Era, the Russian mentality was all about secrecy and non-disclosure and it is still difficult to overcome this cultural issue.

¹³ Non-separation of CEO from chairperson.

¹⁴ International Finance Corporation, part of the World Bank Group

¹⁵ Vernikov (2007); Vernikov (2009); Vernikov, Glushkova (2010); Vernikov (2010); Vernikov (2011)

¹⁶ When state authorities act as direct shareholder or a majority stake in a bank is owned by stateowned enterprises (Vernikov, Glushkova, 2010). They can be directly state-owned - major or sole shareholder is a federal, regional or municipal executive body, or the Central Bank of Russia; or indirectly state-owned - banks owned by stateowned companies and banks, by the Deposit Insurance Agency or by the Bank for Development and Foreign Economic Affairs (Vneshekonombank) (Vernikov, 2010).

¹⁷ The list of the top 350 banks by total assets as in 2012 used in this paper can be found in Appendix C

¹⁸ ‘mdesc’ and ‘codebook’ are commands used to see the missing values, and ‘mvpatterns’ to see that the variables are missing randomly. See Appendix A for further details.

¹⁹ In stata ‘tabulate owner’, ‘generate(dum)’ and ‘xi i.owner’ commands are used to generate ownership dummies. See Appendix A for further details.

²⁰ Fungacova and Poghosyan (2009) used both 50% and 20% thresholds for foreign ownership in their robustness checks and found that both results are in line with each other.

²¹ As in Kwan (2003), De (2003)

²² As in Demirguc-Kunt, Huizinga (1999)

²³ As in Kwan (2003)

²⁴ As in Fungacova, Poghosyan (2009)

²⁵ A non-performing loan is a sum of borrowed money upon which the debtor has not made his/her scheduled payments for at least 90 days. A non-performing loan is either in default or close to being in default.

²⁶ For more details refer to Micco et al (2004)

²⁷ As in Cornett et al (2009)

²⁸ As in Bashir (2001)

²⁹ As in Naceur, Goaied (2001)

³⁰ As in Cornett et al (2009)

³¹ As in Drakos (2003)

³² In Stata, ranking letters from A+ to D are coded with numbers from 1 to 9 accordingly, and U is coded with 0.

³³ This dummy has been used by Naceur, Goaied (2001).

³⁴ Bourke (1989) and Athanasoglou et al (2004) suggest that any functional form of bank profitability is qualitatively equivalent to the linear.

³⁵ As in Athanasoglou et al (2004)

³⁶ The rationale behind RE model is that the variation across entities is assumed to be random and uncorrelated with the predictor variables. Therefore, in RE model it is assumed that explanatory variables are independent of the idiosyncratic error term and of the individual random effect, while in FE model the explanatory variables are independent of the idiosyncratic error term but they are not independent of the individual fixed effect.

³⁷ In stata the command ‘testparm i.year’ is used to see if time effects are needed. It is a joint test to see if the dummies for all years are equal to zero, and if they are then no time fixed effects are needed.

³⁸ The two unit root tests for unbalanced panel data - Im-Pesaran-Shin and Fisher-type tests do not give any consistent results given the relatively short time span used in this study.

³⁹ Athanasoglou et al (2004) compared the FE and the GMM estimates (which address the endogeneity issues) on the profitability of banks in Greece and found that the two methods produced similar results.

⁴⁰ As in Fungacova, Poghosyan (2009)

⁴¹ World Finance Review, 2012.

⁴² Schwaiger and Liebeg (2007) find a significant positive effect of credit risk proxy on NIM in a set of CEEs countries.

⁴³ Revell (1979) introduces the issue of performance and inflation. He finds that the effect of inflation on the profitability of a bank depends on the rate of increase of a bank’s wages and operating expenses. So the extent to which inflation affects bank profitability depends on whether inflation expectations are fully anticipated.

⁴⁴ Especially with Fungacova and Poghosyan (2009) who found that ownership should be considered when analysing the determinants of the net interest margins of Russian banks.

⁴⁵ Refer to the study by Simberova et al (2012) for an extensive list of corporate governance indicators

⁴⁶ I omit, due to space considerations, showing graphs and commands described in the Notes section.