Bank ownership and cost efficiency: New empirical evidence from Russia

Accepted Manuscript Title: Bank ownership and cost efficiency: New empirical evidence from Russia Authors: Mikhail Mamonov, Andrei Vernikov PII: DOI: Reference:

S0939-3625(17)30027-4 http://dx.doi.org/doi:10.1016/j.ecosys.2016.08.001 ECOSYS 602

To appear in:

Economic Systems

Received date: Revised date: Accepted date:

21-9-2015 15-7-2016 16-8-2016

Please cite this article as: Mamonov, Mikhail, Vernikov, Andrei, Bank ownership and cost efficiency: New empirical evidence from Russia.Economic Systems http://dx.doi.org/10.1016/j.ecosys.2016.08.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Bank ownership and cost efficiency: New empirical evidence from Russia Mikhail Mamonova, b,* and Andrei Vernikovc, d,** a

Center for Macroeconomic Analysis and Short-Term Forecasting (CMASF) at the Institute of Economic Forecasting, Russian Academy of Sciences, 47 Nakhimovsky prospekt, 117418 Moscow, Russia

b

National Research University Higher School of Economics, 20 Myasnitskaya Str., 101000 Moscow, Russia

c

Institute of Economics, Russian Academy of Sciences, 117218 Moscow, Russia

d

Moscow School of Management Skolkovo, 143025 Moscow Region, Russia

* E-mail address: [email protected] (M. Mamonov) **Corresponding author. Tel.: +7 495 629 7495; E-mail address: [email protected] (A. Vernikov)

Highlights 

We reinvestigate the nexus between bank ownership and cost efficiency in Russia.



Core state banks are as efficient as private ones and outperform foreign banks.



Shifts in the group efficiency ranking are driven by capitalization and asset structure.



Core state banks gain cost efficiency when they lend less to the economy.



Private domestic banks outperform foreign banks when maintaining lower capitalization.

1

Abstract This paper investigates the cost efficiency of Russian banks with regard to their heterogeneity in terms of ownership form, capitalization and asset structure. Using bank-level quarterly data over the period 2005-2013, we perform stochastic frontier analysis (SFA) and compute cost efficiency scores at the bank and bank group levels. We deduct from gross costs the negative revaluations of foreign currency items generated by official exchange rate dynamics rather than by managerial decisions. The results indicate that the core state banks, as distinct from other state-controlled banks, were nearly as efficient as private domestic banks during and after the crisis of 2008-2009. Foreign banks appear to be the least efficient market participants in terms of costs, which might reflect their lower (and decreasing over time) penetration of the Russian banking system. We further document that the group ranking by cost efficiency is not permanent over time and depends on the observed differences in bank capitalization and asset structure. We find that foreign banks gain cost efficiency when they lend more to the economy. Core state banks, conversely, lead in terms of cost efficiency when they lend less to the economy, which can result from political interference in their lending decisions in favor of unprofitable projects. Private domestic banks that maintain a lower capitalization significantly outperform foreign banks and do not differ from the core state banks in this respect.

Keywords: Cost efficiency, Bank ownership, Capitalization, Asset structure, Stochastic frontier analysis, State-controlled banks, Russia

JEL classification: G21, P23, P34, P52

1. Introduction The Russian banking system operates in an adverse economic environment and keeps shrinking numerically, so each of the remaining banks must improve cost efficiency in order to remain sustainable and competitive. Our research question relates to the comparative efficiency of Russian banks depending on ownership type (public, private or foreign), risk preference, and asset structure. Russia stands out as a special case among transition countries in the

2

sense that public banks, which have largely gone extinct in Central and Eastern Europe (CEE), have a market share of 60% or so. At the same time, foreign banks hold no more than 10% of the banking sector assets, unlike in the CEE region where they may have over 90% in some countries. The main stream of literature on comparative bank efficiency in transition suggests that foreign banks lead in terms of efficiency, whereas state-owned banks lag behind (Bonin et al., 2005; Fries and Taci, 2005; Grigorian and Manole, 2006; Fries et al., 2006). These results may have reflected the high expectations with regard to the privatization of banks in favor of foreign direct investors. Empirical studies on Russian banks have, however, offered mixed evidence regarding the relative performance of state-, private- and foreign-controlled institutions (Styrin, 2005; Golovan et al., 2008; Karas et al., 2010; Fungáčová and Poghosyan, 2011; Mamonov, 2013). In this paper we will perform stochastic frontier analysis (SFA) of Russian bank-level quarterly data over the period 2005–2013 with the aim of computing cost efficiency scores at the bank level and subsequently at the bank group level. Then we will explain the changes in bank efficiency rankings by employing the generalized method of moments (GMM) to estimate a set of distance functions that measure the dependence of the observed differences in SFA scores of banks and bank clusters on either bank-level risk preference or asset structure. There are three reasons for revisiting the interplay between bank ownership and cost efficiency in Russia. First, we focus on the impact of the revaluation of foreign currencydenominated items and securities on the cost efficiency estimations and argue that revaluation proceeds must be removed from the gross costs of the banks as they are generated by exchange rate volatility rather than by managerial decisions. Second, unlike previous studies, we cover the period before, during and after the crisis of 2008 through 2013 in order to gauge the impact of the crisis on bank efficiency. Third, we reveal the mechanisms driving both the within- and between-group heterogeneity of bank efficiency rankings. We study these mecha-

3

nisms in dynamics, whereas previous research has only analyzed them in static, e.g. averaging the impact of different cost efficiency covariates like ownership or market power on bank efficiency scores over the period considered. We stress that the longer the period, the less accurate the averaged efficiency ranking might be. We investigate how specific drivers of cost efficiency such as asset structure and risk preference affect the banks’ efficiency rankings throughout our sample period. The rest of the paper is organized as follows. The next section reviews relevant literature on comparative bank efficiency in Russia. Section 3 explains our methodology and data. Section 4 contains the estimation results and their discussion. Section 5 concludes.

2. Review of selected literature on the bank ownership-efficiency nexus in Russia Within the strong stream of single-country and cross-country studies dedicated to bank efficiency, only the selected papers cover Russia and explore the role of the ownership form as a determinant of efficiency within a stable sample, as distinct from the dynamic effects of ownership changes such as privatization. Empirical research on privatization effects is less relevant in the Russian case because a genuine privatization of banks never took place, especially with regard to the largest state banks. In Central and Eastern Europe, in contrast, it is now hard to study the comparative performance of state banks alongside private and foreign banks over an extended period of time because few, if any, state banks are left. In this paper we focus on the effect of public ownership and not ownership change. The literature so far evidences mixed empirical results: while multi-country studies reinforce the bias against public banks and in favor of foreign banks, single-country studies on Russia usually fail to document this statistically. Using SFA, Fries and Taci (2005) examine the cost efficiency of 289 banks in 15 postsocialist countries in 1994-2001. As a rule, private banks appear more efficient than stateowned banks, but there are differences among private banks: majority foreign-owned ones are

4

more efficient than those with domestic owners. Predictably, a higher share of foreign-owned banks in the banking sector assets is associated with lower costs, and therefore a more efficient financial intermediation, but that only applies to cases where foreign-owned banks have a larger share of the total assets. The latter proviso clearly makes Russia ineligible. Also, a relatively small sample of Russian banks is entered in the general cross-country panel. Fries et al. (2006) estimate the margins and marginal costs of banks in 15 transition countries including Russia for the period 1995–2004. The consistent result is the relative inability of state-owned banks to increase their mark-ups. The authors interpret this as stateowned banks’ lesser ability to attract demand for their services compared to their competitors. The single-country literature on Russia disagrees on the relative efficiency of public, domestic private and foreign banks. Styrin (2005) uses various techniques to calculate X-inefficiency scores of Russian banks for the period 1999-2002. Foreign-owned banks tend to be more efficient than domestic ones. The variable responsible for the affiliation of a bank with the federal or a local government is insignificant in most cases, meaning that banks affiliated with the state do not seem more or less efficient, all other things being equal. Quantitatively the effect is negative, which Styrin (2005) interprets as a sign that many state-owned banks only serve as vehicles for channeling budget funds to state-owned companies or servicing budget accounts and budget expenses. Those activities do not require expertise and/or resources for managing risks and monitoring customers, so there might be an illusion of higher efficiency of state-related banks. Styrin (2005) believes that in Russia state-owned banks are financial intermediaries to a much lesser extent than private banks. Foreign banks did not display higher efficiency than Russian banks in 2002-2005, and within the group of the top 100 banks the foreign ones actually showed lower efficiency (Golovan et al., 2008). That may be due to heavy investments in infrastructure over the period

5

of observation, at the cost of short-term profitability. It is size with the economies of scale determining the gap efficiency rather than other factors including ownership form. Unlike several studies on banking in transition, the article by Karas et al. (2010) is about public and private ownership and not privatization. The sample consists of 747 banks for 2002 and 471 banks for 2006. More or less in line with the mainstream of the economic transition literature, the authors find that foreign banks are more efficient than domestic private banks. However, their other conclusion is counter-intuitive: public banks are not less efficient than domestic private banks. The relative efficiency gap between public banks and domestic private banks becomes significant after the introduction of an explicit deposit insurance scheme. The authors claim that this unorthodox result cannot be attributed to the choice of production process, bank-level characteristics, the econometric approach, or the introduction of deposit insurance. The explanation offered relates to a different set of activities among public and private banks, as the typical activity mix of public banks involves fewer costs than that of private banks. Fungáčová and Poghosyan (2011) analyze interest margin determinants in the Russian banking sector, which happen to be the precursor of efficiency, with a particular emphasis on the ownership structure. In the 1999-2007 period, the impact of commonly used determinants such as market structure, credit risk, liquidity risk and size of operations is found to differ across state-controlled, domestic private and foreign-owned banks. The average net interest margins charged by state-controlled and private domestic banks do not differ significantly. Still, state banks do not exploit their market power, stemming from their large market share, to the full extent when they set interest rates. If bank operational efficiency is proxied by the cost-to-income ratio, then for the period 2004-2012 state banks appear to control their costs better than other bank groups, and that holds for the period before the 2008 crisis as well as for the period after it (Mamonov, 2013). This is another counter-intuitive finding that we would like to re-check.

6

3. Methodology and data 3.1 Estimation of the cost frontier: Revaluation proceeds excluded We use the stochastic frontier technique to compute time-specific rankings in bank cost efficiency. The empirical cost function at the bank level is specified within the production approach, taking into account prices of inputs, quantities of outputs, and equity netputs to control for differences in bank risk preference (Turk Ariss, 2010; Fiordelisi et al., 2011). We prefer the production approach over the intermediation approach because it avoids a possible bias of efficiency estimates due to incomplete assets and liabilities coverage in the intermediation approach (Fortin and Leclerc, 2007). Different from previous research, we focus on revaluation proceeds to render additional precision to the efficiency estimating procedure. Relying on detailed profit and loss accounts of Russian banks in quarterly format over the past decade (see Section 3.3 for details), we analyze revenue and cost structures of banks and find that positive revaluations of foreign currency and securities are the largest component of gross revenues, while negative revaluations of foreign currency and securities are the largest component of gross costs (see Table A.1). Both are more significant than interest income and interest expense, respectively. At the same time, the difference between positive and negative revaluations is close to zero for the majority of Russian banks. This apparent paradox may reflect the fact that the Russian economy, like some other transition and emerging economies, remains volatile and dollarized (Brodsky, 1997; Sutela, 2013), resulting in (i) great instability of the Ruble exchange rate, and (ii) a significant share of bank balance sheet items denominated in foreign currencies.1 These properties launch the permanent mechanism of revaluations (revals hereinafter) — positive on the revenue side and 1

The share of items denominated in foreign currencies in total liabilities rose from 29.5% on the eve of the 2008 financial crisis to 31.2% two years later; it then gradually declined to 21.2% by the end of 2013, which is still economically significant (Table A.1).

7

negative on the cost side of bank balance sheets. Further analysis reveals that almost in all percentiles of revals distribution (Figure 1) the correlation between revals and the Ruble exchange rate is strong, at 81% on average (see Table 1).

Revals are mainly driven by the Ruble instability, which is beyond the control of bank managers. At the same time, it can be argued that the precise currency structure of a bank’s assets and liabilities is within the control of the management, at least to some extent, so that a good manager can shift this structure in favor of a stronger foreign currency. But Russian banks cannot take significant positions in foreign currency because their individual limits on open positions in foreign currency are set and monitored by the regulator. If a bank conducts a large part of its operations in foreign currency, then in the case of domestic currency devaluations the positive revals will have a greater weight within that bank’s revenues, and negative revals will have a proportionately greater weight within the costs. As a result, the net gain for the entire banking sector is negligible at no more than 0.1% of total bank assets (Table A.1). Therefore, banks with greater negative revals are not necessarily those whose managers do not manage the currency structure of assets and liabilities properly. The same banks may well have large revenues stemming from positive revals, but we do not see them because we focus on the cost structure. We are interested in the results of banking activity (attracting deposits, granting loans, etc.) in terms of cost efficiency. In Table A.1 in the Appendix we show all necessary lines of P&L accounts, including costs, as a percentage of total assets in order to highlight the scope of our investigated issue relative to bank size. In Table A.2 in the Appendix we report the descriptive statistics of all variables involved in the cost efficiency estimations.

8

Since we do not recognize currency speculation to be a core function of a commercial bank, we assume that the magnitude of revals2 is largely exogenous and does not reflect managerial efforts, so that revals can be deemed irrelevant and alien to the concept of operating efficiency. Nonetheless, dropping revals from total costs would not affect the econometric estimations if the share of negative revals in total costs was uniformly distributed across banks and all banks had a similar share of negative revals in their gross costs. But this share varies very broadly in our sample — from almost 0% to 95% (see Figure 1), so banks differ substantially in this respect. Consequently, the use of gross data including revals can seriously influence efficiency rankings, especially during financial turmoil. We provide further insights into the behavior of revals with respect to bank size, risk preference, and bank ownership when describing our data in Section 3.3. Additionally, in our baseline estimations of cost efficiency we follow Berger and DeYoung (1997), Maudos and Fernández de Guevara (2007) and Solis and Maudos (2008) and deduct interest expenses from total costs on the assumption that interest expenses reflect not only cost efficiency, but, to some extent, the bank’s market power. For example, when a bank trims the interest rate on attracted funds, it may reflect improved market power rather than cost efficiency management. Nonetheless, as a robustness check, we replicate all our baseline estimations with interest expenses placed back into the cost frontier. In the specific case of Russia, where several large banks have a great market share, it is crucial to account for the differences in market power (Maudos and Fernández de Guevara, 2007; Solís and Maudos, 2008; Turk Ariss, 2010). The “market power” strand of the literature employs the two-stage approach, which separates cost (profit) frontier evaluation from the estimation of its determinant, as do some papers dealing with comparative bank efficiency (Berger et al., 2009), because the effect of market power on efficiency cannot be estimated

2

As we only analyze cost efficiency in this paper, revals hereinafter means negative revaluations as a part of bank costs. We no longer pay attention to positive revals as they are attributed to revenues, not costs.

9

within the one-stage approach, i.e. by estimating the frontier and the inefficiency covariates simultaneously. Bank-level estimates of market power are usually based on the estimates of marginal costs (Lerner index or Boone indicator) obtained after the frontier is evaluated. We thus opt for the two-stage approach, although we recognize the concerns regarding it (Wang and Schmidt, 2002; Belotti et al., 2013) and are aware that several authors prefer its alternative — the one-stage approach (Bonin et al., 2005; Karas et al., 2010). We perform one of the robustness checks with the one-stage approach, dropping market power from the list of inefficiency covariates. Eventually, the empirical cost function takes the following (translog) shape: 3

3

3

3

1 𝑙𝑛𝑂𝐶𝑖𝑡 = 𝛽0 + ∑ 𝛽𝑗 𝑙𝑛𝑌𝑗,𝑖𝑡 + ∑ ∑ 𝛽𝑘𝑙 𝑙𝑛𝑌𝑘,𝑖𝑡 𝑙𝑛𝑌𝑙,𝑖𝑡 + ∑ 𝛾𝑚 𝑙𝑛𝑃𝑚,𝑖𝑡 2 𝑗=1

𝑘=1 𝑙=1

3

𝑚=1

(1)

3

1 + ∑ ∑ 𝛾𝑟𝑞 𝑙𝑛𝑃𝑟,𝑖𝑡 𝑙𝑛𝑃𝑞,𝑖𝑡 + 2 𝑟=1 𝑞=1

3

3

3

3

+ ∑ ∑ 𝛿𝑠𝑢 𝑙𝑛𝑌𝑠,𝑖𝑡 𝑙𝑛𝑃𝑢,𝑖𝑡 + ∑ 𝜑𝑗 𝑙𝑛𝑌𝑗,𝑖𝑡 𝑇 + ∑ 𝜓𝑚 𝑙𝑛𝑃𝑚,𝑖𝑡 𝑇 + 𝛼1 𝑇 + 𝛼2 𝑇 2 𝑠=1 𝑢=1

𝑗=1

𝑚=1

+ 𝜇1 𝑙𝑛𝐸𝑄𝑖𝑡 + 3

3

+𝜇2 (𝑙𝑛𝐸𝑄𝑖𝑡 )2 + ∑ 𝜌𝑗 𝑙𝑛𝑌𝑗,𝑖𝑡 𝑙𝑛𝐸𝑄𝑖𝑡 + ∑ 𝜉𝑚 𝑙𝑛𝑃𝑚,𝑖𝑡 𝑙𝑛𝐸𝑄𝑖𝑡 + 𝜂𝑇𝑙𝑛𝐸𝑄𝑖𝑡 + 𝑣𝑖𝑡 + 𝑢𝑖𝑡 𝑗=1

𝑚=1

where for bank 𝑖 at time 𝑡, 𝑂𝐶𝑖𝑡 are total costs with (i) interest expenses and (ii) negative revals dropped. 𝑌𝑗,𝑖𝑡 is a 𝑗-th output: loans to households and nonfinancial firms (𝑗 = 1), retail and corporate deposits (without government and inter-bank accounts, 𝑗 = 2), fee and commission income as a proxy for non-interest-based output (𝑗 = 3). 𝑃𝑚,𝑖𝑡 is an 𝑚-th factor input price: average funding rate as a price of funds (𝑚 = 1), personnel expenses to total assets ratio as a price for labor (𝑚 = 2) and other non-interest and non-personnel expenses to total assets ratio as a proxy for the price of physical capital (𝑚 = 3). 𝐸𝑄𝑖𝑡 is equity capital as a netput factor reflecting differences in managers’ risk preferences. 𝑇 is the time trend. 𝑣𝑖𝑡 + 𝑢𝑖𝑡 is a

10

composite error term, where 𝑣𝑖𝑡 ~𝒩 (0, 𝜎𝑣2 ) is a random error that follows symmetric normal distribution (by assumption). 𝑢𝑖𝑡 ~𝒩 (0, 𝜎𝑢2 ) captures cost inefficiency and is set to follow a (positive) half-normal distribution. In estimating the empirical cost function we standardly impose linear homogeneity conditions as well as symmetry conditions on factor input prices. Having estimated the parameters of the cost function, we compute cost efficiency scores for bank 𝑖 at time 𝑡: 𝑆𝐹𝐴𝑖𝑡 = 𝑒𝑥𝑝{−𝑢̂𝑖𝑡 }

(2)

where 𝑢̂𝑖𝑡 is an estimate of the inefficiency term. The obtained bank-level cost efficiency scores are aggregated into group-level scores by within-group averaging for each bank ownership status (Section 3.3).

3.2 Introducing bank-level heterogeneity into the relationship between bank ownership and cost efficiency We proceed with a heterogeneity analysis to explain the observable differences in cost efficiency levels, e.g. SFA scores from Equation (2), both within a particular group of banks (core state-controlled banks, other state-controlled banks, and foreign-controlled banks) and among them. The motivation is that efficiency rankings may not be permanent over time and may depend on bank-specific factors. Thus, some banks in one group may be more cost efficient than banks in another group, even if the average ranking is the opposite at the group level. Where this is the case, it becomes important to find out why and when some banks in a less efficient group are more efficient than banks with similar characteristics in a more efficient group. For bank-specific factors, we use the loans-to-assets ratio (LTA) to catch differences in funds allocation between interest-generating and non-interest-based activities, and the equityto-assets ratio (ETA) to manage the variation in risk preference. To test these two factors, we augment the empirical equations employed in Solís and Maudos (2008), Maudos and Fernández de Guevara (2007) and Turk Ariss (2010) by introducing the interaction terms of a bank

11

group dummy variable and either LTA or ETA: 3

3

𝑆𝐹𝐴𝑖𝑡 = 𝛼ℎ,𝑖 + ∑ 𝛽ℎ𝑗 𝐺𝑅𝑂𝑈𝑃𝑗 + ∑ 𝛾ℎ𝑗 𝐺𝑅𝑂𝑈𝑃𝑗 𝑋ℎ,𝑖𝑡 + 𝛾ℎ 𝑋ℎ,𝑖𝑡 + 𝑗=1 𝐾

(3)

𝑗=1 𝑀

+ ∑ 𝛿ℎ𝑘 𝐵𝑆𝐹𝑘,𝑖𝑡 + ∑ 𝜑ℎ𝑚 𝑀𝐴𝐶𝑅𝑂𝑚,𝑡 + 𝜀ℎ,𝑖𝑡 𝑘=1

𝑚=1

where for bank 𝑖 at time 𝑡 𝑆𝐹𝐴𝑖𝑡 is the cost efficiency score from Equation (2). 𝑋ℎ,𝑖𝑡 is the ℎth potential candidate for efficiency heterogeneity factors. Specifically, we are interested in whether bank-level differences in risk preferences (as proxied by equity-to-assets ratios, ℎ = 1) or asset structure (defined through loans-to-assets ratios, ℎ = 2) matter for the bank-level heterogeneity of SFA scores within or between groups of banks. 𝐵𝑆𝐹𝑘,𝑖𝑡 is the 𝑘-th bankspecific factor that may affect cost efficiency: size, share of retail loans in total loans, loans dynamics, loans-to-deposits ratio, and market power (price mark-up as measured by efficiency and the funding-adjusted Lerner index, a modification proposed by Koetter et al., 2012). 𝑀𝐴𝐶𝑅𝑂𝑚,𝑡 is the 𝑚-th macroeconomic factor to control for business cycle, ruble volatility and borrowers’ creditworthiness. As a basic estimator of Equation (3), we exploit a 2-step GMM to address possible endogeneity and heteroscedasticity concerns. As instruments, we use the two first lags of all endogenous (bank-level) variables. Our main hypotheses regarding the chosen heterogeneity factors 𝑋ℎ,𝑖𝑡 are as follows. First, larger equity relative to assets provides potential for maintaining and expanding commercial loans (one of the three outputs included in our cost function). The higher a bank’s equity-to-assets ratio, the greater its outputs can be with the same volume of costs. This implies higher SFA scores. This is in line with Berger and Mester (1997), who claim that prudent banks are likely to have higher efficiency levels. On the other hand, holding more capital may be costly if it implies lower lending activities in the current period (Koetter and Poghosyan, 2009; Williams, 2012). To investigate which of these competing effects predomi-

12

nates in the Russian banking setting and for each bank group, we include equity capital at both steps (cost frontier and efficiency equation). Thus, we do not treat a bank as inefficient for being risk-averse, but still acknowledge that risk aversion may come with additional costs and/or benefits under Berger and Mester’s prudent-efficient hypothesis. Second, intensifying lending activities may facilitate economy-of-scale effects, i.e. a higher loans-to-assets ratio may positively affect cost efficiency (SFA score) (Solís and Maudos, 2008; Williams, 2012). Similarly to the previous case, a bank group could shorten the distance to the reference group by increasing its loans-to-assets ratio. At the same time, increased lending could increase borrower-screening costs and thereby lower cost efficiency (Williams, 2012). As in the previous case, we define the prevailing effect in the Russian banking system empirically. On the basis of Equation (3), we determine the distance of a bank from each group and the referent group in terms of SFA scores. We refer to that as distance functions. In this paper we consider only a time-averaged version of these functions that can be represented as follows: Δ𝑆𝐹𝐴𝑗,𝑖 = 𝛽ℎ𝑗 + 𝛾ℎ𝑗 𝑋ℎ,𝑖𝑡

(4)

For each bank 𝑖 from 𝐺𝑅𝑂𝑈𝑃𝑗 , ∆𝑆𝐹𝐴𝑗,𝑡 < 0 (∆𝑆𝐹𝐴𝑗,𝑡 > 0) implies that this bank is less (more) cost efficient compared to the average privately owned bank. Earlier research typically only analyzes the time-invariant first component on the right-hand side of Equation (4), i.e. 𝛽ℎ,𝑗 (Bonin et al., 2005; Karas et al., 2010).

3.3 Data, bank groups, and the behavior of revaluations Data. For bank-specific factors, we collect disaggregated bank-level data from Russian bank balance sheets and profit-and-loss (P&L) statements that are publicly available via the Central Bank of Russia website (www.cbr.ru). We combine monthly balance sheets with quarterly P&L statements in a quarterly panel dataset.

13

The initial sample includes all Russian banks that disclose their financial accounts data, amounting to 1,248 financial institutions over 2005Q1–2013Q4 that represent on average 95% of the total banking sector assets. This yields 36,422 bank-quarter observations for the pooled sample. Quarterly macroeconomic variables come from the Federal State Statistics Service website (www.gks.ru). We use daily data on the ruble exchange rate relative to a dual currency basket (0.55 US dollar and 0.45 euro) published by Finam (www.finam.ru). We estimate Equation (1) over the 40 quarters of 2005–2013. This fairly long observation period can be broken down into three sub-periods, i.e. before, during, and after the 2008– 2009 financial crisis. Although some changes in the underlying cost function may have occurred due to the destructive nature of the crisis, we do not separate the estimations for these three sub-periods in our basic version, for two reasons. First, the 2008–2009 crisis was expeditiously tackled by the Russian monetary authorities, who released 1.08 trillion rubles (3% of GDP in 2009) in subordinated loans or secondary public offerings (SPOs) to support systemically important banks and developed flexible instruments for liquidity support to soften the impact of the crisis (Solntsev et al., 2010). Second, the more flexible translog form of the cost frontier implies time- and bank-specific relationships between costs and key explanatory variables. Hence, the influence of the crisis may have already been accounted for. We apply common filtering procedures to our panel dataset to deal with outliers. First, we exclude the data on relative indicators (with the exception of bank size) below the 1st and above the 99th percentiles of the initial sample. Further, we drop observations with a loans-toassets ratio smaller than 10% to focus on banks providing credit to the economy and eliminate entities that do not function as genuine banks (Schoors, 2000; Karas and Schoors, 2010). After these filtering procedures, we have unbalanced panel data for 1,038-1,196 entities, and the number of observations ranges from 17,401-20,319 in Equation (3) to 29,082-29,146 in Equation (1).

14

Bank groups. In order to examine variations in the impact of cost efficiency determinants across ownership forms, we divide the sample according to the bank ownership form. We differentiate among four types of banks: core state-controlled, other state-controlled, domestic private and foreign. We extend previous studies on comparative banking in transition by subdividing state banks into two groups: the core (State-1, comprising Sberbank, VTB, and Rosselkhozbank) and other state-controlled banks (State-2, comprising between 28 and 46 institutions). We deem this important in the case of countries with a vast public sector: for instance, China’s “Big Four” state banks are analyzed separately from the other state-controlled banks (Berger et al., 2009). Russia’s state banks are too numerous, their business models too diverse, and their market share too large (about 60%) to appear as a single group in the estimation of comparative efficiency. While Russia’s three largest state banks often act as government agents and pursue a combination of financial and non-financial objectives (Vernikov, 2014), many of the smaller state-controlled banks, particularly the indirectly owned ones, display market behavior similar to private domestic institutions. The structure of the banking industry suggests that each of the two groups of state banks will be more homogeneous than the entire category. State-1 have between 35% and 43% of total banking sector assets in Russia, and State-2 have 19% (see Table 2).

The group of foreign-controlled banks (Foreign) comprises between 27 and 48 entities with 8-12% of total banking sector assets. We build our own sample with a focus on institutions that are fully or predominantly owned by foreign banks.3 The assumption is that performance characteristics of banks controlled by foreign strategic investors should be coherent and well pronounced. The remaining group of banks privately owned by Russian residents

3

More detail on the coverage of banks is provided in Mamonov and Vernikov (2015) and Vernikov (2015).

15

(Private) covers some 745 to 920 banks with a market share ranging from 31% to 42% of total assets. We revise the composition of each group for each quarter to reflect possible migrations. Revals behavior. In order to shed more light on the specifics of revals behavior in Russian banks, we split the sample into two parts according to the size of assets: abovemedian (large banks) and below median (small banks). Then we analyze revals and risk preference in each subsample in two dimensions: the whole banking sample and within each of the four groups (panels 1 and 2 of Table 3, respectively). We observe that, irrespective of the stage of the business cycle in Russia, small banks are better capitalized and operate mostly in national currency, so that revals are much lower compared to the large banks. State-1 and foreign banks demonstrate the largest, and quite similar, magnitudes of revals, while State-2 and private banks operate with the lowest revals. This difference can be explained by the intensity of cross-border operations: State-1 and foreign banks have large shares of operations with the rest of the world, while State-2 and private banks exhibit the opposite, on average. Irrespective of the crisis of 2008-2009, even large State-2 and private banks possess lower revals than small foreign or State-1 banks. These findings allow us to hypothesize that dropping revals from costs will most seriously affect the positions in the cost efficiency ranking held by State1 and foreign banks as compared to the other two groups of banks.

4. Results 4.1 Bank-level and group-level cost efficiency Table 4 presents SFA scores computed both at the bank and group levels for three distinct percentiles of respective distributions (25th, 50th, and 75th).4 These values reveal the scope of differences between less efficient (p25) and more efficient (p75) banks as average 4

The descriptive statistics of the variables included in the empirical cost functions are shown in Table A.2.

16

over the whole sample period (2005Q1-2013Q4) and within the sub-periods before and after the economic crisis of 2008-2009 to account for possible changes that may have occurred during the crisis. The average bank-level SFA scores computed over the whole sample period were 74.3%, 83.9%5 and 90.5 at 25th, 50th and 75th percentiles, respectively, implying that the Russian banking system was quite efficient in managing its operating costs. The same holds before and after the crisis of 2008-2009.

In terms of group-specific medians (p50) over the whole sample period, the leading position goes to private domestic banks (84.4%), followed by non-core state banks (82.6%), core state banks (78.4%), and foreign banks (62.9%). Below the group-specific medians (in p25), private banks are still the most cost efficient, but above these (in p75) they are overthrown by the national champion Sberbank from the State-1 group, whose SFA score is approaching the frontier (95.4%). Before the crisis, the leading positions were held by the State2 group on average, not private banks, that is, the ranking was different; among banks, Sberbank was the absolute leader. After the crisis, private banks became the most cost-efficient group, but Sberbank did not lose its absolute leadership in terms of efficiency. Our empirical result that foreign banks are the least efficient group among Russian market participants deserves some interpretation as it runs contrary to the mainstream literature on banking in transition (Bonin et al., 2005; Fries and Taci, 2006; Grigorian and Manole,

5

We cannot compare this estimate with those obtained in other studies on Russian banks because none of them exclude revals from operating costs as we do. In an earlier version of this paper, we showed that keeping revals in total costs decreases the median SFA score from 83.9% to 68.3% (Mamonov and Vernikov, 2015), which is comparable with other studies. Turk Ariss (2010) estimates Russian banks’ SFA score to be 83% on average. Kumbhakar and Peresetsky (2013) arrive at an estimated average SFA score of 81% when comparing the cost efficiency of banks in Russia and Kazakhstan. The period and the scope might explain these differences. Given that Russia is an emerging economy and Russian banks were still rather primitive in the mid-2000s, SFA scores above 80% appear on the high side, as they imply quite limited room for improvement in cost efficiency. Schaeck and Cihák (2014) estimate an average EU banking system SFA score of 88% for 1995-2005. Our estimated average SFA level of 68% with revals included looks much more credible.

17

2006; Fries et al., 2006; Karas et al., 2010). As shown theoretically by Mian (2006) and empirically by Lensink et al. (2008), substantial institutional differences between home and host countries (developed and transition economies in this case) can lead to a negative effect of foreign ownership on banking efficiency, resulting from additional costs borne by foreign banks as compared to domestic banks. We see two potential explanations of the reduced cost efficiency of foreign banks in Russia: (a) excessive capital adequacy with a relatively small loan portfolio of foreign banks in the initial period of penetration into the Russian market, impeding the exploitation of economies of scale; and (b) risk aversion of foreign banks in a volatile Russian market with poor protection of property rights. The validity of these explanations, however, must remain a topic of future research. Finally, we analyze the trajectories of group-specific averages of SFA scores in quarter-by-quarter format (see Figure 2).

We discover, first, that the spreads between the efficiency of different groups of banks appear relatively narrow. Keeping revals in the bank financial results would have blurred this effect (Mamonov and Vernikov, 2015). This finding is consistent with the notion that all players within a banking system are potentially exposed to the best available technology, so that the status of banks (state-controlled or private) does not preclude them from adopting best practices. Second, the quarterly evolution of bank groups’ average efficiency scores shows more clearly than in Table 1 that group rankings are unstable over the period considered. They may change from quarter to quarter, so that average efficiency scores over, say, 2 or 3 years may provide only rough information on the comparative performance. Third, bank group efficiency rises during a crisis period, which comports with the view that economic crises discipline economic agents by forcing them to eliminate unnecessary costs accumulated in previous periods. The highest observed increase in cost efficiency happened in the State-1

18

group, whereas the average efficiency score of private banks remained almost at the same level.

4.2 Distance functions: How does group ranking depend on bank-specific factors? In this section, we present the estimation results on the distance functions of Equation (4) obtained using 2-step GMM. The descriptive statistics of the variables employed in the regressions appear in Table A.2, and the underlying coefficients for the distance functions are presented in Table A. 3. Table A.4 contains three versions of Equation 3. In the first model (I), we simply regress the SFA scores on the group dummies without interacting them with other bank-specific factors as in previous studies. In the second model (II), we include all dummies and their interaction terms with the bank-level equity-to-assets ratios in order to measure bank risk tolerances. In the third model (III), we include all dummies and their interaction terms with the bank-level loans-to-assets ratios as a measure of asset structure.6

4.2.1 Homogenous relations: Previous research findings revisited Model I in Table A.4 represents the average ranking of our four bank groups over the sample period. The core state banks (State-1) are found to be slightly more cost efficient than private banks; the estimated coefficient before the respective dummy is positive and marginally significant. The estimated difference between the two groups is only 2.7 p.p. in terms of SFA scores. This result supports the findings of Karas et al. (2010) and suggests that core state banks are no less cost efficient than private banks in Russia.7

6

From a technical viewpoint, all three presented models satisfy the necessary requirements. The sets of instruments employed at the first stage of regressions are valid according to the Hansen test, as none of the P-values are below the 10% threshold. These sets of instruments are exogenous, as predicted by the Kleibergen-Paap LM statistics in the respective regressions (P-values are below the 1% level). We obtain quite large values for the centered R2, up to 56%. 7 These findings imply that even large observed differences in the average efficiency scores outlined in the previous section (SFA score at 50% for State-1 vs. 65% for Private when revals are kept, and 75% vs. 81% other-

19

The other state-controlled banks (State-2) seem to be more cost efficient than the private banks. The estimated difference of 1.7 p.p. in terms of SFA scores is very significant, but quite small (even smaller than the difference between State-1 and private banks discussed above). Foreign banks on average display no differences to private banks in terms of cost efficiency, i.e. the estimated coefficient is negative, but insignificant. These findings contradict those of Karas et al. (2010), who stress that foreign banks outperformed all other groups in Russia prior to the 2008-2009 crisis. 8

4.2.2 Heterogeneous relations based on differences in either risk preference or asset structure We now consider Models II and III in Table A.4. The interpretation of the respective coefficients provides little information on the ranking of banks, so we move to analyzing the underlying distance functions (Equation 4). For each bank group separately, we generate the distribution of values obtained from the respective distance functions, depending on the equity-to-assets (ETA) or loans-to-assets (LTA) ratios of banks, and then extract the 10th, 25th, 50th, 75th and 90th percentiles from these distributions for further analysis. The results are laid out in Table 5 (panels 1–2 for ETA and panels 2-4 for LTA). Risk preference. The core state banks (State-1) mainly show ETA ratios ranging from 8.8% (p10) to 21.2% (p90) over the sample period (see panel 2 of Table 5). Up to the 15.3% (p75) in this range, we see no statistical difference between State-1 and private banks in terms of cost efficiency (panel 1). Differences emerge and increase above p75; State-1 outperforms

wise) can disappear when we take into account internal specifics of these groups’ risk preference, asset structure, market powers, and other bank-level characteristics unrelated to costs. The core state banks possess greater market power than private banks (Anzoátegui et al., 2012), and are in fact no less cost-efficient than private banks despite formally lower efficiency scores. 8 Our findings may stem from the higher dependence of foreign banks on cross-border operations (mostly with their parent banks in their home countries), resulting in a period-average share of negative revals in total expenses of 58% (compared to just 23% for the other groups).

20

the other three groups. Specifically, a State-1 bank with an ETA ratio of 15.3% (21.2%) is 4.3 p.p. (6.7 p.p.) more cost efficient than the average private bank with an estimated SFA score of 84% (see Table 4). Foreign banks appear to be the least cost-efficient group only in case they operate with ETA ratios below the p50; that is, a foreign bank with an ETA ratio of 8.2% (11.1%), which corresponds to p10 (p25), is 2.5 p.p. (1.8 p.p.) less cost efficient than the average private bank. Decreasing leverage by increasing the ETA ratio above p50 allows foreign banks to outperform private banks and, above p75, even State-2 banks. For example, a foreign bank with an ETA ratio of 41.5% (p90) is 5.5 p.p. more cost efficient than the average private bank (although it makes little sense to maintain a sky-high ETA ratio 3.5 times above the banking sector average, and thereby substantially restrain profitability). The results for State-1 and foreign banks follow the prudent-efficient hypothesis of Berger and Mester (1997). Asset structure. The State-1 banks operate with one of the largest LTA ratios in the Russian banking system, while foreign banks are less specialized in allocating loans to the economy than all the other groups (panel 4 of Table 5). Within the p10–p90 distributions, the State-1 bank LTA ratios range from 36.8% to 71.2%, while foreign bank LTA ratios are only 6.4% in p10 and 70.7% in p90. Here our estimations of efficiency distance functions yield an important result: State-1 banks can become the most cost-efficient group in case they lend less to the economy, while foreign banks become the most cost-efficient group when they lend more to the economy (panel 3).

We suppose that the growing efficiency of foreign subsidiary banks as they lend more is logical. The economy of scale makes sense, especially if we examine traditional commercial banks geared towards lending and other core banking business. Our finding proves that the subsidiaries of foreign commercial banks, as opposed to other types of foreign-controlled

21

banking entities, are ‘normal’ commercial banks pursuing healthy business models. What is unusual is the decreasing efficiency of core state banks in dynamics. We do not interpret this as a depressing effect of loans on bank efficiency. We might actually be looking at banks pursuing different business models. For instance, an expansion of retail/consumer/mortgage lending might require additional costs reflecting investments in technology and infrastructure, at least for a certain period. On the other hand, for systemically important state banks a surge in policy lending might constrain the growth of profitability. This explanation is consistent with the growing literature on the political economy of finance predicting that government bank lending is inefficiently allocated (Coleman and Feler, 2015). Another possible explanation would be that, in the case of large state banks, a lesser share of commercial loans in assets corresponds to a larger than average share of financial instruments and other asset classes typical of investment banking that bring higher returns. That puts those banks at an advantage compared to others in terms of efficiency. Finally, state banks can be prone to corruption in the lending process in the form of kick-backs and/or related lending to a greater degree than peer banks. Another notable result comes from the comparison of State-1 and State-2 banks. As our estimations show, State-2 banks are more cost efficient than State-1 banks in the p50–p75 range of LTA ratios. This could reflect a lesser degree of political interference in bank decision-making. Unlike in the case of State-1 banks, the government does not compel State-2 banks to lend to government-approved projects.

4.3

Robustness checks We carried out several robustness checks. First, staying within the initially chosen cost

function with 3 outputs, we re-estimate our equations by replacing the production approach with the intermediation approach (Equation 1) and by applying a Tobit rather than a GMM

22

estimator to account for the censored nature of SFA scores that are always bounded between 0 and 100% (Equation 3). Second, we include additional outputs, namely securities, in our translog cost function (Equation 1) as the fourth output, given that this class of asset may be important for banks less geared to lending to the economy. Next, we include foreign assets (loans to and securities of non-residents) as the fifth output to account for the fact that Russian banks rely on this class of asset more than on loans to residents in periods of ruble instability. In each case, we additionally re-estimated the respective cost function on the post-crisis sub-period to account for possible effects of the 2008-2009 crisis in shifting the efficiency frontier. Third, we drop from Equation (3) all but one macroeconomic control, the GDP growth rate, to address possible multicollinearity concerns. Fourth, we return interest expenses into the evaluation of the cost frontier in order to understand whether our results on comparative efficiency are sensitive to accounting for market power differences between our four groups of banks. In Table A.5 in the Appendix we replicate Table 4 with SFA scores computed on the basis of total costs. Likewise, we replicate Figure 2 with the dynamics of SFA scores in all four ownership groups (Figure A.1). Finally, we omit the Lerner index from the list of inefficiency covariates and reestimate Equations (1) and (3) simultaneously, i.e. within the one-stage approach, taking into account bank-level fixed effects and replacing the (positive) half-normal distribution of the inefficiency term by the (positive) truncated-normal alternative. For this purpose, and similarly to Karas et al. (2010), we employ the model of Battese and Coelli (1995). In all these cases our baseline results remain qualitatively unchanged. 9

9

All these exercises, including the estimation results, are described in our working paper (Mamonov and Vernikov, 2015) and are available upon request. We do not spell them out here in order to save space.

23

5. Concluding remarks This paper introduced three amendments to the SFA computation of comparative bank efficiency in Russia. First, in order to avoid the distorting effect of the revaluations of all foreign currency items and securities on Russian banks’ balance sheets (revals), we deduct these two components from total costs when estimating the cost frontier. Second, we analyzed the performance of core state-controlled banks separately from other state-controlled banks. Third, within the group of foreign banks, we focused on those controlled by strategic foreign investors (i.e. subsidiaries of foreign banks). Fourth, we analyzed empirically how bank-level differences in risk preference and asset structure can affect rankings of both groups of statecontrolled banks, foreign and private banks in terms of cost efficiency scores. Our empirical results shed new light on the issue of comparative bank efficiency in Russia. First, bank efficiency improves during financial turmoil relative to normal circumstances. Second, foreign-controlled banks on average appear to be the least efficient market participants. Third, Russia’s core state-controlled banks led by Sberbank are more efficient than other state-controlled banks and nearly as efficient as private domestic banks during and after the 2008-2009 financial crisis. Fourth, foreign-controlled banks can become more cost efficient than others when they increase their equity-to-assets and loans-to-assets ratios above the sample median level. Core state-controlled banks, conversely, show superior cost efficiency when their loans-to-assets ratio falls below the sample median level. Some of our results are consistent with previous research (Karas et al., 2010), while others challenge the conventional wisdom with regard to the general level of Russian bank efficiency, the performance of foreign-controlled banks (Bonin et al., 2005; Fries and Taci, 2005; Grigorian and Manole, 2006) and bank behavior during crises. The most striking finding is the inferior efficiency performance of banks controlled by strategic foreign investors. This result may be attributable to institutional differences between Russia and the home countries of some foreign banks present in Russia (Mian, 2006; Lensink et al., 2008) as well as the

24

inability of foreign banks to take advantage of economies of scale. This issue requires further research. Another important finding is that Russia’s large state-controlled banks are not necessarily poor performers. These empirical findings might have research and policy implications. From a research perspective, this paper offers evidence that bank rankings in terms of efficiency may be misleading unless the effects of revaluation of foreign currency and securities are neutralized. Hopefully, subsequent estimations of comparative performance and efficiency estimations will use refined bank revenue data. From the policy perspective, our empirical results may motivate regulators to adjust industrial policy with regard to banks. The prejudice against state banks in favor of foreign banks should give way to a more balanced industrial policy aimed at a better performance of all national banks. On the other hand, even with this more enlightened approach, there may be less room for improvements in cost efficiency than is widely believed. In any case, we believe that the approach shown here is applicable to other dollarized emerging markets.

Acknowledgments We gratefully acknowledge, without implication, comments and suggestions from anonymous referees, Richard Frensch (the editor), Zuzana Fungáčová, Sergei Golovan, Ikka Korhonen, Israel Marques, Henry Penikas, Anatoly Peresetsky, Anna Pestova, Maria Semenova, Michael Skully, Vladimir Sokolov, Laura Solanko, Oleg Solntsev, and participants at the EBES conference in Barcelona (2014), the 16th International Academic Conference at the Higher School of Economics in Moscow (2015), the BOFIT research seminar at the Bank of Finland in Helsinki (2015), the 2015 RCEA Money and Finance Workshop in Rimini

25

(2015), the First World Congress of Comparative Economics in Rome (2015), the 5th International Banking Workshop “Banking in Emerging Markets: Challenges and Opportunities” in Moscow (HSE and BOFIT, 2015), and the XXIV International Rome Conference on Money, Banking and Finance (2015). Mikhail Mamonov acknowledges financial support received from HSE within the framework of the Basic Research Program 2016.

26

References Altunbas, Y., Evans, L., Molyneux, P. 2001. Bank ownership and efficiency. Journal of Money, Credit and Banking 33, 926-954. Battese, G., Coelli, T. 1995. A model for technical inefficiency effects in a stochastic frontier production for panel data. Empirical Economics 20, 325-332. Belotti, F., Daidone, S., Ilardi, G., Atella, V. 2013. Stochastic frontier analysis using Stata. STATA Journal 13, 719-758. Berger, A., DeYoung, R. 1997. Problem loans and cost efficiency in commercial banks. Journal of Banking and Finance 21, 849-870. Berger, A., Mester, L. 1997. Inside the black box: What explains differences in the efficiencies of financial institutions? Journal of Banking and Finance 21, 895-947. Berger, A., Hasan, I., Zhou, M. 2009. Bank ownership and efficiency in China: What will happen in the world’s largest nation? Journal of Banking and Finance 33, 113-130. Bertay, A., Demirgüç-Kunt, A., Huizinga, H. 2015. Bank ownership and credit over the business cycle: Is lending by state banks less procyclical? Journal of Banking and Finance 50, 326-339. Bonin, J., Hasan, I., Wachtel, P. 2005. Bank performance, efficiency and ownership in transition countries. Journal of Banking and Finance 29, 31-53. Brodsky, B. (1997). Dollarization and monetary policy in Russia. Review of Economics in Transition 6, 49-62. Caner, S., Kontorovich, V. 2004. Efficiency of the banking sector in the Russian Federation with international comparison. Economic Journal of the Higher School of Economics 8, 357-375. Сhernykh, L., Cole, R. 2011. Does deposit insurance improve financial intermediation? Evidence from the Russian experiment. Journal of Banking and Finance 35, 388-402. Claessens, S., van Horen, N. 2012. Being a foreigner among domestic banks: Asset or liability? Journal of Banking and Finance 36, 1276-1290. Claeys, S., Vander Vennet, R. 2008. Determinants of bank interest margins in Central and Eastern Europe: A comparison with the West. Economic Systems 32, 197–216. Coleman, N., Feler, L. (2015). Bank ownership, lending, and local economic performance during the 2008-2009 financial crisis. Journal of Monetary Economics 71, 50-66. Fiordelisi, F., Marques-Ibanez, D., Molyneux, P. 2011. Efficiency and risk in European banking. Journal of Banking and Finance 35, 1315-1326. Fortin, M., Leclerc, A. 2007. Should we abandon the intermediation approach for analyzing banking performance? GREDI Working Paper 07-01, Université de Sherbrooke. Fries, S., Taci, A. 2005. Cost efficiency of banks in transition: Evidence from 289 banks in 15 postcommunist countries. Journal of Banking and Finance 29, 55-81. Fries, S., Neven, D., Seabright, P., Taci, A. 2006. Market entry, privatisation and bank performance in transition. Economics of Transition 14, 579-610. Fungáčová, Z., Poghosyan, T. 2011. Determinants of bank interest margins in Russia: Does bank ownership matter? Economic Systems 35, 481–495. Golovan, S., Karminsky, A., Peresetsky, A. 2008. Cost efficiency of Russian banks. Models with risk factors. Ekonomika i matematicheskiye metody 44, 28-38. [in Russian]. Grigorian, D., Manole, V. 2006. Determinants of commercial bank performance in transition: An application of data envelopment analysis. Comparative Economic Studies 48, 497-522. Jiang, C., Yao, S., Feng, G. 2013. Bank ownership, privatization, and performance: Evidence from a transition country. Journal of Banking and Finance 37, 3364-3372. Karas, A., Schoors, K., Weill, L. 2010. Are private banks more efficient than public banks? Evidence from Russia. Economics of Transition 18, 209-244. Karas, A., Schoors, K. 2010. A guide to Russian bank data. SSRN Working Paper Series No. 1658468. Available at: http://ssrn.com/abstract=1658468 Karas, A., Pyle, W., Schoors, K. 2012. Deposit insurance, banking crises, and market discipline: Evidence from a natural experiment on deposit flows and rates. Journal of Money, Credit and Banking 45, 179-200. Kumbhakar, S., Peresetsky, A. 2013. Cost efficiency of Kazakhstan and Russian banks: Results from competing panel data models. Macroeconomics and Finance in Emerging Market Economies 6, 88113.

27

Lensink, R., Meesters, A., Naaborg, I. 2008. Bank efficiency and foreign ownership: Do good institutions matter? Journal of Banking and Finance 32, 834-844. Mamonov, M. 2013. State banks vs the private banking sector: Who is more efficient?, Bankovskoye delo 5, 23-30. [in Russian] Mamonov, M., Vernikov, A. 2015. Bank ownership and cost efficiency in Russia, revisited. BOFIT Discussion Paper 22/2015, Helsinki. Maudos, J., Fernández de Guevara, J. 2007. The cost of market power in banking: Social welfare loss vs. cost inefficiency. Journal of Banking and Finance 31, 2103-2125. Mian, A. 2006. Distance constraints: The limits of foreign lending in poor countries. Journal of Finance 61, 1465-1505. Schaeck, K., Cihák, M. 2014. Competition, efficiency, and stability in banking. Financial Management 43, 215-241. Schoors, K. 2000. A note on building a database on Russian banks: Fieldwork against the odds. PostCommunist Economies 12, 241-249. Solís, L., Maudos, J. 2008. The social costs of bank market power: Evidence from Mexico. Journal of Comparative Economics 36, 467-488. Solntsev, O., Pestova, A., Mamonov, M. 2010. Stress test: Will Russian banks need new government support? Problems of Economic Transition 54, 68-94. Styrin, K. 2005. What explains differences in efficiency across Russian banks? EERC Working Paper Series, Moscow. Sutela, P. (2013). Financial and credit markets. In Oxford Handbook of the Russian Economy (Alexeev M., Weber S., eds.). Turk Ariss, R. 2010. On the implications of market power in banking: Evidence from developing countries. Journal of Banking and Finance 34, 765-775. Vernikov, A. 2012. Impact of state-controlled banks on the Russian banking sector. Eurasian Geography and Economics 53, 250-266. Vernikov, A. 2014. ‘National champions’ in Russia’s banking services market. Problems of Economic Transition 57, 3-25. Vernikov, A. 2015. A guide to Russian bank data: Breaking down the sample of banks. SSRN Working Paper Series No 2600738. Available at: http://ssrn.com/abstract=2600738. Wang, H., Schmidt, P. 2002. One-step and two-step estimation of the effects of exogenous variables on technical efficiency levels. Journal of Productivity Analysis 18, 289-296.

28

80

45

70

40

Rubles per 1 USD (right axis)

60

35

Banking system average

30

5th percentile

50 40 25 30

25th percentile

20

20

50th percentile 15

0

10 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4

10

75th percentile 95th percentile

Figure 1. Negative revals (% of total costs) and the Ruble exchange rate to US dollar

29

2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4

Banks' SFA scores, % 90

100 - Efficiency frontier

80

70

60

50

40

Crisis

30

20

10

Private State-1 State-2

30 Foreign

Figure 2. SFA scores for different bank groups (arithmetic averages within each group; ranging from 0 for the least efficient to 100 for the most efficient)

Figure A.1. SFA scores for different bank groups when interest expenses are returned into the cost frontier

31

Table 1 Negative revals and their correlation with the nominal exchange rate Negative revals (% of total costs) Distribution Average 5th percentile 25th percentile 50th percentile 75th percentile 95th percentile

Average of 2005Q1–2013Q4 48.9 1.4 8.1 19.5 35.0 59.7

32

Correlation with the Ruble exchange rate to US dollar 0.81 0.50 0.68 0.77 0.81 0.84

Table 2 Breakdown of sample banks Period (end of) 2005 2006 2007 2008 2009 2010 2011 2012 2013

Core statecontrolled banks (State-1) No. %* 3 3 3 3 3 3 3 3 3

36.8 35.0 36.7 38.3 39.8 39.4 40.8 41.5 42.6

Other statecontrolled banks (State-2) No. %* 28 30 33 45 46 41 37 36 36

11.7 12.7 12.0 17.7 18.5 17.4 17.5 17.0 17.7

Domestic privately-owned banks (Private) No. %* 745 865 891 871 920 908 880 857 820

41.7 42.5 39.9 32.1 31.3 33.1 32.0 32.4 31.7

Notes: * Group share of total assets of the sample for the respective quarter

33

Foreign subsidiary banks (Foreign) No. %* 27 27 34 37 46 48 45 43 42

9.8 9.8 11.4 11.9 10.4 10.1 9.7 9.1 8.0

Total No.

%*

803 925 961 956 1015 1000 965 939 901

100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Table 3 Revals behavior with respect to bank size, ownership, and risk preference

Size (assets)

2007Q4 (before the crisis) Equity-toRevals, % of assets, % total costs (median) (median)

Panel 1. Banking system below median above median

2009Q4 (at the end of the crisis) Equity-to- Revals, % of assets, % total costs (median) (median)

2013Q4 (after the crisis) Equity-toRevals, % of assets, % total costs (median) (median)

17.8 11.2

4.2 11.0

26.5 13.3

23.0 44.4

23.2 10.9

10.1 30.7

State-1 below median above median

12.9 — 12.9

49.9 — 49.9

11.4 — 11.4

73.8 — 73.8

11.5 — 11.5

62.6 — 62.6

State-2 below median above median

9.8 11.2 9.7

4.8 4.2 5.6

11.7 22.2 10.7

33.6 20.0 38.6

10.8 12.3 10.7

22.2 2.4 29.2

Foreign below median above median Private below median above median

12.2 52.5 9.7 13.8 17.5 11.3

45.7 37.8 46.1 7.6 4.1 10.6

15.5 64.0 14.6 17.1 23.8 13.3

73.7 64.7 74.1 34.6 23.1 43.1

16.4 32.8 14.1 13.9 21.8 10.6

54.6 63.2 52.8 19.3 10.1 28.5

Panel 2. Bank groups

34

Table 4 Averages of bank- and group-level cost efficiency scores Whole period (2005Q1–2013Q4) Percentile

Before 2008–2009 crisis (2005Q1–2008Q2)

After 2008–2009 crisis (2010Q1–2013Q4)

p25

p50

p75

p25

p50

p75

p25

p50

p75

74.3

83.9

90.5

73.7

83.6

90.6

74.6

84. 1

90.3

State-1

65.3

78.4

95.4

59.8

71.0

92.1

66.6

State-2

72.4

82.6

89.5

76.8

84.2

89.6

69.6

Foreign

47.0

62.9

79.3

47.0

58.6

77.0

48.3

Private

75.3

84.4

90.7

74.5

83.8

90.8

75.9

Bank-level Group-level

35

79. 7 81. 8 65. 1 84. 6

95.9 90.0 81.0 90.5

Table 5 GMM post-estimation results: Distances between groups of banks in terms of cost efficiency (p.p. of SFA scores) determined on the basis of observable heterogeneity in risk preference or asset structure, averages for 2005Q1–2013Q4 Percentile a

p10

p25

p50

Panel 1: Distances as a function of risk preference (equity-to-assets ratios, ETA) –0.370 0.403 1.614 State-1 1.753*** 1.711*** 1.654*** State-2 –2.471*** –1.781** –0.709 Foreign Panel 2: Percentiles of ETA distributions within a particular group of banks 8.8 10.1 12.3 State-1 6.7 8.8 11.5 State-2 8.2 11.1 15.6 Foreign 8.2 11.0 16.5 Private Panel 3: Distances as a function of asset structure (loans-to-assets ratios, LTA) 6.140*** 4.454** 0.223 State-1 4.006*** 2.708*** 1.793*** State-2 –18.552*** –11.989*** –3.586*** Foreign Panel 4: Percentiles of LTA distributions within a particular group of banks 36.8 43.7 61.1 State-1 22.0 39.4 51.7 State-2 6.4 24.1 46.7 Foreign 23.3 39.4 54.8 Private

p75

p90

3.368** 1.514***

6.679*** 1.258***

1.493*

5.548***

15.3 18.1

21.2 30.3

24.7

41.5

27.1

44.3

–1.058

–2.247

1.072***

0.410

1.758**

5.316***

66.3 61.4

71.2 70.3

61.1

70.7

66.7

75.8

Notes: ***, ** and * indicate that an estimate is significant at the 1%, 5% and 10%, respectively. Robust standard errors are not provided for reasons of space. a

The number of bank-quarter observations employed to calculate the values of the distance functions are 108 for State-1 banks, 1,204 for State-2 banks, and 1,177 for foreign banks. The referent group (private banks) accounts for 26,624 observations.

36

Appendix Table A.1 Breakdown of income and costs of Russian commercial banks (% of total assets)

Total income

Before 2007Q4

2008–2009 crisis During After 2009Q4 2011Q4 2013Q4

40.7

105.4

65.7

53.9

Interest income

6.9

9.3

6.7

7.7

Income from operations with securities

2.7

2.7

1.4

2.5

Positive securities revaluation

1.0

0.3

0.4

0.1

Income from operations in foreign currency Income from positive revaluation of assets and negative revaluation of liabilities both denominated in foreign currency Fee and commission income

15.0

76.9

43.3

30.9

11.7

68.4

37.5

26.8

1.8

1.6

1.4

1.4

Income from decreasing of loan loss provisions (+LLP)

10.7

12.2

9.6

8.4

Other income

3.6

2.8

3.5

2.9

Total costs

38.4

104.9

64.1

52.5

Interest expenses

3.2

5.1

3.1

3.8

Expenses due to operations with securities

2.1

2.1

1.3

2.4

Negative securities revaluation

0.4

0.2

0.5

0.1

Expenses due to operations in foreign currency Expenses due to negative revaluation of assets and positive revaluation of liabilities both denominated in foreign currency Fee & commission expenses

14.8

76.7

43.1

30.8

11.8

68.5

37.5

26.7

0.2

0.2

0.2

0.3

Expenses from raising loan loss provisions (–LLP)

11.5

15.4

9.8

9.5

Personnel expenses

1.9

1.5

1.6

1.5

Other expenses

4.8

4.0

5.0

4.3

Profit (after LLP and taxation)

2.3

0.4

1.7

1.4

Net interest income

3.4

3.7

3.0

3.4

Net income from operations with securities

0.9

1.2

0.7

0.6

Net securities revaluation

0.7

0.5

0.0

0.0

Net income from operations in foreign currency

0.2

0.2

0.2

0.2

Net foreign currency revaluation

–0.1

–0.1

0.1

0.1

Net fee & commission income

1.7

1.4

1.2

1.1

Net income from decreasing of loan loss provisions

–0.8

–3.3

–0.3

–1.1

Personnel expenses (with “–” sign)

–1.9

–1.5

–1.6

–1.5

Net other income

–1.2

–1.2

–1.6

–1.4

Net foreign currency position

0.3

Assets in foreign currency

29.8

4.0 35.2

2.9 30.3

0.9 22.1

Liabilities in foreign currency 29.5 31.2 27.4 21.2 Source: Own calculations based on CBR database on bank balance sheets and profit-and-loss statements.

37

Table A.2 Descriptive statistics of the variables in the cost function (2005Q1–2013Q4) Unit

Symbol

Mean

Std dev

Min

Max

Obs

Banks

Dependent variables Total costs minus interest expenses minus revals* Explanatory variables

RUB bln

𝑂𝐶𝑖𝑡

7.7

69.8

0.0

2904.0

30784

1196

Retail and corporate loans

RUB bln

𝑌1,𝑖𝑡

18.2

206.7

0.0

10015.4

30045

1159

Retail and corporate accounts and deposits

RUB bln

𝑌2,𝑖𝑡

16.6

205.1

0.0

10374.8

30635

1191

Fee and commission income

RUB bln

𝑌3,𝑖𝑡

0.5

5.0

0.0

220.6

30635

1189

Average funding rate

%

𝑃1,𝑖𝑡

4.9

2.8

0.0

50.1

29365

1152

Price for personnel expense

%

𝑃2,𝑖𝑡

4.1

3.3

0.1

49.5

30784

1196

Price of physical capital

%

𝑃3,𝑖𝑡

23.7

22.4

0.2

180.0

30784

1196

RUB bln

𝐸𝑄𝑖𝑡

3.8

40.8

0.0

1954.2

30745

1196

Equity capital

38

Table A.3 Descriptive statistics of the variables in the cost efficiency equations (2005Q1–2013Q4), % Mean

Std dev

Min

Max

Obs

Banks

Equity-to-assets ratio

18.6

12.0

1.9

79.8

22629

1038

Loans-to-assets ratio Loans-to-deposits ratio

55.1 107.3 31.8 0.1

16.7 83.0 23.5 1.1

10.0 10.5 0.0 0.0

96.0 996.0 100.0 31.6

22629 22629 22629 22629

1038 1038 1038 1038

17.0

37.3

–96.8

94.9

22316

1033

3-month ruble volatility GDP (annual growth rate)

0.6 3.4

0.5 4.9

0.1 –11.2

2.2 8.6

36 36

Real households income (annual growth rate)

6.1

4.9

–4.9

15.4

36

Firms’ profit-to-debt ratio

4.7

2.1

–1.7

10.4

36

Bank-specific factors*

Share of retail loans in total loans Bank size (in terms of assets) Funding- and efficiency-adjusted Lerner index of market power** Macroeconomic controls

Notes: * All values reported in this table were obtained through applying filtering procedures as described in Section 3.3. ** Negative values of the Lerner index for some banks in our sample may reflect either a cross-subsidy strategy to gain a higher market share (Solís and Maudos, 2008) or greater inefficiency of small banks, i.e. players who cannot dictate prices for their loans and suffer from negative margins. Notably, as shown by Solís and Maudos (2008), the Lerner index averaged across all banks operating within the market for loans in the Mexican banking system achieved huge negative values, including –0.41 in 2003. The cross-subsiding strategy implies covering such negative margins in one market with positive margins in others. For example, in 2003, the Lerner index for the deposits market in Mexico was about +0.48. We assume similar patterns are observable within the Russian banking market.

39

Table A.4 GMM estimation results: The determinants of within- and between-group heterogeneity of cost efficiency, 2005Q1-2013Q4 (dependent variable is bank-level SFA score) I

II

III

State-1

2.704* (1.406)

–5.365** (2.648)

15.123*** (4.984)

State-2

1.672*** (0.241)

1.895*** (0.572)

5.643*** (1.223)

Foreign

–0.021 (0.693)

–4.456*** (1.159)

–20.925*** (1.892)

0.426*** (0.011)

0.417*** (0.011)

0.419*** (0.011)

Dummy variables for bank ownership status

Bank-specific factors Equity-to-assets ratio (ETA) ETA × State-1

0.569*** (0.166)

ETA × State-2

–0.021 (0.035)

ETA × Foreign

0.241*** (0.059) 0.439*** (0.008)

Loans-to-assets ratio (LTA)

0.439*** (0.008)

0.428*** (0.008)

LTA × State-1

–0.244*** (0.085)

LTA × State-2

–0.074*** (0.022)

LTA × Foreign

0.371*** (0.037)

Loans-to-deposits ratio

–0.106*** (0.003)

–0.106*** (0.004)

–0.101*** (0.004)

Share of retail loans in total loans

0.009*** (0.003)

0.011*** (0.003)

0.009*** (0.003)

Bank size

0.483*** (0.064)

0.513*** (0.063)

0.533*** (0.066)

Funding- and efficiency-adjusted Lerner index of market power a

0.012*** (0.003)

0.012*** (0.003)

0.013*** (0.003)

0.202 (0.160)

0.205 (0.159)

0.177 (0.158)

GDP (annual growth rate)

0.053*** (0.015)

0.053*** (0.014)

0.050*** (0.014)

Real households income (annual growth rate)

–0.066*** (0.017)

–0.066*** (0.017)

–0.059*** (0.017)

0.028 (0.046)

0.029 (0.046)

0.018 (0.049)

61.283*** (0.396) 19573 (967) 0.557 6, 12 0.143 0.000

61.361*** (0.395) 19573 (967) 0.559 9, 15 0.221 0.000

61.552*** (0.412) 20319 (978) 0.549 9, 15 0.167 0.000

Macroeconomic factors 3-month ruble volatility

Firms’ profit-to-debt ratio Intercept No. of observations (No. of banks) Centered R2 No. of endog. variables, excl. instr. P-value for Hansen J-stat P-value for Kleibergen-Paap LM stat

Notes: ***, ** and * indicate that an estimate is significant at the 1%, 5% and 10% level, respectively. Robust standard errors are provided in parentheses under the coefficients. Privately owned domestic banks are the referent group. SFA scores are defined within the production approach.

40

a

Cumulative effect of four quarters (0, –1, –2, and –3).

41

Table A.5 Averages of bank- and group-level cost efficiency scores when interest expenses are returned into the cost frontier Whole period (2005Q1-2013Q4)

Before 2008-2009 crisis (2005Q1-2008Q2)

p25

p50

p75

p25

p50

p75

p25

p50

p75

73.5

82.8

89.7

72.6

82.3

89.9

74.2

83. 2

89.7

State-1

60.1

77.6

94.9

58.5

70.3

91.6

61.9

State-2

71.1

81.7

89.1

74.0

83.2

89.3

68.3

Foreign

44.0

59.7

77.3

42.8

55.9

73.7

45.5

Private

74.6

83.4

90.0

73.5

82.8

90.1

75.4

Percentile Bank-level

After 2008-2009 crisis (2010Q1-2013Q4)

Group-level

42

79. 2 81. 2 63. 2 83. 8

94.9 89.8 79.7 89.9