Banking efficiency and Managerial behavior: Evidence from Central and Eastern European Banks

Transcription

1 Banking efficiency and Managerial behavior: Evidence from Central and Eastern European Banks Published in Kredit und Kapital, N.4, 2008 Stefania Rossi Markus S. Schwaiger Gerhard Winkler 1

2 Motivations of the research Studies on bank efficiency date back to the 1960s. yet they remain an appealing research area from both, a macro and a micro economic standpoint. From the micro point of view, achieving a higher level of efficiency becomes crucial nowadays due to the squeeze experienced in interest margins. From the macro point of view, bank efficiency represents a socially optimal target, since it reduces the costs of financial intermediation. 2

3 As regards academic research, efficiency analysis is an appealing theme, which has evolved over the last decades, in two separate streams: non-parametric frontier analysis (Data Envelopment Analysis): a linear programming technique which approximates an efficient frontier in a nonparametric way. parametric frontier analysis: an econometric estimation of a cost or production frontier. 3

4 The purpose of this study is to shed light on the efficiency of the banking sectors in countries that have recently acceded (10 countries) or are in the process of accession to the European Union (excluding Turkey). We will refer to these countries as Central and Eastern European Countries (CEECs), without Malta and Cyprus and with the Baltic states subsumed in this expression. 4

5 The literature on banking has so far mainly focused on the efficiency of OECDs countries. Berglöf and Bolton, 2002; Bonin, Hasan and Wachtel, 2003; Buch, 2000; Fries and Taci, 2002 and 2003; Fries, Neven and Seabright, 2002; Weill,

6 Characteristics and the recent developments into the banking markets of CEECs The CEE banking market is characterized by: An increased competition among financial institutions, due to: 1. institutional upgrading in all economic sectors after the collapse of the socialist regimes; 2. the preparation of new member countries for accessing European Monetary Union; 3. the privatization and concentration process occurred in the financial market; 6

7 a relatively high degree of concentration high degree of profitability, while the system is at the same time well equipped with equity (as shown in capital adequacy ratios (CARs) widespread presence of foreign ownership that has emerged over the last years. 7

8 Banking activity in CEECs Oest er r eichische Nat ional bank 8

9 Table 3: Characteristics of the banking sector in selected CEE countries 9

10 As a results we have a two-tier banking system: (a) a tier of well capitalized well managed foreign owned private financial institutions; (b) a second tier of banks still in governments hands and saddled with an overhang of non-performing loans. 10

11 Developement of Banking Sectors Sharply expanding total assets Rising equity ratio Return on equity recently rising/stable Interest rate margins decrease Loan- and deposit-ratios on the rise 11

12 Aims and Scope The purpose of this analysis is to overcome the shortcomings of the literature and to analyze the effect of the consolidation process on the banking system of the accession economies. 12

13 The main innovations of our paper are: Providing panel data evidence over the period on cost efficiency for a large sample of banks in CEECs; using two different cost functions (TL and Fourier); giving explanations in terms of the factors underlying differing efficiency scores across banking institutions. 13

14 Dataset consists of an unbalanced panel of balance sheet and income data taken from BankScope Database (BvD, Fitch/Ibca) for 9 Central and Eastern European Countries (CZ, EST, HU, LV, LT, RO, SI, SK, PL) from the years 1995 to

15 Table 4: Distribution of banks across countries Countries Number of banks Bulgaria 0 Cyprus 19 Czech Republic 44 Estonia 14 Hungary 35 Latvia 28 Lithuania 12 Malta 10 Poland 72 Romania 34 Slovenia 21 Slovakia 27 TOTAL

16 Dataset characteristics of our sample 16

17 Some methodological remarks Oest er r eichische Nat ional bank A production plan is called efficient if it is not possible to produce more using the same input, or to reduce these inputs leaving the output unchanged. The duality theory (Diewert 1974; Shephard 1970) has shown that, under given conditions the property of the production function can be studies indirectly, through cost or profit functions. C= C (y, p) gives an indirect representation of the feasible technology, since it indicates the minimum cost of producing an output vector y, given a price vector p in the input markets. 17

18 Some methodological remarks So far the theory.! Oest er r eichische Nat ional bank Observable production plans and cost level do not follow from perfectly rational and efficient decisions: factors as errors, lags between the choice of the plan and its implementation, inertia in the human behavior, distorted communications, may cause the so-called X-inefficiency (Leibenstein 1966; Leibenstein, 1978) to drive real data away from the optimum production plan. 18

19 Some methodological remarks Oest er r eichische Nat ional bank The stochastic frontier models (Aigner et al. 1977; Meusen and Van der Broek, 1977; Stevenson, 1980) allows to estimate a minimum cost frontier including a composite disturbance, which in turn can be subdivided in two parts: beside a one-sided component (u) representing inefficiency as in the deterministic model; (v) accounting for the noise in the data, and usually distributed as a normal. 19

20 20 Oest er r eichische Nat ional bank Some methodological remarks Therefore for the s-th producer what we observe empirically is the following cost function: s s s s s s s s v u p y C p y C C ), ( ), (

21 Variable measurement and definition of input and output In modeling the bank cost function one of the most debated questions regards the definition of the inputs and outputs of the multi-product financial firms. The discussion concentrates particularly on two issues: how input and output have to be measured the role of deposits 21

22 Variable measurement and definition of input and output The literature faces these issues according different approaches: The production approach The intermediation approach The modified production approach We follow the latter approach: the interests paid on deposits are accounted as input, while the volume of deposits are considered as output. (see table 6) 22

23 Cost function specifications and estimation methodology In modeling the cost function we use the different functional forms: the Translog cost function (TL); the Fourier Flexible form (FF). 23

24 Cost function specifications and estimation methodology The s-th firm total cost TL function can be written as follows: ln( TC) k 1h 1 st kh [ ln y ln 0 p ks 3 i 1 ln i p hs is 3 i 1 k 1 k 1 ln p 3 3 ik k ln y is ks ln p 1 2 ks 3 3 ln y i 1 j 1 st ij ] D V st is ln y U st js 24

25 Cost function specifications and estimation methodology The s-th firm total cost Fourier can be written as follows: ln( TC) k 1h 1 st kh [ ln y ln 0 p ks 3 i 1 ln i p hs is 3 i 1 k 1 k 1 ln p 3 3 ik k ln y is ks ln p 1 2 ks 3 3 ln y i 1 j 1 st ij ] D V st is ln y U st js i a i cos(y i ) + i b i sin(y i ) + k c k cos(p k ) + k d k sin(p k ) + ij e ij [cos(y i ) + cos(y j )] + ij f ij [sin(y i ) + sin(y j )] + ij g ij [cos(y i ) - cos(y j )] + ij h ij [sin(y i ) - sin(y j )] + kl i kl [cos(p k ) + cos(p l )] + kl l kl [sin(p k ) + sin(p l )] + kl m kl [cos(p k ) - cos(p l )] + kl n kl [sin(p k ) - sin(p l )] ( 25

26 Cost function specifications and estimation methodology Regressions (2) and (3) are estimated applying the stochastic frontier approach (Battese and Coelli, 1992); In this model, the measure of cost efficiency may be written as: EFF st E( TCst Ust, st )/ E( TCst Ust 0, st ) where phi are the regression parameters. 26

27 Cost function specifications and estimation methodology The inefficiency term, Ust can be expressed as an explicit function of a vector of variables Zst which may influence the firm s efficiency: U st 0 n 1 dsize l 1 dot is is js js i 1 j 1 b 1 p bs BalI bs st 27

28 Empirical findings: stochastic cost function Table 7 All the outputs and input price coefficients present the expected positive sign; The elasticity of production costs to unit staff (bp1=0.62) is larger than the elasticity to the capital price; Among outputs, loans (ay2= 0.37) are more cost absorbing than deposits and other earning assets. 28

29 Empirical findings: Efficiency scores by country Tables: 8-11 On average banks operate far from the efficient frontier. Lithuania has the most efficient banking system of all the countries selected, followed by Slovenia, Poland and Romania. The worst performance is instead achieved by Slovakia and the Czech Republic. There is a slight tendency of efficiency to increase over time when we use control variables in the cost function specification. 29

30 Empirical findings: determinants of inefficiency Table: 12 Small and medium banks are more efficient than large banks. The variable non-interest expenses positively affects the efficiency of banks. Higher problems loans are connected with a lower level of efficiency. Commercial banks are the most efficient banks 30

31 Empirical findings: determinants of inefficiency XEFFst= + 1LLPst + 2CAPst + 3LTAst + 4STDEVROAs + + 5SMALLBANKst + 6LARGEBANKst + 7MARKETSHAREst + 8S1s + 9S2s + 10S3s + 11S4s + 12S5s + 13S6s + 14S7s GDPCAPt + 16CONCENTRATIONt + 17OWNERSHIPt + 18TRENDt + εst (4) where εst stands for the error term and all the other variables are defined above. 31

32 Empirical findings: determinants of inefficiency Small and medium banks are more efficient than large banks. The variable non-interest expenses positively affects the efficiency of banks. Higher problems loans are connected with a lower level of efficiency. Commercial banks are the most efficient banks 32

33 Modelling the management behaviour Four different hypotheses concerning different kinds of management behaviour can be formulated: a) the bad management hypothesis implies that bad managers do not adequately control for operating expenses and poorly manage their loan portfolio: low efficiency should cause an increase in problem loans. b) The skimping hypothesis suggests that the quality of a bank s loan portfolio is determined by the costs associated with the monitoring of lending activities: higher levels of cost efficiency therefore cause higher levels of problem loans. 33

34 Modelling the management behaviour c) The bad luck hypothesis suggests that external factors increase problem loans which in turn cause a decline in cost efficiency as monitoring costs are increased in response. d) The moral hazard hypothesis suggests that managers of weakly capitalized banks are less risk adverse and engage in risk taking behaviour. Therefore, low bank capitalization causes an increase in problem loans. 34

35 Modelling the management behaviour These hypotheses are captured in the following equations: LLPst = f1(llp s lag, XEFF s lag,cap s lag, LTA s lag, Trend t)+ e 1s t (5) XEFFst = f2 (LLP s lag, XEFF s lag,cap s lag, LTA s lag, Trend t)+ e 2s t (6) CAPst = f 3 (LLP s lag, XEFF s lag,cap s lag, LTA s lag, Trendt)+ e 3s t (7) 35

36 Modelling the management behaviour Equation (5) tests the bad management hypothesis (we would expect a negative sign of the lagged X-efficiency variables), as well as the skimping hypothesis (we would expect a positive sign of the lagged X-efficiency variables). Equation (5) also tests the moral hazard hypothesis when only a sub-sample of low capitalized banks are used. In this case, a negative sign is expected between the bad loan and the lagged value of the capitalization variables (CAP). 36

37 Modelling the management behaviour Equation (6) tests the bad luck hypothesis: a negative relationship is expected between X- efficiency and the lagged values of problem loans. As in Berger and DeYoung (1997) and Williams (2004) equation (7) is included to complete the model and to allow us to obtain further information on managerial behaviour. 37

38 Empirical evidence: Cost and profit estimates The evidence emerging from Table 4 can be summarized as follows: (a) The banking systems of CEECs present, as expected, significant levels of cost and profit inefficiency, indicating that on average banks operate far above (below) the cost (profit) efficient frontiers. Compared to studies available for banks in Western European countries and the U.S. (Berger and Humphrey, 1997), banks in CEECs display higher variation as well as comparably lower levels of efficiency (see Weill, 2004). 38

39 Empirical evidence: Cost and profit estimates (b) Efficiency levels vary considerably across countries, as is revealed by inspecting the average efficiency scores by country. These differences in cost and profit efficiency levels are significant in almost all cases. As far as cost efficiency is concerned, values range from 0.58 (the Czech Republic) to 0.89 (Slovenia). For profit efficiency, values range from 0.29 (Romania) to 0.57 (the Czech Republic). 39

40 Empirical evidence: Cost and profit estimates (c) Overall, the profit efficiency levels of the CEE banks are well below cost efficiency levels. This result is consistent with most other empirical papers estimating both cost and profit functions for the Western banking sectors (Berger and Humphrey, 1997; Maudos et al., 2002). 40

41 Empirical evidence: Cost and profit estimates (d) Additionally, Table 4 shows that the profit efficiency ranking across countries is rather different from the cost efficiency ranking: countries that present fairly high cost efficiency scores seem to be performing less well in terms of profit efficiency, and vice versa. This negative correlation between cost and profit efficiencies is supported by other empirical studies, such as those for US banks. 41

42 Empirical evidence: Cost and profit estimates Looking at the evolution of efficiency we find a positive and significant increase over time of cost and profit efficiency for the overall sample of banks, equal to 6 per cent and 4 per cent respectively. Czech Republic and Slovakia are the countries with the worst performance in terms of cost efficiency, and both have a high burden of non performing loans affecting the quality of bank s loan portfolios. On the other hand, Slovenia, Estonia and Lithuania, leading countries in terms of cost efficiency, present relatively low levels of loan loss provisions, 42

43 Empirical evidence: Possible determinants of cost and profit efficiency Our results on the determinants of efficiency show a negative, albeit small, relationship between loan loss provisions LLP (as a proxy for problem loans) and the cost efficiency level, indicating that higher problem loans are connected with lower efficiency levels. On the profit side we do not find a significant impact of the variable LLP on profit efficiency. (MARKETSHARE), which is a measure of bank s size, has a positive effect on cost efficiency. 43

44 Empirical evidence: Possible determinants of cost and profit efficiency The variable SPECIALIZATION turns out to be significant using a joint F-test in the case of cost efficiency, but it does not appear relevant in an analysis of profit efficiency. The degree of concentration (CONCENTRATION) have a systematic and significant impact on efficiency. We find a negative relation between cost efficiency and GDP per capita (GDPCAP) - which is a proxy for the demand and supply of banking services and a positive correlation of GDPCAP with profit efficiency. 44

45 Empirical evidence: Possible determinants of cost and profit efficiency Regarding the impact of foreign bank penetration (OWNERSHIP) on efficiency we observe a negative correlation between foreign owned bank assets and cost efficiency and a positive and significant correlation with profit efficiency. To sum up the high variation in efficiency can be explained by bank features, market characteristics and general environmental conditions. 45

46 Empirical evidence: Management behaviour results The correlation we found between bad loans and cost efficiency opens up a crucial question: is this negative correlation a sign of bad management or bad luck? In order to address this question we use the managerial behavior equations with Granger causality tests. 46

47 Empirical evidence: Management behaviour results We employ the model proposed by Berger and DeYoung (1997) and used in Williams (2004) based on a Granger causality approach for testing the different management hypotheses (see equations 5-7). The three equations of the model were estimated separately using the Arellano-Bond dynamic panel data model. 47

48 Empirical evidence: Management behaviour results The only hypothesis that is supported by our analysis, both for cost and profit efficiency, is the bad luck hypothesis. This is the case since only equation (6) results in significant coefficients of the expected sign. The coefficients on the trend terms are again positive and significant indicating the increase in efficiency over time as already discussed. 48

49 Policy implications and conclusions In terms of regulatory policy implications the evidence for the bad luck hypothesis is a signal that bank inefficiency and failures in these markets are inter alia associated with external shocks beyond the control of management. Although there is a general trend in terms of improving efficiency scores across the region, regulatory and supervisory rules should therefore focus on reducing banks exposure to these unforeseen events. This could for instance be done by increasing the diversification of loan portfolios through limits on loan concentration or by promoting mergers with foreign institutions. 49

50 Policy implications and conclusions This conclusion suggests that for these countries the target should be to reduce the exposure of banks to external shocks. Foreign bank penetration and loan diversification may help lower the risk of financial crises by reducing the potential negative effects deriving from the adverse business cycle. Foreign strategic capital may also strengthen the banking system and improve the level of financial intermediation. 50

51 Policy implications and conclusions Moreover, foreign bank penetration may help increase competition and international integration which may be beneficial by stabilizing the credit base for these countries. The effect of having joined the European Union will be crucial for these countries which can benefit from a more competitive and stable market and a more favourable macroeconomic environment. 51

52 Policy implications and conclusions Future research on bank managerial behaviour in these countries should be able to take advantage of better data quality on bank risk arising from the introduction of the Basel II accord and common reporting standards (IFRS) for these countries. Furthermore, it should be verified, once a long enough time series exists, whether improvements in banking efficiency were a lasting development and whether the dynamics of banks managerial behaviour has changed since EU membership. 52