Local bank competition and small business lending after the onset of the financial crisis

Transcription

1 Local bank competition and small business lending after the onset of the financial crisis Jaakko Sääskilahti 1 The Jyväskylä University School of Business and Economics Abstract July 2014 This paper examines the effects of the financial crisis on small business lending. In particular I examine whether the magnitude of the changes in volumes and prices of small business loans is dependent on pre-crisis local bank competition and market structure. I use a unique data set of Finnish cooperative banks and the locations of all branches of banks in Finland between 2004 and The data enable to examine the effect of competition environment based on the homogeneous set of the banks regarding culture and business model, possibility to control uniformly other affecting factors, and significant differences in competition environments. I find that the monthly volumes of new business loans decreased and the average loan margins increased during the crisis period. More important, I find evidence of the larger decrease in volumes and increase in margins in the banks operating in the more competitive local markets. The results are more robust for the prices than for the volumes. Keywords: financial crisis, small business lending, local bank competition JEL classification: G01, G21, L11 1 I thank OP-Pohjola group and Toni Honkaniemi, Arto Kuhmonen, Kirsi Saari and Ari Saarinen for providing the bank data. I thank Manuel Bagues, Ari Hyytinen, Juha Junttila, Panu Kalmi, Kari Kemppainen and Jaakko Pehkonen, participants at the Allecon seminar (Jyväskylä) and the Finnish Economic Association XXXVI Annual Meeting (Kuopio) for valuable comments. Financial support from the OP-Pohjola Group s Research Foundation is gratefully acknowledged. address: jaakko.s.saaskilahti@student.jyu.fi Tel.:

2 1 1 INTRODUCTION The availability and cost of bank loans is crucial for many small businesses because they do not often have other possibilities for external funding (Carbó-Valverde et al. 2009). Small business lending has a local nature due to informational issues (Petersen & Rajan 2002) and thus local small businesses are often exposed to the behavior of competing local banks or branches. When uncertainty is high especially during crises, the loan availability for opaque small businesses is often reduced the most. Changes in bank driven effects on lending i.e, in loan supply during crises are mostly explained by bank balance sheet factors (e.g. Bernanke & Lown 1991; Jimenez et al. 2012b). Studies concerning the availability and costs of bank loans during the current crisis focus especially on the significant role of bank funding structure (Ivashina & Scharfstein 2010; Cornett et al. 2011) and additionally examine the effects of loan losses (Puri et al. 2011; Santos 2011) and more indirect factors like changes in risk tolerance (Albertazzi & Marchetti 2010). Market structure and bank competition can affect loan availability and prices (e.g. Hannan 1997; Carbo-Valverde et al. 2009; Rice & Strahan 2010; Chong et al. 2013). As changes in competition during the crisis (see Ruckes 2004) 2 and characteristics relevant for crisis time lending like risk-taking, relationship banking and funding structure can be dependent on competition environment (see e.g. Keeley 1990; Boyd & Nicolo 2005; Petersen & Rajan 1995; Graig and Dinger 2013), this imply that different competition environments can also 2 Besides this theoretical consideration, for example, The ECB Bank Lending surveys point that the reduced pressure of competition was one reason for tightened lending standards during the current crisis.

3 2 explain heterogeneous changes in lending behavior during crises which has not considered in previous literature. In this paper, I examine first how market structure and pre-crisis competition of local banks are associated with heterogeneous changes in the volumes and prices of small business loans during the crisis. Second, I examine whether the relationship between competition environment and changes in lending during the crisis is due to different changes in competition and/or various characteristics of banks depending on local market structure and competition. I employ unique data of the Finnish local cooperative banks containing monthly information about the volumes and average margins of new small business loans during the period 2004 through I combine local economic characteristics and number of branches of cooperative and other banks for each individual cooperative bank. My empirical identification is based on the homogeneous cooperative banks regarding common culture and consistent information but heterogeneous local competition environments. The data provide fully comparable and detailed lending information and various control variables affecting loan demand and supply. As the measurement of bank competition is not unambiguous, I use two kind of measures. Herfindahl-Hirschmann index (HHI) is a market structure measure based on the number of branches of various banks in the operating area of each cooperative bank. The Lerner index is a more direct competition measure widely used in the banking literature (e.g. Carbo et al. 2009). My results indicate a relevant role of competition environment in changes in lending during the crisis. First, I find a decrease in the volumes of new withdrawn small business loans relative to the total assets and an increase in the margins of new business loans after the

4 3 onset of the financial crisis. More important, I find larger changes in the most competitive local bank markets 3 and the results are more robust for prices. Second, I find that the averages of competition measures and various types of deposit and loan prices converged between the most competitive and other banks. Additionally, funding structures are quite different in the most competitive and other banks which tentatively suggest also this indirect effect of competition environment. Instead, risks and changes in risk taking do not seem to explain the results. Overall, the results suggest that competition environment explains the heterogeneous effects of the crisis on small business lending at least when basic problems observed in the previous literature, particularly bad funding conditions and large loan losses, are not very large. More generally, the findings provide tentative support for the effect of competition on pro-cyclical lending which is relevant for bank regulation and supervision. Moreover, as the heterogeneous effects of crisis on various local banks or bank markets can cause heterogeneous effects of crisis on local economies (see Gozzi & Goetz 2010), local competition environment may have a role in this. 4 In the next section, I review previous theoretical and empirical literature to lay the foundations for my paper. Section 3 describes the data and empirical approach. The results of the main empirical analysis are presented in section 4. I make auxiliary analysis and discuss the results in section 5 and conclude in section 6. 3 In my data operating area of each bank forms a single local market and thus competitiveness of local bank and local market are used as synonyms. 4 Adams and Amel (2011) study how local market structure affect the pass-through of the federal funds rate and point the importance to understand heterogeneous effects in various areas.

5 2 COMPETITION ENVIRONMENT AND LENDING IN NOR- MAL AND CRISIS TIMES 4 In this chapter I consider based on previous theoretical and empirical research various possible channels through which competition environment can affect variations in loan availability and prices between good and crisis times. Competition environment can affect, firstly, the behavior of banks and its changes in crisis times and secondly, bank-specific characteristics that affect lending differently in good and crisis times. Prices and other lending standards The market power hypothesis suggests that competition lowers loan prices and thus improves availability (e.g. Carbo-Valverde et al. 2009). Many empirical papers support this hypothesis. Hannan (1997) examines US small business loans at the MSA level and finds that under more competition, banks lower loan rates. Corvoisier and Gropp (2002) use country level European data and find a positive relationship between concentration and loan prices. Degryse and Ongena (2005) find that firm s distance to competing banks is positively related to its loan rates. Petersen and Rajan (1995) find a little contrary to the hypothesis that higher concentration i.e., lower competition decreases loan rates for young and small firms in the U.S. but the effect diminishes as firms get older. Ruckes (2004) describes theoretically how price competition varies between recessions and booms but, to my knowledge, there is no empirical analyses how price changes during crises are related to changes in competition. Lending standards i.e., the strictness of various loan terms is one way to explore changes in loan availability. The ECB Bank Lending survey asks lending standards from banks and divides the relevant factors affecting them into the three groups: competition, balance sheet position and risk perception (Berg et al. 2005). The lending standards of this survey include

6 5 prices, collateral requirements, maturity, size, covenants and loan-to-value ratio of loans. Empirical studies find that changes in lending standards based on surveys 5 explain changes in credit growth and loan availability (Lown and Morgan 2006; Demiroglu et al. 2012) The relationship between competition and other lending standards than price has not been considered very much. The theories of Ruckes (2004) and Dell Ariccia and Marquez (2006) explain how strategic competitive behavior influences variations in lending standards over business cycles but they do not explicitly consider the effect of level of competition on lending standards at various times. Empirical works support the view of countercyclical lending standards and its role in procyclical lending behavior but do not actually examine the role of competition (Asea & Blomberg 1998; Maddaloni & Peydro 2011; Dell Ariccia et al. 2012). Jimenez and Saurina (2006) describe increasing competition as one of the possible reasons for looser lending standards in good times but do not examine its role separated from other determinants. Risk-taking The traditional view is that bank competition increases risk-taking. According to the competition-fragility view, more competition leads to lower profit margins and franchise value which encourages to more risk-taking (Keeley 1990). In contrast, the competition-stability view suggests that higher prices in less competitive markets may cause higher repayment problems and tendency to moral hazard and adverse selection (Boyd and Nicolo 2005). The model of Martinez-Miera and Repullo (2010) combines these opposite views and proposes a U-shaped relationship between competition and risk-taking. 5 For example, the Fed makes similar surveys

7 6 Berger et al. (2009) illustrate how the empirical results are sensitive to various measures of competition and risks and empirical studies actually find U-shaped, inverse U-shaped, positive and negative relationships (e.g. Jimenez et al. 2013; Tabak et al. 2013;Kick and Prieto 2013). Fiordelisi and Mare (2014) study the relationship between competition and risk in European cooperative banks and find that competition increases stability. Risk-taking behavior during good times can affect realized risks and thus willingness and ability to lend during crises. Jimenez and Saurina (2006) describe how competition can reduce the net interest margin and thus profitability of banks which can encourage to raise the volumes at the expense of quality. When this does not lead to problem loans in the short run, strong loan growth can continue. However, this causes greater losses during a crisis. Appetite for risk can also change during crises. Bernanke et al. (1996) develop model for so called flight to quality behavior which means that during bad times the relative decline in lending is larger to borrowers with high agency costs. Their empirical analysis supports the theory. Albertazzi and Marchetti (2010) examine this with Italian data during the current financial crisis and find that large and low capitalized banks moved to less riskier loans. Funding structure and costs When banks compete also for deposits, competition environment can affect the funding structure and costs. The market power hypothesis proposes lower deposit rates in more concentrated markets and the empirical evidence supports that (e.g. Berger & Hannan 1989; Hannan 1997; Hannan & Prager 2004). There is only a little empirical evidence on how competition affects banks funding structures. Graig and Dinger (2013) describe how tight deposit competition raises deposit rates and thus incentives to use wholesale funding. However, they do not actually examine how competition affects the relative amount of deposit

8 7 and wholesale funding, but how deposit rates affect risks. They find that higher deposit rates, that implicates higher deposit competition, increase the risks of banks. According to many previous studies, bank funding structure had a significant role in lending behavior during the current crisis. Banks with relatively large share of wholesale funding and thus small share of core deposits were the most vulnerable to freeze in global money markets and reduced lending more (e.g. Ivashina & Scharfstein 2010; Iyer et al. 2014). Relationship banking The essential part of banking and lending decisions is asymmetric information between lender and borrower. Banks develop close relationships with customers to ease asymmetric information (Boot 2000). According to the information hypothesis, high bank competition can weaken the incentives to invest in soft information and thus relationship building which means weaker loan availability in more competitive markets (Petersen & Rajan 1995). In contrast, the model of Boot and Thakor (2000) suggests that competition encourages to relationship lending in order to maintain satisfactory prices in competitive markets. Empirical studies support both kind of hypothesis (see Degryse et al p ). Asymmetric information generally increases during crises and theoretical and empirical literature indicates the increased role of relationship banking. The model of Bolton et al. (2013) suggests that relationship banking diminishes the effect of crisis on lending. Relationship banks charge higher interest margins during normal times but do not change their loan terms as much as transaction banks during crises. Empirical analyses find that relationship banking diminished the common increase in loan prices and decrease in availability during the current crisis (Puri et al. 2011; Gambacorta & Mistrulli 2011; Bolton et al. 2013; Cotugno et al. 2013)

9 8 Summary and purposes of this paper The previous literature shows that the relationship between competition and lending behavior is multifaceted. The previous theoretical and empirical studies imply that tighter competition decreases interest rates on loans and deposits and loosens lending standards. Instead, the effect of competition on risk-taking and relationship lending is controversial. In this paper, I first examine the effect of competition environment on changes in lending during the crisis and thus the general effect of various channels discussed above. Second, I consider the contributions of various factors. The previous literature implies that competition decreases during the crisis. I study whether the change in competition is different in various competition environments. The previous studies concerning especially the current crisis indicate that changes in risk-taking, realized risks, relationship banking and funding structure are relevant factors for lending behavior during crises. The effect of competition environment can come also through these factors even if previous literature is controversial about the relationship between these and competition. I examine, as far as possible, the role of these indirect factors of competition environment in changes in lending during the crisis. 3 DATA AND EMPIRICAL APPROACH In this section, I first describe the data used in this paper and provide more detailed description about small business loan data and competition measures. Then, I present my empirical specifications.

10 3.1 The cooperative banks of OP-Pohjola group 9 OP-Pohjola group consists of an amalgamation of independent cooperative banks that own OP-Pohjola Central Cooperative. This central institution controls and supervises cooperative banks and produces e.g., information services. OP-Pohjola Central Cooperative has also many subsidiaries of which Pohjola Bank is the most important. Pohjola is a commercial bank that acts as OP-Pohjola Group s central bank and manages Group s liquidity and international operations. It also manages corporate lending for large and mid-size firms. The group structure and the same business model make, in one sense, these banks homogeneous but at the same time, the banks are quite heterogeneous especially regarding size and operational environment. 7 The following characteristics are relevant to my analysis. First, the individual cooperative banks are purely local and have non overlapping operating areas that are defined at the zip code level and cover the whole Finland. The general principle is that they do not compete against each other. However, many banks are very small and they can have shared business loans. Additionally, there are regional corporate banks that manage larger business loans in wider regional operating areas. Second, the cooperative banks get guidelines and constraints from the Central Cooperative but make independent decisions. Third, liquidity issues and market funding of the cooperative banks are managed with Pohjola. The cooperative banks have a checking account with overdraft facility and they can take a short term debt funding from Pohjola. The price of a certain short term debt at a given time is the same for all banks and reflects the price of Pohjola s own wholesale funding. This indicates that the availability of short term funding is not a problem for the 6 Situation at the end of Hyytinen and Toivanen (2004) use the data of the same cooperative banks to study whether the banks use branch network to invest monitoring and/or market power and how this affect loan interest rates and credit losses

11 10 individual cooperative banks but costs vary over time according to the price of wholesale funding of Pohjola which certainly can affect willingness to use this kind of funding in various times. Fourth, it is possible that cooperative banks reacted differently to the crisis than competing banks due to the different business models Data sources The unique bank data collected from inside the OP-Pohjola Central Cooperative include detailed monthly lending information for all cooperative banks between January 2004 and October The price and volume information is available for both the outstanding and new withdrawn business loans, mortgages and consumer loans. The data contain also detailed balance sheet, income statement and deposit information about each cooperative bank. Additionally, the information about small business lending is complemented by the loan-specific data that is available since September The data about locations of the branches of all Finnish banks is from the establishment data of business register of Statistics Finland. Yearly data between 2004 and 2010 include enterprise (bank) name and business ID, establishment (branch) name and code, municipality and zip-code. Local economic data are at the municipal level and collected from the statistical databases StatFin and Altika of Statistics Finland. The number of the branches of all banks is joined to each cooperative bank based on zip codes. The local economic data are combined at the municipality level so that local variables are calculated based on all municipalities where a cooperative bank operates. 8 See Groeneveld and de Vries (2009) and EACB (2010)

12 Small business lending data The dependent variables are the new withdrawn business loans a month divided to total assets in the previous month and the average margin on monthly new business loans. The corporate loans of the cooperative banks are mainly small business loans because Pohjola Bank manages the loans for larger companies 9. The new withdrawn loans consist mainly of new contracts but include also drawdowns of existing loan commitments. This measure describes better the changes in loan supply and demand than the change in outstanding loans used in many previous papers (e.g. Cornett et al. 2011) because the latter measure is affected also by the repayments of loans. As our interest is in the purely new lending activity, the effect of drawdowns of existing loan commitments is diminished by controlling the amount of off-balance sheet items that include these limits. The average margin of new business loans is the average interest margin on floating rate loans with either one of the Euribor or banks own prime rate as a reference rate. As the margin is the pricing element of loan contracts, it is appropriate measure for changes in banks loan pricing Based on loan specific data between 10:2008 and 06:2013, 89 percent of new corporate loan contracts were under euros and 96 percent under one million. The shares of total volume were 43 and 69 percent, respectively. 10 In fact, the average total interest rates on small business loans decreased during the crisis due to significantly loosened monetary policy and decreased market rates.

13 12 The general developments of the volumes and margins of the new business loans in the cooperative banks are presented in figure 1. Vertical lines in October 2008 indicate the beginning of the crisis period which is determined based on the collapse of Lehman Brothers on 15 September A The average of new business loans/total assets m1 2006m1 2007m1 2008m1 2009m1 2010m1 Time B The average new business loan margin m1 2006m1 2008m1 2010m1 Time Figure 1: The development of new business loans in Finnish cooperative banks The figure shows the average development of the volumes and margins of new business loans in the cooperative banks of OP-Pohjola group for the January 2004/2005 to September Panel A shows the 12 months moving average of banks average monthly loan volumes that is measured as the percentage of new withdrawn business loans divided by total assets in the previous month. Panel B shows index of monthly average margin of cooperative banks new withdrawn business loans.

14 Competition data Bank competition measures are typically classified based on the Traditional and New Empirical Industrial Organization (IO) literature. The traditional IO literature uses market structure measures 11 whereas new measures try to gauge market power more directly 12. Carbo et al. (2009) compare various competition measures and find that these are only weakly related which highlights the complexity of the issue. The recent literature often argues that national concentration is an inappropriate measure of competition (Claessen & Laeven 2003; Schaeck et al. 2009; Bikker et al. 2012). However, Schaeck et al. (2009) note that if bank markets are local, national concentration measures can in principle be a wrong way to gauge competition and refer to findings that show the relationship between local concentration and competition. I use both kinds of measures and take into account the different characteristics of them in analysis. 13 The market structure measure of this paper is based on locations of the bank branches of the cooperative banks and all other banks in Finland. The Herfindahl Hirschman Index (HHI) is computed assuming that the share of branches represents the market shares, like e.g. in Degryse and Ongena (2007) and Chong et al. (2013). I modify this by using the weighted effects of different banks branches based on banks total business loan market shares in Finland related to their total number of branches. For example, a weight of certain bank s 11 E.g. concentration ratios, number of banks or Herfindahl indices (Beck et al. 2010) 12 The new IO measures include the Lerner index, Panzar and Rosse H-statictics. Other used competition measures are e.g. net interest margin and return on assets (see e.g. Carbo et. al 2009) 13 Detailed information about the calculation of competition measures is presented in appendix 1

15 14 branch is over one if the bank s share of business loans is higher than the share of branches in the whole Finland. The Lerner index is a more direct competition or market power measure based on the gap between the marginal price and the marginal cost of an individual bank (e.g. de Guevara et al. 2005; Carbo et al. 2009; Beck et al. 2013). The Lerner index can be computed in various ways depending on the purposes and data availability. It is mostly computed by using total revenues or interest revenues and the marginal cost of total funding and operating costs. 14 I construct the Lerner index by using total interest revenues. 3.5 Control variables and descriptive statistics I control the key transmission channel of crisis on loan supply observed in previous literature, i.e., the dependence on short term market-based funding (e.g. Iyer et al. 2014) and other bank characteristics that can affect loan supply and pricing (e.g. Jimenez et al. 2012a; Gambacorta & Mistrulli 2014). Size, capitalization, liquidity and non-performing loans are the basic control variables and funding structure is a relevant one particularly in the context of the current crisis. Off-balance sheet commitments is an important control in my setting because my volume measure, new withdrawn loans, includes also drawdowns of existing loan commitments. In addition, I use local economic variables and province-time trend to control local loan demand. Bank variable definitions are based on previous studies (e.g., above mentioned) and characteristics of the data. The funding structure variable is the share of short term market based funding that consist in the cooperative banks mainly of short term loans from Pohjola and 14 In some papers the Lerner index is calculated more precisely to different products (see e.g. Jimenez et al. 2013)

16 15 negative checking account in Pohjola. The capitalization variable is the regulatory capital adequacy ratio. Liquid assets consist of cash, demand deposits in banks (especially in Pohjola) and debt securities eligible for refinancing with central banks. The bank size is measured as logarithm of total assets and non-performing loans are the loans having been in default for 90 days. The relative size of off-balance sheet is measured by the amount of irrevocable commitments to customers divided by total assets. The local economic variables include the unemployment rate, the growth of sales in establishments, the personal income and the growth in the number of corporations (see e.g. Adams & Amel 2011; Keeton 2009) Table 2 provides descriptive statistics of the variables used in estimations separately for periods before and during the crisis. The competition and local economic variables are measured yearly and all other variables monthly. The average of new business loans divided by total assets decreased from 0,34 % before the crisis to 0,28 % during the crisis and average loan margin increased from 1,29 percentage points to 1,58. The average Lerner index is considerably lower during the crisis which should mean increased competition. However, this reflects rather the problem of the Lerner index in measuring competition over time. For example, in this case, the sharp reduction in market rates during the crisis combined with large share of floating rate loans is likely to have a significant effect on the values of the Lerner indices. For this reason, my analysis is based on banks differences in the Lerner indices in certain time and relative changes in the Lerner indices after the onset of the crisis.

17 16 Table 1. Descriptive statistics before and during the crisis Dependent variables mean standard deviation number of observations before crisis crisis before crisis crisis before crisis crisis new business loans/total assets (%) 0,34 0,28 0,34 0, margin of new business loans (%-points) 1,29 1,58 0,49 0, Competition variables Lerner index 0,26 0,11 0,10 0, Branch HHI 0,65 0,68 0,27 0, Bank characteristics short term market-based funding/assets (%) 5,02 5,82 5,02 8, capital adequacy ratio (%) 21,27 23,04 5,88 6, liquid assets/total assets (%) 3,01 2,30 2,90 2, the log of the total assets 18,18 18,41 1,10 1, nonperforming loans/total assets (%) 0,66 0,59 0,55 0, OBS commitments/total assets (%) 4,80 4,87 1,86 1, Local economic conditions unemployment rate (%) 10,06 10,68 3,79 3, personal income (thousand euros) 22,74 23,84 3,04 2, the change in number of corporations (%) 2,79 1,13 2,42 2, the change in sales of establishments (%) 6,39-3,43 9,40 13, Empirical specifications I use difference-in-difference (DID) methods to analyze whether the changes in margins and volumes of new small business loans were different in different local bank competition environments after the onset of the financial crisis. The empirical identification is based on the data on the local cooperative banks of the same group with independent decision making and heterogeneous local competition environments. Some of the cooperative banks compete with many other banks including large global banks but some of the cooperative banks operate in local markets with less or not at all competing banks. Besides the market structure, there can be also other factors affecting the tightness of competition or differences in market power in different local markets. For that reason, I use both market structure, HHI, and more direct market power measure, Lerner index, as the competition variable. At the first stage, I estimate gradually the following econometric model:

18 17 (1)Y it = α + β 1 Crisis t + β 2 Comp i + β 3 Comp i xcrisis t + β 2 Short debt i + J β 3 Short debt i xcrisis t + j=1 δ j X j,i,t 1 + k=1 θ k R k,i,t + ε it, K where Y it is in one specification the monthly volume of new withdrawn business loans divided by total assets of the bank in the previous month and in other specification the average margin of new business loans in bank i at time t. I start by comparing the changes in average price and volume measures in the most competitive and in the other banks during the crisis. The pre-crisis time period is between 2004:01 and 2008:09 and the crisis period between 2008:10 and 2010:09 when Crisis dummy takes value 1. Comp dummy takes value one for the most competitive banks and zero for the other banks. The classification is based on either the average HHI or the average Lerner index before the crisis such that the most competitive are banks at the first quintile 15. The serious problems in money markets and thus the difficulties in short term market based funding of banks especially after the collapse of Lehman Brothers was a key characteristic of this crisis period. In order to better identify the effect of competition environment on lending, I take next into account this by setting dummy variable Short debt to one for the banks that were the most dependent on short term market-based funding before the crisis. The classification is based on one year averages of the share of short term market based funding of total assets before the beginning of the crisis and the most dependent banks are those at the top quintile. 15 Lower values of the Lerner index and the HHI mean higher competition.

19 18 The differences between the banks and local environments can cause differences in loan supply and demand and I control these by using one month lagged bank-specific balance sheet factors X and local market factors R described in the previous section. Further, in order to control better the differences in regional economic (loan demand) conditions, the changes in common lending conditions and unobservable differences between banks, I estimate the following model: (2)Y it = α i + λ t + β 1 Crisis t + β 2 Comp i + β 3 Comp i xcrisis t + β 4 Short debt i + J β 5 Short debt i xcrisis t + j=1 δ j X j,i,t 1 + k=1 θ k R k,i,t + γ l Province l t before L l=1 κ l Province l t crisis + +ε it, K L l=1 + I add first to the model 1 the province-time trends separately for periods before and during the crisis such that t before =1,,57 between 2004:01 and 2008:09 and zero otherwise and t crisis =1,,24 between 2008:10 and 2010:09 and zero otherwise. Finally, I include bank fixed effects to control unobservable fixed differences between banks and their local environments and time fixed effects to control the common changes in lending environment like significantly loosened monetary policy, increased uncertainty and new regulatory initiatives after the onset of the financial crisis. Comp and Short debt dummies are not used when bank fixed effects α i are included in this final model. 4 EMPIRICAL ANALYSIS This section provides results of the main empirical analysis for the effect of competition environment on the changes in volumes and prices of small business loans during the current financial crisis. In addition, the section includes several robustness checks.

20 Main results Tables 2 and 3 present bivariate results of the mean DID estimates of volumes and margins of new business loans for the competitive and other banks. 16 The classification is based on the Lerner index in panels A and on the HHI in panels B. Table 2 indicates that the volume of new business loans with relation to total assets is on average significantly lower during the crisis than before the crisis in both the competitive and other banks. The more important for the purpose of my study, the interaction terms of competition and crisis i.e., the DID estimates are negative and also statistically significant when the competition classification is based on the Lerner index in panel A. Whereas the average of new business loans divided by total assets in previous month decreased 19 % (0,410 ->0,333) among the competitive banks, the decrease was 14 % (0,298 ->0,255) among the other banks. Table 3 shows that the margins of new business loans increased significantly after the onset of the financial crisis. The average loan margin is lower in the competitive banks but the difference almost vanished during the crisis. The results are very similar with both competition measures and indicate that the average loan margin of the competitive banks increased about 43 basis points and 24 basis points in the other banks and the differences i.e., the DID estimates are also statistically significant. 16 The same results as figures when the classification is based on the Lerner index are presented in the appendix 2.

21 20 Table 2. Business loans volumes before and after the onset of the financial crisis (bivariate tests) This table presents mean difference-in-difference (DID) estimates when the dependent variable is the percentage of monthly new business loans divided by total assets in the previous month that is winsorized at the 99th percentile. The cooperative banks are classified into two groups: competitive, when cooperative bank s pre-crisis average of competition measure is at the first quantile; other, otherwise. The classification is based on the Lerner index in panel A and on the HHI in panel B. The DID estimates are printed in bold. Standard errors (not reported) are clustered at the bank level. ***, **, and * denote that the coefficients are statistically significantly different from zero at the 1%, 5%, and 10% level, respectively. Panel A: New loans/ Total assets(t-1), Competition measure: Lerner index Competitive Other Difference Before October ,410 0,298-0,113*** After October ,333 0,255-0,078*** Difference -0,078*** -0,043*** -0,035** Panel B: New loans / Total assets(t-1), Competition measure: HHI Before October ,404 0,299-0,105*** After October ,339 0,253-0,086*** Difference -0,065*** -0,046*** -0,019 Table 3. The margins of new business loans before and after the onset of the financial crisis (bivariate tests) This table presents mean difference-in-difference (DID) estimates when the dependent variable is the average margin of monthly new business loans that is winsorized at the 99th percentile. The cooperative banks are classified into two groups: competitive, when cooperative bank s pre-crisis average of competition measure is at the top quantile; other, otherwise. The classification is based on the Lerner index in panel A and on the HHI in panel B. The DID estimates are printed in bold. Standard errors (not reported) are clustered at the bank level. ***, **, and * denote that the coefficients are statistically significantly different from zero at the 1%, 5%, and 10% level, respectively. Panel A: Margin of new business loans, Competition measure: Lerner index Competitive Other Difference Before October ,085 1,323 0,238*** After October ,518 1,561 0,044** Difference 0,433*** 0,238*** 0,195*** Panel B: Margin of new business loans, Competition measure: Branch HHI Before October ,028 1,340 0,312*** After October ,459 1,578 0,119*** Difference 0,431*** 0,238*** 0,193*** Tables 4 and 5 present results for volumes and prices when the effect of dependence on short term market-based funding is taken into account and the bank-specific and local environment controls are included. I use mainly the competition dummy based on the Lerner index and make comparison using the HHI in the model with all controls. Table 4 indicates that the coefficients of the interaction term of the competition and crisis dummies remain negative and rather the same size in all specifications which means that

22 21 decrease in volumes during the crisis is higher in the most competitive banks than in the other banks. The values of these coefficients are about that are roughly half of the average decrease of all banks during the crisis (see Table 1). The coefficients are not statistically significant in all estimations which at least partly reflect the high volatility of the volume measure. The coefficients of the interaction term of dependence on short term marketbased funding and crisis are negative. One more important result is that the coefficients of relative amount of off-balance sheet items are positive and statistically significant. This is an important control variable as the purpose is to examine the volumes of new loans and the dependent variable is new withdrawn business loans that includes also drawdowns of existing loan commitments. The results in table 5 show that the coefficient of the crisis-competition interaction term is positive and statistically significant in all models and also when the HHI is used as the competition classification measure. This confirms the result of the bivariate analysis that the increase in the average margins of new small business loans during the crisis was higher in the most competitive banks than in the other banks. The values of coefficients are quite similar in different specifications and indicate that the increase of the average margin is about 19 basis points higher in the most competitive banks. The economic relevance becomes clear when this is compared to the average increase of all banks that is 29 basis points (Table 1).

23 22 Table 4. Business loan volumes before and after the onset of the financial crisis (multivariate tests). This table presents regression of loan volumes on crisis, competition, dependence on short term marketbased funding, and their interactions plus bank and local environment controls. The dependent variable is defined as the monthly new business loans divided by total assets in the previous month and it is winsorized at the 99 th percentile. Crisis takes value one after september 2008 and zero otherwise. Competitive takes value one when cooperative bank s pre-crisis average of competition measure is at the top quantile and zero otherwise. Standard errors,clustered at the bank level,are reported in parentheses. ***, **, and * denote that the coefficients are statistically significantly different from zero at the 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) (5) (6) Crisis *** *** *** (0.0080) (0.0082) (0.0092) (0.0155) (0.0375) (0.0372) Competitive (Lerner) *** *** (0.0213) (0.0238) (0.0273) (0.0268) Crisis x Competitive (Lerner) ** * (0.0144) (0.0155) (0.0166) (0.0166) (0.0173) Crisis x Competitive (HHI) (0.0162) Short market-based debt dummy *** * * (0.0292) (0.0245) (0.0249) Crisis x Short market-based debt ** * * * (0.0170) (0.0174) (0.0171) (0.0177) (0.0170) Capital adequacy ratio t *** *** (0.0013) (0.0012) (0.0019) (0.0019) Liquid assets/total assets t *** *** * * (0.0015) (0.0015) (0.0011) (0.0011) The log of the total assets t *** *** (0.0085) (0.0087) (0.0666) (0.0673) Nonperforming loans/total assets t ** * (0.0091) (0.0087) (0.0069) (0.0069) OBS commitments/total assets t *** *** *** *** (0.0046) (0.0045) (0.0034) (0.0034) Local economic variables no no yes yes yes yes Time-province trends no no no yes yes yes Time-fixed effects no no no no yes yes Bank fixed effects no no no no yes yes Number of observations R-squared (w) (w) F-statistic

24 23 Table 5. Business loan prices before and after the onset of the financial crisis (multivariate tests). This table presents the regression of business loan prices on crisis, competition, dependence on short term market-based funding, and their interactions plus bank and local environment controls. The dependent variable is defined as the monthly average margin of new business loans and it is winsorized at the 99 th percentile. Crisis takes value one after september 2008 and zero otherwise. Competitive takes value one when cooperative bank s pre-crisis average of competition measure is at the top quantile and zero otherwise. Standard errors,clustered at the bank level,are reported in parentheses. ***, **, and * denote that the coefficients are statistically significantly different from zero at the 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) (5) (6) Crisis *** *** *** *** *** *** (0.0171) (0.0189) (0.0215) (0.0338) (0.0700) (0.0708) Competitive (Lerner) *** *** (0.0358) (0.0361) (0.0440) (0.0357) Crisis x Competitive (Lerner) *** *** *** *** *** (0.0334) (0.0308) (0.0310) (0.0289) (0.0294) Crisis x Competitive (HHI) *** (0.0325) Short market-based debt dummy * (0.0340) (0.0289) (0.0289) Crisis x Short market-based debt (0.0299) (0.0305) (0.0308) (0.0305) (0.0317) Capital adequacy ratio t ** ** (0.0025) (0.0024) (0.0038) (0.0039) Liquid assets/total assets t (0.0038) (0.0036) (0.0027) (0.0027) The log of the total assets t *** *** (0.0157) (0.0137) (0.1209) (0.1245) Nonperforming loans/total assets t *** ** (0.0235) (0.0198) (0.0111) (0.0110) OBS commitments/total assets t *** *** *** *** (0.0080) (0.0071) (0.0048) (0.0049) Local economic variables no no yes yes yes yes Time-province trends no no no yes yes yes Time-fixed effects no no no no yes yes Bank fixed effects no no no no yes yes Number of observations R-squared (w) (w) F-statistic Robustness checks I conduct various robustness checks to consider possible problems in the main analysis. The difference-in-difference analysis based on the two groups, the competitive and the other banks, assumes the nonlinear effect of the competition measures because the competition variables take only the value one or zero. Now I use continuous values instead of dummies for the pre-crisis competition measures and make otherwise the similar estimations. Results in Tables 6 and 7 are comparable to Tables 4 and 5.

25 24 Table 6 shows that the coefficients of the interaction terms of the crisis and the pre-crisis average of the competition measures are positive and mainly statistically significant. As higher value of the competition measures means lower competition, these results indicate that the decrease in new small business loans during the crisis and tightness of pre-crisis competition are positively related. 17. To assess the economic significance, I compare the predictions of the model 3 for the effects of the crisis in the tenth and ninetieth percentile of the average pre-crisis value of the Lerner index. The decrease of the volume of lending is 17 % evaluated in the tenth percentile and 7 % in the ninetieth percentile. In sum, the results are in line with my main analysis. Table 7 shows that the coefficients of the interactions terms of the crisis and the pre-crisis average of competition measures are negative and highly statistically significant in all specifications which means higher increase in loan margins in the more competitive banks. Evaluated at the tenth and ninetieth percentile of the average pre-crisis value of the Lerner index, the coefficient in the fifth column indicates that the increase in loan margin is 21 basis points lower in the less competitive banks. The difference is 13 basis points when evaluation is based on the average pre-crisis values of the HHI in column I have dropped observations if the HHI equals one which explains the lower amount of observations in columns 6 in Tables 7 and 8. The dropped observations consist mainly of very small banks.

26 25 Table 6. Business loans volumes before and after the onset of the financial crisis (cf. table 4). This table presents the regression of loan volumes on crisis, competition, dependence on short term marketbased funding, and their interactions plus bank and local environment controls. The dependent variable is defined as the monthly new business loans divided by total assets in the previous month and it is winsorized at the 99 th percentile. Crisis takes value one after september 2008 and zero otherwise. The competition measures are calculated as pre-crisis average. Standard errors,clustered at the bank level,are reported in parentheses. ***, **, and * denote that the coefficients are statistically significantly different from zero at the 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) (5) (6) Crisis *** *** *** (0.0180) (0.0211) (0.0218) (0.0242) (0.0416) (0.0516) Average Lerner before crisis *** *** (0.0814) (0.0943) (0.1220) (0.1204) Crisis x Average Lerner before crisis *** ** ** ** (0.0681) (0.0726) (0.0710) (0.0679) (0.0706) Crisis x Average HHI before crisis (0.0450) Short market-based debt/assets *** ** * (year average before crisis) (0.1596) (0.1211) (0.1337) Crisis x Short market-based debt *** ** ** * ** (0.0885) (0.0853) (0.0914) (0.0895) (0.1151) Capital adequacy ratio t *** *** (0.0015) (0.0014) (0.0019) (0.0021) Liquid assets/total assets t ** *** (0.0014) (0.0014) (0.0011) (0.0018) The log of the total assets t *** *** (0.0080) (0.0083) (0.0666) (0.0872) Nonperforming loans/total assets t * (0.0090) (0.0087) (0.0069) (0.0111) OBS commitments/total assets t *** *** *** *** (0.0044) (0.0043) (0.0034) (0.0034) Local economic and bank controls no no yes yes yes yes Time-province trends no no no yes yes yes Time-fixed effects no no no no yes yes Bank fixed effects no no no no yes yes Number of observations R-squared (w) (w) F-statistic

27 26 Table 7. Business loan prices before and after the onset of the financial crisis (cf. table 5). This table presents the regression of business loan prices on crisis, competition, dependence on short term market-based funding, and their interactions plus bank and local environment controls. The dependent variable is defined as the monthly average margin of new business loans and it is winsorized at the 99 th percentile. Crisis takes value one after september 2008 and zero otherwise. The competition measures are calculated as pre-crisis averages. Standard errors,clustered at the bank level,are reported in parentheses. ***, **, and * denote that the coefficients are statistically significantly different from zero at the 1%, 5%, and 10% level, respectively. (1) (2) (3) (4) (5) (6) Crisis dummy *** *** *** *** *** (0.0430) (0.0470) (0.0480) (0.0485) (0.0743) (0.0978) Average Lerner before crisis *** *** ** ** (0.1495) (0.1584) (0.2355) (0.2259) Crisis x Average Lerner before crisis *** *** *** *** *** (0.1625) (0.1645) (0.1666) (0.1398) (0.1362) Crisis x Average HHI before crisis *** Short market-based debt/assets (year average before crisis) (0.1731) (0.1807) (0.1619) (0.0991) Crisis x Short market-based debt *** *** (0.1570) (0.1635) (0.1614) (0.1647) (0.2691) Capital adequacy ratio t (0.0029) (0.0028) (0.0037) (0.0037) Liquid assets/total assets t (0.0038) (0.0036) (0.0026) (0.0026) The log of the total assets t *** *** (0.0163) (0.0150) (0.1191) (0.1462) Nonperforming loans/total assets t *** ** (0.0235) (0.0199) (0.0110) (0.0142) OBS commitments/total assets t ** *** *** *** (0.0081) (0.0072) (0.0047) (0.0054) Local economic variables no no yes yes yes yes Time-province trends no no no yes yes yes Time-fixed effects no no no no yes yes Bank fixed effects no no no no yes yes Number of observations R-squared (w) (w) F-statistic Next I take into account that also the effects of the bank characteristics can be different in good and crisis times by using the interactions terms of the crisis and all the explanatory variables (see Cornett et al. 2011). Additionally, I use lagged values of the competition measures instead of pre-crisis averages. I use yearly data because the competition variables are measured annually. All the explanatory variables are one year lagged and also time fixed