Community Banks as Small Business Lenders The Tough Road Ahead

Community Banks as Small Business Lenders:TheTough Road AheadO.Emre Ergungor∗Federal Reserve Bank of ClevelandWorking Paper No.:02-03∗Economist,Research Dept.P.O.Box6387,Cleveland,OH44101-1387.Ph:(216)579-3004,Fax: (216)579-3050.E-mail:ozgur.e.ergungor@I would like to thank Mark Sniderman,James Thomson,Joseph Haubrich,Robert Avery,Stanley Longhofer,Phil Strahan,Troy Adair,Peter Rupert and Ed Nosal for many valuable comments.The usual disclaimer applies.The views stated herein are not necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System.Community Banks as Small Business Lenders:TheTough Road AheadAbstractThis paper investigates the viability of community banks as small business (relationship)lenders.Theory suggests that competition reduces the benefits ofbank-borrower relationships,making relationship loans more risky and lessprofitable.In support of this theory,Ifind that community banks do not earn asufficiently high return on their small business lending to justify the relativelyhigh risk associated with this line of business.These results raise doubts aboutthe viability of the community bank business model and suggest thatcommunity banks willfind it difficult to channel additional credit to smallbusinesses.(JEL G21,G28)Keywords:small business lending,relationship lending,community banksI IntroductionCommercial banks are the most important supplier of credit to small businesses (Bitler,Robb,and Wolken,2001).In particular,community banks with less than$1 billion in total assets hold,on average,17percent of their assets in small business loans(SBLs)and supply40percent of small business credit.1Although community banks control less than half of the market,research-to-date suggests that the SBLs supplied by community banks are fundamentally different from the loans made by larger munity banks are better than larger banks in establishing close relationships with borrowers,and processing soft,unverifiable information(Stein,2002;Berger,Miller,Petersen,Rajan,and Stein,2002).In 1Larger banks hold only about6percent of their assets in SBLs.Data come from June2002Call Reports.1contrast,larger banks rely more heavily on centralized,automated credit scoring that evaluates borrowers based on“hard”,quantifiable information.Consequently, community banks seem essential to the fulfillment of small,informationally opaque businesses’funding needs.To see whether community banks will remain as a reliable source of funding for these small enterprises in the future,we need to address the following two questions:First,does small business lending add value to community banks?Second,is small business lending a crucial line of business for community banks?To answer these questions,I study how bank performance is affected by small business lending.The evidence I present suggests a negative response to both questions.Ifind that small community banks with total assets less than$150Million that are more actively involved in small business lending than otherwise identical banks never outperformed the less active banks in any year between1996and2002. These active banks earn lower risk-adjusted income,have less profitable business loan portfolios and lower Sharpe ratios.In a recent article,DeYoung,Hunter,and Udell (2003)argue that community-bank business model is viable but a large number of community banks are poorly managed.Ifind that even the well-managed banks (based on X-efficiency)fail to outperform their peers when they devote a larger fraction of their assets to small business lending.The evidence is ambiguous for large community banks($150Million-$1Billion in assets).Banks with total assets larger than$1Billion do not seem to be affected by small business lending.Based on this2evidence,the main message of the paper is that small community banks may be losing their traditional advantages in processing soft information due to structural changes in the marketplace.2This may also explain why earlier papers fail to provide a clear indication that small businesses are desperately dependent on community bank funding.3The rest of the paper is organized as follows.Section II reviews the literature and explains why community banks still devote17percent of their assets to small business lending despite the threat to their future performance.Section III describes the dataset and the method used in the analysis.Section IV presents the results. Section V concludes.II Small Businesses and the New Financial Services IndustryLending to small businesses entails an information-intensive credit evaluation process. Little information is publicly available on thefinances of small businesses because 2Myfindings should not be interpreted as SBLs are bad investments in general.After all,large banks and non-bank lenders,such asfinance companies,are eager to increase their presence in themarket.I am only pointing at some potential problems in the community bank business model.3For example,smallfirms in areas with few small banks are no more credit-constrained in the long-run thanfirms in areas with many small banks although short-term disruptions-such as late payment of trade credit-may occur(Jayaratne and Wolken,1999;Berger,Miller,Petersen,Rajan, and Stein,2002).Scott(2003)finds that even though small banks emphasize soft information while large banks do not,soft information produced by small banks is not valued highly by smallfirms.3they are not traded in the stock market,financial analysts do not track their performance,and theirfinancial statements and contracts(labor,customer and so on)are kept private.The lack of public information erects a barrier between small businesses and investors,which thwarts these small institutions from accessing capital markets but creates a raison d’ˆe tre for banks(Greenbaum,Kanatas,and Venezia, 1989;Sharpe,1990;Haynes,Ou,and Berney,1999;Stein,2002;Cole,Goldberg,and White,2003).4Community banks acquire borrower-specific information by devoting substantial resources to getting to know their customers and developing relationships with them. What is the value of this information?The evidence suggests that when a bank knows more about afirm than other banks,it can charge a little extra for its services without fearing the loss of that customer(Berger and Udell,1996).So,community banks lend to small businesses despite the cost of getting to know them because they hope to recover their initial losses from profitable future business.Yet,large banks and nonbanks(mercialfinance companies)are eating away community banks’turf.According to the American Bankers Association’s sixth annual community bank competitiveness survey,37percent of the900respondents in50 states said large banks were their chief competitors for SBLs.How will competition change the nature of community bank lending?Petersen and Rajan(1995)argue that greater competition reduces banks’profits and 4Venture capital is not within the scope of this paper.4decreases costly relationship lending.After all,banks are not willing to incur the initial losses that come with new relationships if they do not expect to profit from them.However,what alternative source of income does a bank have in the absence of relationship lending and how does competition affect that alternative?In a recent study,Boot and Thakor(2000)address these questions.Banks,they argue,engage in both relationship and arm’s-length(transaction)lending.Unlike relationship loans—which are tailored to the needs of a particular borrower based on a bank’s previous experience with that customer—transaction loans are commodities that can be replicated by any bank because they do not necessitate gathering of detailed information about the borrower.Boot and Thakor suggest that the ability to tailor products to an established borrower’s particular needs is the way relationship loans provide better protection from competition.Theyfind that competition reduces bank’s profits from both relationship and arm’s-length lending.However,the effect is uneven across the two alternatives.As expected,an increase in competition reduces a bank’s profits from transaction lending more than its profits from relationship lending.Thus, competition encourages banks to shift from transaction to relationship loans.Thefindings of Berger,Goldberg,and White(2001)appear to support this argument. They observe higher small business(relationship)lending by small banks when competition increases in their market due to merger activity or new entry. However,the nature of relationship lending itself changes with increasing competition.5It becomes more important,but each loan has less added value for the borrower.In other words,as the bank’s already strained resource-pie is divided among a larger group of relationship borrowers,the bank will serve less pie to each customer.Boot and Thakor’s analysis has serious implications for the futurefinancial performance of community banks.Thefinancial performance of banks depends on thefinancial performance of businesses they extend credit to.Because small businesses rely on relationship-based services,their performance is likely to deteriorate when competition forces banks to reduce the overall quality of these services.This,in turn,weakens the banks’earnings performance and asset quality. To make matters especially worse for community banks,large-bank competitionaffects mostly the high-quality end of a community bank’s loan rge banks use centralized,automated credit scoring which is not an information-intensive credit evaluation technique,but helps them generate large volumes of SBLs—primarily,loans less than$100,000—at a lower cost than small banks for borrowers whose credit histories receive a passing grade(see Levonian,1997; Akhavein,Frame,and White,2004;Frame,Srinivasan,and Woosley,2001).The disadvantage of community banks is that the capital cost of this technology issignificant;so,it is not suitable for many small banks with low-volume operations. Also,while community banks felt large-bank competition primarily in themicro-lending area,this is likely to rge banks are exploring ways to use this technology with larger credits and gaining ground.A1998survey found that the6median loan size scored increased from$100,000in1997to$150,000in1998(see Zuckerman,1996;Feldman,1997;Ely and Robinson,2001).Petersen and Rajan (2002)report that,mainly as a consequence of growing use of information technology, the physical distance between highest-quality borrowers and their lenders has been increasing since1970s.Afirm that began borrowing from its lender in1993is34 percent farther away from the lender than an otherwise identicalfirm thatfirst borrowed from its lender in1983.Over time,even the lower quality borrowers began to increase their distance from their lenders.5This trend may cause community banks to replace the loans large banks take away from them with less profitable/riskier credits.Clearly,this will further impair the performance of the community banks. Loan pricing is not the only area where community banks are pressured by larger banks.One resource in short supply in the commercial lending market today is commercial loan officers.6Banks cut their loan-officer-training programs over the past decade to reduce costs.This made economic sense at the time because the use of credit scoring models had reduced large banks’dependence on loan officers,and small businesses did not care much about their relationships with a loan officer during the good times.According to the1999National Small Business Survey conducted by PSI Global,only10percent of small businesses said their relationship with a loan officer was the primary reason for choosing a lender.With the slowing economy,however, 5In contrast,Degryse and Ongena(2002)find that the distance betweenfirms and their banks inBelgium has not increased substantially in the1975-1997period.6American Banker,February26,2001,“Wanted:A Few Good Commercial Loan Officers”7that same fraction almost doubled to19percent a year later,while the market for commercial loan officers was the tightest in14years.The problem is especially distressing for community banks,which cannot afford the signing bonuses and other benefits offered by large banks to experienced,high-quality loan officers who are fundamental for profitable relationships(Myers,2001).Thus,with increasing competition but without the necessary resources,our discussion raises the possibility that community banks might be losing their traditional advantage in soft-information processing.If that is the case,we should observe some deterioration in the the quality and profitability of relationship loans.In the remainder of the paper,I present the signs of deterioration.III Data and MethodCommunity banks are usually defined as banks with total assets less than$1billion.I abide by this definition but carve out a“small community bank”category which consists of banks with total assets less than$150million.These are the banks that are often obligated to make SBLs because they cannot devote more than10percent of their capital to any single borrower.Finer partitions do not affect the results. Banks with total assets between$150million and$1billion and those with assets greater than$1billion will be referred to as“large community banks”and“large banks”,respectively.I analyze the effect of small business lending on bank performance in the1996-20028period.I estimate two separate models.First,in Specification A,I examine the effect of a bank’s SBL portfolio in1996on its future performance.Taking1997or any other year as a starting point does not affect the results.Second,in Specification B,I examine how a change in the size of a bank’s small business lending portfolio and other market-and bank-specific factors from1996to1999affects the change in the bank’s performance over the same period and in the future(2000,2001and2002).In other words,Specification A explains how well the banks that were heavily invested in small business lending in1996performed later on,while Specification B shows how the banks that became more aggressive were affected by this strategy.I start the analysis in1996because starting from earlier years significantly reduces the number of banks that were operational in the entire period.7Individual bank balance sheet and income statement data are obtained from the December Bank Call Reports,with the exception of small-business-lending data, which are reported only in June Call Reports and deposit market data which are reported in the Summary of Deposits.Other bank-specific data,such as geographic location,age,and Bank Holding Company affiliation information come from the Federal Reserve Bank Structure and Relationships Tables.December1996Call Reports contain data from9,375commercial banks.From those,only6,512remained active in2000and their number declined to6,223in2001and6,020in2002.7Small business lending data start in1994.9A Dependent VariablesI have three variables that measure how well banks perform with larger small business lending.Thefirst one,Roa,is the return on assets.The second one,Sharpe, is a reward-to-risk ratio defined asSharpe=Time-series mean of quarterly ROAs Standard deviation of quarterly ROAswhere the time-series is the quarterly ROAs from1997to2002.While Roa and Sharpe measure the overall bank performance,the third variable, Yield,captures the profitability of business loans.It is similar to the risk-adjusted yield in Carter,McNulty,and Verbrugge(2004)and defined as:Yield=Interest revenue on business loans−net business-loan charge-offsTotal business loansThe only difference is that I am not subtracting the risk free rate because it is not necessary in my specification(see below).Admittedly,all three variables have their strengths and weaknesses.Roa and Sharpe incorporate all the factors that drive a bank’s performance including the net effect of nonperforming loans and the non-interest costs associated with delivering the SBLs. However,their weakness is that they may be influenced by factors other than small business lending which one cannot fully control for.Yield is an attempt to measure the risk-adjusted return on business loans.Its strength is its focus on business loans as an isolated business line.Its weakness,however,is the omission of the costs10associated with delivering business loans;i.e.it is a revenue-based measure not profit. The risk-adjustment(net charge-offs)is also problematic.Higher charge-offs than peers may signify that management has adopted a more proactive approach to charging-offloans and may not necessarily indicate increased loan problems. Conversely,low charge-offs may not signify the lack of loan problems but rather the reluctance to charge-offloans.Nevertheless,in the absence of better measures,these three variables are reasonable performance proxies despite theirflaws.The reader must keep their strengths and weaknesses in mind while interpreting the results.B MethodI estimate Specification A using OLS:x1+ (1) Y t=α+βSBL SmallBus1996+β1where Y t∈{Roa t,Y ield t,Sharpe},t∈{1997,1998,...,2002},and SmallBus1996is the share of SBLs in the bank’s total assets in1996.For all three performance variables,a negativeβSBL would suggest that banks heavily invested in small business lending cannot outperform their peers.88A word of caution is in order.All small business lending is not necessarily relationship lending; we know that other technologies are available(such as asset-based orfinancial statement lending; see Berger and Udell,2002).However,due to the difficulty of constructing appropriate controls for the other technologies,all SBLs are treated as relationship loans in the literature.This is also the11x1is a set of explanatory variables,which capture bank-and market-specific factors set at their1996values that may affect a bank’s future performance.In particular,x1includes the standard deviation of quarterly ROAs in the1993-1996period as a measure of risk(Std Dev)and three variables as a measure of market concentration and size structure:Herf,MediumShare,and LargeShare.Herf is the deposit-market Herfindahl Index,which measures the market concentration in the MSA or county where the bank is located.MediumShare is the asset share of banks with total assets between$150Million and$1Billion in the total banking assets of the market(MSA or county).All banks that have a branch in the market are included in the total banking assets.The third variable,LargeShare,is the share of banks with total assets greater than$1Billion in the total banking assets.Berger,Rosen,and Udell(2001) have shown that SBL interest rates are lower in markets dominated by large banks but how this translates into overall bank profitability in unclear.I also include an estimate of the banks’X-efficiency(Efficiency)calculated using the alternative profit efficiency approach and Fourier-flexible functional form.I prefer the alternative profit efficiency approach over the standard profit efficiency because it provides a way of controlling for unmeasured differences in output quality and market power—which is what information-intensive lending is about.I deviate from the standard practice of using six years of data and use only four because as I go back to 1991to obtain the efficiency in the1991-1996period,I lose more than150 assumption in this paper.12observations.The estimates are admittedly more noisy but the results are notaffected by this deviation(see the discussion of Efficiency in Specification B below). Because the steps involved in calculating the X-efficiency are complicated,I omit that discussion here but refer the reader to Berger and Mester(1997).For the sake of brevity,all the other bank-and market-specific factors included in x1 are described in Table1.With Specification B,I examine how the changes in x1and SmallBus in the1996-1999period affected future performance measures.Denoting the set of explanatory variables included in Specification B by x2,I estimate the following model using OLS:Y t=α+βY1996Y1996+β∆SBL∆SmallBus1999−1996+βSBL SmallBus1996+β2x2+(2)where t∈{1999,2000,2001,2002}and∆SmallBus1999−1996is the difference between SmallBus1999and SmallBus1996.The initial value of the Sharpe ratio on theright-hand side,Sharpe1996,is defined over the1993-1996period.x2contains x1and some additional variables that measure the change in x1from1996to1999.The variables in x2are listed and defined in Table2.Summary statistics are in Table3. Note that I do not define a∆Efficiency variable because I have only three years (1997,1998and1999)to calculate the change in efficiency from1996to1999.My preferred approach is to recalculate the X-efficiency in the1994-1999period,using six years of data.I also recalculate Std Dev over a longer time period(1994-1999)and13drop the more noisy estimate over the1993-1996period.Again,for all three performance variables,a negativeβ∆SBL would suggest that a bank cannot outperform its peers by increasing its small business lending.IV AnalysisIn this section,I present the results of the analysis.For the sake of brevity,the tables contain the results for only a selected number of years and a selected set of explanatory variables although all of them were included in the analysis.9Tables4and5present the regression coefficients for Specification A and B respectively.Small business lending does not add value to community banks.Small community banks,which devote a larger share of their assets to small business lending could not outperform their peers in any year in the1997-2002period(Spec.A).In fact,in every year except1999,they slightly underperformed their peers.Their underperformance is statistically significant but economically small;a one standard deviation increase in SmallBus in1996(11.5percentage points)leads to a decline of 0.05percent in Roa(five percent of sample mean in1996),0.2percent in Yield(two percent of the sample mean)and a decline of0.11in Sharpe(three percent of the sample mean)in2001.The declines may be small but the crucial observation is that SBLs do not add value.Results from Specification B are parallel to those in Specification A.Small 9Available upon request.14community banks that became more aggressive in small business lending over the 1996-1999period could not outperform their otherwise identical peers that did not change the size of their SBL portfolio.All three performance indicators lead to the same conclusion;Yield declines,Roa and Sharpe areflat with increasing small business lending(∆SmallBus).All three measures show signs of deterioration if the initial small business lending is high(SmallBus).As before,the economic significance of these declines is not large.For example,one standard deviation increase in∆SmallBus over1996-1999leads to a0.7percent decline in Yield in2000which is about seven percent of the mean.The evidence is ambiguous for the larger banks.Specification A suggests that Yield of large community banks declines significantly with larger small business lending but Specification B only weakly confirms thisfinding.Roa does not seem to be affected by small business lending in Specification A but Specification B suggests that large community banks benefited from more aggressive small business lending during the economic slowdown in2000,2001,and2002.Banks with assets greater than$1 billion attain significantly higher Sharpe ratios with increasing small business lending but other measures do not detect a significant effect in other years—althoughcoefficients have a positive sign.These results suggest that the small community banks were hit hard by the changes in the banking industry.Note that this contradicts thefindings of Carter,McNulty, and Verbrugge(2004)whofind that smaller banks earn greater Yield than larger15banks.There is one major reason for this difference.Instead of pooling all banks together and studying the effect of size by interacting SmallBus by the size variable,I analyze all size groups separately;i.e.,I allow bank and market-specific effects to play different roles in each size group.A Robustness CheckIn this section,I focus solely on small community banks and subject my results to a wide variety of robustness checks.I investigate whether the negative effects of small business lending are driven by a sub-sample of small banks.In Table6-Panel A,I divide the sample into three groups of equal size.Small lenders (0.1%<SmallBus<13.1%),medium lenders(13.1%<SmallBus<22.8%),and heavy lenders(SmallBus>22.8%).Doing so enables me to address the question of whether SBLs are an essential line of business for community banks.If small business lending is essential,small lenders should be better of by making more loans.Yet,I fail tofind any indication that small business lending might be beneficial to any particular group of lenders.Table6-Panel B is a linearity check.The results from Specification B would lose their meaning if they were significant only among those banks that reduced their small business lending(∆SmallBus<0).In other words,due to conditions that were specific to those banks and their markets,these institutions might have reduced their small business lending and improved their performance.However,the effect of small16business lending may be positive among the institutions that increased their lending (∆SmallBus>0).To make sure that banks that increased their lending have indeed been adversely affected,I divide the sample into two groups;one which includes only those institutions with∆SmallBus<0and one which includes those with∆SmallBus>0.The results are in Table6-Panel B and show that those institutions with positive∆SmallBus did not benefit from higher small business lending.In fact, the effect of aggressive lending on Sharpe,which was insignificant in Table5,is negative and significant among banks with∆SmallBus>0.As a further robustness check,note that my small business lending variable covers all loans less than$1million.Yet,credit scoring mostly affects micro-business loans (Feldman,1997;Akhavein,Frame,and White,2004).Anecdotal evidence suggests that community banks are responding by shifting their loan portfolios towards loans larger than$100,000(e.g.,Zuckerman,1996;Ely and Robinson,2001).Tofind out whether community banks are beaten up in micro-lending but performing much better with larger loans,I create a large-SBL ratio,LSBRatio,which measures the share of loans between$100,000and$1Million in the total SBL portfolio.∆LSBRatio is the change in that variable from1996to1999.Table6-Panel C presents the results.Those institutions that were heavily invested in large SBLs in1996could not outperform their rivals in later years(Specification A). Becoming more aggressive in this area did not pay offeither(Specification B).In fact,small community banks may earn a slightly better Yield if they stay in the17micro-lending business.As an additional robustness check,I address the argument in DeYoung,Hunter,and Udell(2003)that community-bank business model is viable but a large number of community banks are poorly managed.My strategy is to analyze the interaction of Efficiency with SmallBus.A large,positive interaction term would imply that the negative coefficient of SmallBus is driven by inefficient banks and that efficient ones may actually benefit from small business lending.The results are in Table6-Panel D. The interaction term is positive and significant in some regressions but even the most efficient banks cannot outperform their peers by increasing their small business lending.If we focus on the case where Efficiency equals1,βSmallBus+βSmallBus×Eff andβ∆SmallBus+β∆SmallBus×Eff suggest that even the most efficientfirm is slightly hurt by aggressive small business lending in some years and it does not reap any benefits in other years.In Table6-Panel E,I check whether the results are solely driven by rural or MSA banks.There is still no indication that small business lending adds value to small community banks in either sub-sample.V ConclusionThis paperfinds that small business lending does not add value to community banks. Neither does it seem to be a crucial line of munity banks cannot outperform their rivals by making more SBLs.The results are strong for banks with18。