Local Market Structure and Small Bank Risk: Is there a Connection?

Size: px
Start display at page:

Download "Local Market Structure and Small Bank Risk: Is there a Connection?"

Transcription

1 Local Market Structure and Small Bank Risk: Is there a Connection? Jim Rauch Gustavus Adolphus College 800 West College Drive Department of Economics Saint Peter, MN jrauch@gac.edu and Jill M. Hendrickson University of the South 735 University Avenue Department of Economics Sewanee, TN jhendric@sewanee.edu January 2003 Draft Abstract This study uses fourteen quarters of Call Report data between 1999 and 2002 to test whether local market structure affects loan quality at small banks. The hypothesis is that smaller banks located in markets dominated by large banks can discriminately choose the best loan applicants and, consequently, make better quality loans. Using regression methods, the empirical results support the hypothesis by finding that small banks in local markets dominated by large banks have higher quality loan portfolios, ceteris paribus. 1

2 Local Market Structure and Small Bank Risk: Is there a Connection? Introduction The United States' history of commercial banking evolved such that the small banker was a central source of credit for businesses and consumers alike. However, as the enterprise of banking grew and larger banks operated along side small banks, the smaller banks earned a reputation for making more small, rather than large, business loans. Recent trends within the industry are only reinforcing this belief. For example, in 1992, small business lending made up 19.7% of small banks assets, while in 2002 the percentage jumped to 27.4% (FDIC Statistics on Banking). 1 A number of explanations may account for this increase: looser bank credit standards; a vibrant economy; growing numbers of small businesses; bank consolidation and abandoned small business borrowers. Irrespective of the reason, a valid question is whether small banks have assumed too much risk. Akhigbe, McNulty, and Verbrugge (2001) examine this issue and find evidence that between 1986 and 1996, loan quality is lower at small metropolitan statistical area (MSA) banks versus large MSA banks. The purpose of this paper is to investigate what determines loan quality at small banks, and in particular whether local size market structure affects the quality of their commercial loan portfolio; a question not thoroughly investigated in the literature. A model similar to Akhigbe, McNulty, and Verbrugge (2001) is used to examine four measures of loan quality at small banks from 1999 to 2002: nonperforming business loans as a percentage of business loans; loan loss provisions as a percentage of loans; net 1 Small banks are defined as having less than $1 billion in assets and small business lending is measured by commercial and industrial loans and loans secured by nonfarm nonresidential real estate. 2

3 charge-offs on business loans as a percentage of business loans; and other real estate owned as a percentage of loans. Local size market structure is proxied as the percentage of deposits in the MSA held by large banks. The higher this percentage, the fewer small banks operating in the market and the higher the expected loan quality. The model finds local size market structure does affect loan quality. Specifically, small banks in local markets dominated by large banks have higher quality loan portfolios than small banks in local markets dominated by small banks, ceteris paribus. Related Literature Before delving into the hypothesis of the paper, some background on the current state of small business lending by banks is necessary. Not all small business loans are created equal; different sized banking institutions tend to make different types of loans and to use different processes in making lending decisions. Banking literature distinguishes between two types of loan decision processes: relationship loans and ratio loans. A relationship loan refers to a loan made based on both objective and subjective information about the prospective borrower (see Allen et al. 1991, Nakamura 1993, and Berger and Udell 2002). Typically, such loans are made by small banks. In contrast, a ratio loan is one made according to predetermined underwriting procedures and is objective in nature. Not surprisingly, such loans tend to be made by large commercial banks. Why do small banks have a comparative advantage making relationship small business loans? One reason is that small bank decision makers are likely more familiar with local economic conditions as well as the character of the prospective borrower. Conversely, large banks often make credit decisions hundreds of miles from the small 3

4 business in a centrally based credit department. Small banks can also use their proximity to more effectively gather information from suppliers, customers, and even competitors of the firm. Another advantage for small banks is there are fewer chains of command between the loan officer and bank manager, allowing the manager to more easily monitor the loan officer. This organizational structure is conducive for relationship lending since soft information, which is accumulated over time, is difficult to transfer over many layers of management (Berger and Udell 2002). To compensate for the flexible underwriting standards used for relationship loans as well as their higher risk, they typically carry higher interest rates (Berger and Udell 1996). Ratio small business loans, conversely, are more apt to be made by large banks. Ratio borrowers usually have an established history and encounter more standardized underwriting procedures (Berger and Udell 1996, Cole et al. 1997, DeYoung and Hunter 2001). Many are credit scored, underwritten homogeneously, and some even securitized (Mester 1997). Cole et al. (1997) finds financial ratios such as firms debt-to-assets and cash-to-assets to be a statistically significant predictor of whether a loan is approved at large banks, though these variables are insignificant for small banks. Large banks, due to their superior technology, have a comparative advantage making ratio loans in mass at relatively low rates. Large banks also make these loans out of necessity since it is easier to monitor loan officers in a large, complex organization when a consistent loan approval system based on readily available hard information is implemented. However, because of the relatively minimal and inflexible underwriting requirements associated with credit scored loans, credit is usually limited to less than $100,000 per customer (Mester 1997). Levonian (1997) found large banks in the western 4

5 U.S. have increased their holdings of small business loans with original amounts under $100,000 over 26 percent from June 1995 to June 1996 while other banks (small and medium) in the same region increased their holdings only about 3 percent. The same study finds that small business loans between $100,000 and $1 million shrank at large banks by 5 percent from 1995 to 1996, but increased by 7 percent at other banks. A more recent study by Ely and Robinson (2001) found small banks ratio of small business loans for less than $100,000 to total loans shrunk 16.15% from 1994 to 1999 compared to only 4.73% for large banks. However, the ratio of small business loans for $250,00 to $1 million to total loans grew for small banks by 22.12% over this period and shrunk by 8.09% for large banks. It appears small banks are finding a niche in the large small business loan market and the large banks in the small small business loan market (which are more likely to be ratio loans). Given small banks have the comparative advantage making relationship loans, the hypothesis of this study is that small banks located in local markets dominated by large banks have the pick of the litter with respect to relationship borrowers, and therefore make higher quality loans. Berger et al. (2002) lend some support to this hypothesis. They find small businesses are less likely to take discounts on trade credit in markets overwhelmingly dominated by large banks. They view this as a sign of bank credit constraint, and many of the constrained small businesses are likely relationship borrowers. Not only will small banks experience high demand from relationship borrowers in these markets, they may also have an incentive to be more discerning with respect to whom they lend to. Petersen and Rajan (1994) find banks in a competitive environment make fewer loans to unestablished businesses since banks have difficulty 5

6 recouping initial concessionary rates later when the business matures. A similar scenario may play out here; small banks may be reluctant to lend to unestablished small businesses in markets where many large banks are vying to provide loans to mature businesses. As a result of these forces, small banks are predicted to have a higher quality loan portfolio in markets with relatively many large banks. In local markets dominated by small banks conversely, relationship borrowers have more banks to choose from, and small banks are predicted to make riskier loans, ceteris paribus. 2 In other related literature, a number of scholars explore whether market concentration affects bank risk. Most of them conclude higher concentration levels (less competition) are consistent with lower bank risk. Edwards and Heggestad (1973), observe large banks between 1954 and 1966 and find risk decreases as local concentration rises. Keeley (1990) also examines large banks, but tests how easing state branching restrictions (and presumably increasing competition) affects risk. He finds lower market-to-book and equity ratios as well as higher costs of capital for banks in states which have eased state branching restrictions. The results of Jayarante and Strahan (1998, 1999) conflict with Keeley (1990). Examining all banks between 1982 and 1992, they find deregulation lowers statewide nonperforming loans as a percentage of statewide total loans. Five studies, to our knowledge, specifically examine risk (or risk-adjusted returns) at small banks. Rhoades and Rutz (1982) find variance of earnings for unit banks decrease as local concentration rises. Similarly, Bergstresser (2002) using data from 1980 through 1994, finds small banks located in highly concentrated MSAs make fewer 2 Berger et. al. (2001) were the first, to our knowledge, to measure local market structure using percentage 6

7 construction loans (a proxy for risky loans). Ergungor (2002) examines risk-adjusted returns at community banks from 1996 through He finds small business lending increases return on assets (ROA), but reduces community banks risk-adjusted ROA. Moreover, he finds an insignificant relationship between ROA (controlling for risk) and market median bank size (his proxy for size structure of the local banking market) using stepwise OLS. In a similar study, Carter, McNulty, and Verbugge (2002) compare riskadjusted returns on business loans at large and small banks from 1996 to As in Ergungor (2002), they find risk-adjusted yields fall as the proportion of small business loans to total loans rise at all banks. Somewhat surprisingly, however, they find small MSA banks have higher risk-adjusted yields even though they have a higher proportion of small business loans. They reason small banks have an information advantage in evaluating credit. As for market concentration, they find a consistently positive, but rarely significant, relationship between it and risk adjusted yields in their estimated regressions. Akhigbe, McNulty, and Verbrugge (2001), also compare large and small banks. They hypothesize small banks loan portfolios are of higher quality given they have an information advantage over large banks. However, observing a sample of Florida banks from 1986 to 1996, they find small MSA banks have loan portfolios of lesser quality, ceteris paribus, whereas small non-msa banks have higher quality loan portfolios. They also test whether market concentration affects loan quality, and find no significant relationship between the two. The variables they found which significantly affect the quality of a bank s loan portfolio besides bank size and location are: county of deposits held by large banks. 7

8 unemployment rate, the ratio of the bank s real estate loans to total loans, and the passage of the 1991 Federal Deposit Insurance Corporation Improvement Act. All in all, the research generally finds higher concentration levels are consistent with lower small bank risk or higher risk-adjusted returns. Rhoades and Rutz (1982) and Bergstresser (2002) find a significant relationship between the two, while Carter, McNulty, and Verbugges (2002) results are significant in only some of their tests. Meanwhile, Ergungor (2002) and Akhigbe, McNulty, and Verbrugge (2001) do not find a significant relationship. The purpose of this paper is to further explore what affects risk at small banks, and in particular how local size market structure affects business loan quality at small banks. What differentiates this study is the measure of local market structure the percentage of local bank deposits held at small banks (as opposed to the market Hirschman-Herfindahl Index (HHI) and market median bank size used in the previous studies). It is hypothesized that small banks located in local markets dominated by large banks realize a higher quality commercial loan portfolio. Data and Methodology To determine if size market structure affects the quality of business loans at small banks (those with less than $1 billion in assets), this study utilizes several data sources. Small bank characteristics and loan quality measures come from the FDIC s quarterly Report of Condition and Income (Call Reports) beginning with the first quarter of 1999 through the second quarter of Local market structure data are taken from FDIC s Summary of Deposits. The Bureau of Economic Analysis and the Bureau of Labor Statistics provide data for local economic conditions. Following prior research, we define a local market as the Metropolitan Statistical Area (MSA). Recent studies have 8

9 shown small businesses are expanding their geographical market for accessing credit (Cyrnak and Hannan 2000, Petersen and Rajan 2002). Nonetheless, nearly 70 percent of loans are still accessed from financial institutions within 30 miles of the small business (Wolken and Rohde 2002). As a result, we are excluding MSAs with more than 3600 square miles in land area. 3 The model of this study is similar to Akhigbe, McNulty, and Verbrugge (2001) with some minor nuances. The dependent variables attempt to capture loan quality while the independent variables capture size market structure, bank characteristics, and MSA-level economic conditions. While the independent variables are the same for each regression model, four different measures of loan quality are utilized as the dependent variable. Thus, there are four different models of loan quality though each contains the same set of independent variables. The independent variables are lagged one period because it takes time for variable change to affect loan quality and to minimize potential simultaneity bias in the regression. The four measures of loan quality are: total loan loss provisions as a percentage of total loans (LLP); nonperforming business loans as a percentage of business loans (NPL) 4 ; net business charge-offs as a percentage of business loans (CHGOFF) 5 ; and other real estate owned as a percentage of total loans (OREO). 3 However, we include the MSAs of Philadelphia (3,856 square miles), Denver (3,761 square miles), Detriot (3,897 square miles), and Los Angeles (4,060 square miles) to more thoroughly represent heavily populated MSAs. 4 Business loans are defined as: commercial and industrial loans plus loans secured by nonfarm, nonresidential real estate. Nonperforming loans are the sum of: loans past due 30 through 89 days and still accruing; loans past due 90 days or more and still accruing; and nonaccrual loans. 5 Net charge-offs are charge-offs minus recoveries. 9

10 The independent variables may be characterized as MSA-level economic conditions, bank characteristics, and local market structure. Three variables capture local economic conditions: POPGROW is the percentage change in MSA population from 1990 to The MSA 2000 unemployment rate (UNRATE) and per capita income (PERCAP) are included to capture local macroeconomic conditions that may impact loan quality. The two bank characteristics that the literature identifies as important in explaining loan quality are the ratio of commercial loans to total loans (BLN/TLN) and the ratio of real estate loans to total loans (RELN/TLN). Finally, and most important for this study, is the percentage of deposits in the MSA held by large banks in 2001, PERCDEP, which is the proxy for size market structure. 6 A negative relationship between PERCDEP and loan quality is expected. In other words, as the percentage of large banks share of MSA deposits rise, quality loans at small banks increase since the few small banks in these market will have the pick of the litter with respect to relationship borrowers. We can more formally write the regression models as: 6 The FDIC s Summary of Deposits nd Quarter is used to calculate this variable. Since banks with $1 billion in assets held deposits of 82.2% of assets on average, $822 million in deposits is used as a proxy for banks with less than $1 billion in assets (FDIC Statistics on Banking, nd Quarter). 10

11 (1) LLP it = β 0 + β 1 POPGROW it + β 2 UNRATE it + β 3 PERCAP it + β 4 BLN / TLN it + β 5 RELN / TLN it + β 6 PERCDEP it + µ (2) NPL it = β 0 + β 1 POPGROW it + β 2 UNRATE it + β 3 PERCAP it + β 4 BLN /TLN it + β 5 RELN / TLN it + β 6 PERCDEP it + µ (3) CHGOFF it = β 0 + β 1 POPGROW it + β 2 UNRATE it + β 3 PERCAP it + β 4 BLN / TLN it + β 5 RELN / TLN it + β 6 PERCDEP it + µ (4) OREO it = β 0 + β 1 POPGROW it + β 2 UNRATE it + β 3 PERCAP it + β 4 BLN / TLN it + β 5 RELN / TLN it + β 6 PERCDEP it + µ where i refers to the MSA and t refers to the quarter. The definition, mean, and standard deviation for all variables are found in Table 1. As indicated above, the data consists of 14 quarters of pooled cross section data; available bank data for each MSA is aggregated for each quarter. Consequently, for each MSA there are 14 observations; one for each quarter. Much of the data comes from Call Reports, which are reports filed voluntarily by banks. Banks that do not make commercial loans are excluded, as are banks that have omitted data pertinent to the model. Because each bank does not always report for each quarter, the pooled data is not balanced. This, in turn, means that the appropriate regression technique must take into account the unbalanced nature of the data set. Thus, feasible generalized least squares (FGLS), also known as the estimated generalized least squares, is utilized. The FGLS is estimated using the Cochrane-Orcutt method (see Table 2). Further, the pooled data is not serially correlated as evidenced by the Durbin-Watson tests (see Table 2) and there is no evidence of heteroskedasticity. 11

12 Empirical Results Table 2 contains the estimated coefficients for all four loan quality regressions. The only variable to be statistically significant in all four models is the ratio of real state to total loans. The positive estimated coefficient in three of these suggests that increased holdings of real estate loans leads to a reduction in loan quality, as expected. Two macroeconomic variables, per capita income and the unemployment rate, are statistically significant in three of the four models. The positive estimated coefficient on UNRATE indicates that higher levels of unemployment hurt loan quality which is also expected. The negative estimated coefficient in PERCAP indicates that increases in per capita income improve the quality of MSA level bank loans. Most important for this study is the fact that the other variable to be statistically significant in three of the four models is the size market structure proxy. The negative estimated coefficient to PERCDEP suggests that in markets with fewer small banks, loan quality at small banks improves when measured by LLP, NPL, and CHGOFF, ceteris paribus. It is interesting to note that the size market structure proxy is statistically insignificant only in explaining variation in the ratio of other real estate owned to total loans; the OREO model. This is not surprising given that the other measures of loan quality tend to directly capture loans that have gone bad and are consequently certainly of poor quality. OREO, on the other hand, perhaps is more of a measure of bank risk than it is of loan quality. Conclusions Although a significant amount of consolidation has all ready occurred in the banking industry within the last 20 years, there is likely more to come, albeit at a decreasing rate. 12

13 The results of this study suggest as there are fewer small banks, their competitive position should improve. We find in MSA markets dominated by large banks, small banks commercial loan portfolios are of higher quality, ceteris paribus. The results suggest large and small banks make different types of small business loans, consistent with previous research, as small banks in markets dominated by large banks have the pick of the litter with respect to relationship borrowers and realize a higher quality loan portfolio. The end result is small banks will likely continue to carve a niche for themselves within the banking industry in the future. Large banks may very well squeeze this niche since they will be able to more confidently lend to marginally creditworthy small businesses as credit scoring models improve. These models will not capture all the soft information used to make relationship loans, however, thus preserving small banks role. Nonbank lenders may also squeeze small banks niche, but assuming the majority of these lenders are located in markets dominated by large banks, the results of this study suggest small banks can meet this challenge. Future research is warranted. Comparing the results of this study to previous studies examining risk and market structure is difficult. Most of the existing studies measure local market structure differently from this study by considering how much the market is dominated by any group of banks (i.e., HHI). In contrast, our study measures market structure by how much the market is dominated by large banks. Ergungor (2002) uses a measure similar to ours, but relates local market median size bank to small bank risk-adjusted ROA, and not risk specifically. He finds an insignificant relationship between the two. An interesting future study would be to relate our measure of market structure to small banks risk 13

14 adjusted returns. Our results are, nonetheless, consistent with the majority of the previous research which finds higher concentration levels are consistent with lower bank risk, the main difference being measure of market structure. 14

15 Table 1. Descriptive Statistics for Loan Quality Models Variable Definition Mean Standard Deviation Loan loss provisions as a percentage of total loans (LLP) Nonperforming business loans as a percentage of business loans (NPL) Net Business charge-offs as a percentage of business loans (CHGOFF) Other real estate owned as a percentage of total loans (OREO) Percentage change in MSA population from 1990 to 2000 (POPGROW) 2000 MSA unemployment rate (UNRATE) 2000 MSA per capita income (PERCAP) Commercial loans to total loans (BLN/TLN) Real estate loans to total loans (RELN/TLN) Percentage of deposits in MSA held by large banks in 2001 (PERCDEP) , ,

16 Table 2. FGLS Regression Results for Loan Quality Models Independent Variables LLP model Estimated coefficient POPGROW (1.55) UNRATE (2.81)* PERCAP (1.94)** BLN/TLN (0.543) RELN/TLN (2.504)* PERCDEP (3.61)* NPL model Estimated coefficient (1.02) (1.99)** (2.02)** (1.44) (4.36)* (2.11)* CHGOFF model Estimated coefficient (0.925) (3.04)* (0.584) (0.966) (1.98)** (2.84)* OREO model Estimated coefficient (0.842) (1.66) (2.02)** (0.857) (2.34)* (0.948) R-squared DW Notes: ** significant at the 10% level. * significant at the 5% level. Absolute values of the t statistic appear in parentheses. 16

17 References Akhigbe, A., McNulty, J., and Verbugge, J., "Small Bank Loan Quality in a Deregulated Environment: The Information Advantage Hypothesis," Journal of Economics and Business 53, Allen, L., A. Saunders, and G. F. Udell, The Pricing of Retail Deposits: Concentration and Information, Journal of Financial Intermediation, pp Berger, A., N. Miller, M. Petersen, R. Rajan, and J. Stein, "Does Function Follow Organizational Form? Evidence From the Lending Practices of Large and Small Banks," Working Paper. Berger, A., and G.F. Udell, Forthcoming Small Business Credit Availability and Relationship Lending: The Importance of Bank Organizational Structure, Economic Journal. Berger, A., R.J. Rosen, and G.F. Udell, The Effect of Market Size Structure on Competition: The Case of Small Business Lending, Working Paper 63, Federal Reserve Board, Finance and Discussion Series., Universal Banking and the Future of Small Business Lending, edited by A. Saunders and I. Walter, Financial System Design: The Case for Universal Banking, Burr Ridge, IL, Irwin Publishing, Bergstresser. D., "Market Concentration and Loan Portfolios in Commercial Banking," Working Paper. Carter, D., J. McNulty, and J. Verbugge, "Do Small Banks have an Advantage in Lending? An Examination of Risk-Adjusted Yields of Business Loans at Large and Small Banks," unpublished manuscript. Cole, R.A., L.G. Goldberg, and L.J. White, Cookie-Cutter versus Character: The Micro Structure of Small Business Lending by Large and Small Banks, manuscript, Berman and Company, University of Miami, and New York University. Cyrnak, A., and T.H. Hannan Non-local Lending to Small Businesses, Federal Reserve Board Working Paper. DeYoung, R., and W. C. Hunter Deregulation, the Internet, and the Competitive Viability of Large Banks and Community Banks, Working Paper , Federal Reserve Bank of Chicago. Edwards, F., A. Heggestad, Uncertainty, Market Structure, and Performance: The Galbraith-Caves Hypothesis and Managerial Motives in Banking. Quarterly Journal of Economics 87, Ely, D., and K. Robinson, Consolidation, Technology and the Changing Structure of Bank s Small Business Lending, Economic and Financial Review Federal Reserve Bank of Dallas, First Quarter,

18 Ergungor, O., "Community Banks as Small Business Lenders: The Tough Road Ahead," Working Paper 02-03, Federal Reserve Bank of Cleveland. Jayaratne, J., and P. Strahan, Entry Restrictions, Industry Evolution, and Dynamic Efficiency: Evidence from Commercial Banking. Journal of Law and Economics 41, , The Benefits of Branching Deregulation. Federal Reserve Bank of New York Policy Review, 3, Keeley, M Deposit Insurance, Risk, and Market Power in Banking. American Economic Review 80, Levonian, M. E., Changes in Small Business Lending in the West, Federal Reserve Bank of San Francisco Economic Letter. Mester, L., What s the Point of Credit Scoring? Federal Reserve Bank of Philadelphia Business Review, September/October, Nakamura, L. I., Commercial Bank Information: Implications for the Structure of Banking, Michael Klausner and Lawrence J. White, eds., Structural Change in Banking, Homewood, Ill.: Irwin. Petersen, M., and R. Rajan, The Benefits of Firm-Creditor Relationships: Evidence from Small Business Data, Journal of Finance 49, The Information Revolution and Small Business Lending: Does Distance Still Matter? Journal of Finance, forthcoming. Rhoades, S., R. Rutz, Market Power and Firm Risk, a Test of the Quiet Life Hypothesis. Journal of Monetary Economics 9, Wolken, J.D., and D. Rohde Changes in the Location of Small Businesses Financial Service Suppliers between 1993 and 1998, Federal Reserve Board internal memorandum. 18