ARE GOOD OR BAD BORROWERS DISCOURAGED FROM APPLYING FOR LOANS? EVIDENCE FROM US SMALL BUSINESS CREDIT MARKETS. Working Paper No.



Similar documents
An Empirical Analysis of Insider Rates vs. Outsider Rates in Bank Lending

Who Needs Credit and Who Gets Credit? Evidence from the Surveys of Small Business Finances

Warwick Business School

Collateral and relationship lending in loan pricing: Evidence from UK SMEs

Factors Determining Bank Debt vs Bond Debt of Canadian Corporations

IS THERE ETHNIC DISCRIMINATION IN THE UK MARKET FOR SMALL BUSINESS CREDIT? Working Paper No. 96 April Stuart Fraser

TERMS OF LENDING FOR SMALL BUSINESS LINES OF CREDIT: THE ROLE OF LOAN GUARANTEES

WOMEN-OWNED BUSINESSES AND ACCESS TO BANK CREDIT: A TIME SERIES PERSPECTIVE

Small Business Borrowing and the Owner Manager Agency Costs: Evidence on Finnish Data. Jyrki Niskanen Mervi Niskanen

Loan Officer Turnover and Credit Availability for Small Firms

THE INCIDENCE OF PERSONAL COLLATERAL IN SMALL BUSINESS LENDING.

Gender and the Availability of Credit to Privately Held Firms: Evidence from the Surveys of Small Business Finances

THREE ESSAYS IN APPLIED FINANCE SENA DURGUNER DISSERTATION

Do Banks Price Owner Manager Agency Costs? An Examination of Small Business Borrowing*

Entrepreneurial Optimism, Credit Availability, and Cost of Financing: Evidence from U.S. Small Businesses

Mitigating Information Asymmetries through Collateral Pledges

Does Distance Still Matter? The Information Revolution in Small Business Lending

NBER WORKING PAPER SERIES BANKRUPTCY AND SMALL FIRMS ACCESS TO CREDIT. Jeremy Berkowitz Michelle J. White

Does Market Size Structure Affect Competition? The Case of Small Business Lending

This version: August 12, 2010

Women-Owned Businesses and Access to Bank Credit: Evidence from Three Surveys Since 1987

The Determinants of Collateral: a Decision Tree Analysis of SME Loans

DIVIDEND POLICY, TRADING CHARACTERISTICS AND SHARE PRICES: EMPIRICAL EVIDENCE FROM EGYPTIAN FIRMS

Borrower-lender distance and its impact on small business lenders during the financial crisis

Bank Credit, Trade Credit or No Credit: Evidence from the Surveys of Small Business Finances

Kauffman Dissertation Executive Summary

Does Distance Still Matter? The Information Revolution in Small Business Lending

Determinants of Capital Structure in Developing Countries

Chapter 14. Understanding Financial Contracts. Learning Objectives. Introduction

Adverse Selection in the Credit Card Market: Evidence from Randomized Trials of Credit Card Solicitations. Abstract

Agribusiness Trade Credit - A Paradox

Literature Review. Introduction

Entrepreneurial Optimism, Credit Availability, and Cost of Financing: Evidence from U.S. Small Businesses

1 Determinants of small business default

The Use of Trade Credit by Businesses

Credit Scores and Credit Market Outcomes: Evidence from the Survey of Small Business Finances and the Kauffman Firm Survey

The Determinants and the Value of Cash Holdings: Evidence. from French firms

American Finance Association

Ownership and Asymmetric Information Problems in the Corporate Loan Market: Evidence from a Heteroskedastic Regression.,

Finance for Small and Medium-Sized Enterprises

COMPETITIVE AND SPECIAL COMPETITIVE OPPORTUNITY GAP ANALYSIS OF THE 7(A) AND 504 PROGRAMS

Federal Reserve Bank of Chicago

Discrimination in Access to Finance: Evidence from the United States Small Business Credit Market

ORGANIZATIONAL DISTANCE AND USE OF COLLATERAL FOR BUSINESS LOANS. Documentos de Trabajo N.º 0816

The Importance of Adverse Selection in the Credit Card Market: Evidence from Randomized Trials of Credit Card Solicitations

Econometric Analysis from the UK Survey of SME Finances

Have the GSE Affordable Housing Goals Increased. the Supply of Mortgage Credit?

Small Bank Comparative Advantages in Alleviating Financial Constraints and Providing Liquidity Insurance over Time

BANKS AND SMES: EMPIRICAL QUANTITATIVE APPROACH ON BANKS BEHAVIOUR AS LENDERS TO SMALL BUSINESSES DURING CRISIS TIMES

A Preliminary Causal Analysis of Small Business Access to Credit during Economic Expansion and Contraction

Access to Capital among Young Firms, Minority-owned Firms, Women-owned Firms, and High-tech Firms

Econometric analyses using data from the UK Survey of SME Finances and the SME Finance Monitor

Medical Bills and Bankruptcy Filings. Aparna Mathur 1. Abstract. Using PSID data, we estimate the extent to which consumer bankruptcy filings are

Measuring BDC s impact on its clients

Center for Economic Institutions Working Paper Series

Discussion Papers on Entrepreneurship, Growth and Public Policy

Corporate Finance and Econometric Analysis

Business collateral and personal commitments in SME lending.

Impact of Receivership Costs on the Optimal Capital Structure for Small Businesses

Bankruptcy and Small Firms Access to Credit

``New'' data sources for research on small business nance

The Role of Loan Guarantee Schemes in Alleviating Credit Rationing in the UK

SME Credit Availability Around the World: Evidence from the World Bank s Enterprise Survey

Statistics in Retail Finance. Chapter 2: Statistical models of default

Collateral, Type of Lender and Relationship Banking as Determinants of Credit Risk

Lines of Credit and Relationship Lending in Small Firm Finance

Regression Analysis of Small Business Lending in Appalachia

How Do Small Businesses Finance their Growth Opportunities? The Case of Recovery from the Lost Decade in Japan

econstor zbw

trade credit vis-à-vis bank debt. 4 Fishamn and Love (2003) and Demirgüç and Maksiovic (2001) pointed that trade credit is more prevalent in

Banks and the Role of Lending Relationships: Evidence from the U.S. Experience

EVALUATING CHANGES IN BANK LENDING TO UK SMES OVER ONGOING TIGHT CREDIT?

CREDIT MARKET COMPETITION, COLLATERAL AND FIRMS' FINANCE. Gabriel Jiménez, Vicente Salas and Jesús Saurina. Documentos de Trabajo N.

During the last couple of years, concern

Small Firms in the Credit Crisis: Evidence from the UK Survey of SME Finances.

The Role of Loan Guarantee Schemes in Alleviating Credit Rationing in the UK

Financial Capital Injections among New Black and White Business Ventures: Evidence from the Kauffman Firm Survey

Lending to small businesses: the role of loan maturity in addressing information problems *

Improving Access to Credit for SMEs: An Empirical Analysis of the Viability of Pooled Data SME Credit Scoring Models in Brazil, Colombia & Mexico

What Do We Know about the Capital Structure of Privately Held US Firms? Evidence from the Surveys of Small Business Finance

Accounts Receivable and Accounts Payable in Large Finnish Firms Balance Sheets: What Determines Their Levels?

INTERNET APPENDIX TIME FOR A CHANGE : LOAN CONDITIONS AND BANK BEHAVIOR WHEN FIRMS SWITCH BANKS. This appendix contains additional material:

A COMPARISON OF NON-PRICE TERMS OF LENDING FOR SMALL BUSINESS AND FARM LOANS

The Impact of Entrepreneurs Personal Wealth Allocations in Determining Their Firms Capital Structures

How and why do small firms manage interest rate risk? a

An Empirical Investigation of Trade Credit Use: A Note

W. Scott Frame Financial Economist and Associate Policy Advisor Federal Reserve Bank of Atlanta Atlanta, GA U.S.A.

CHAPTER 2 LOAN DEMAND IN APPALACHIA

EIB PAPERS. Europe s changing financial landscape: The financing of small and medium-sized enterprises. Volume 8 No

Board of Governors of the Federal Reserve System. International Finance Discussion Papers. Number December 2013

Credit Card Market Study Interim Report: Annex 4 Switching Analysis

Market size structure and small business lending: Are crisis times different from normal times?

Public Policy in Support of Small Business: The American Experience

Issue Brief. Access to Capital for Women- and Minority-owned Businesses: Revisiting Key Variables. Advocacy: the voice of small business in government

General guidelines for the completion of the bank lending survey questionnaire

Kauffman. Firm Survey. The Use of Credit Card Debt by New Firms. August Sixth in a series of reports using data from the Kauffman Firm Survey

Bankruptcy and small firms access to credit

FAIR TRADE IN INSURANCE INDUSTRY: PREMIUM DETERMINATION OF TAIWAN AUTOMOBILE INSURANCE. Emilio Venezian Venezian Associate, Taiwan

1. State and explain two reasons why short-maturity loans are safer (meaning lower credit risk) to the lender than long-maturity loans (10 points).

Response to Critiques of Mortgage Discrimination and FHA Loan Performance

Transcription:

ARE GOOD OR BAD BORROWERS DISCOURAGED FROM APPLYING FOR LOANS? EVIDENCE FROM US SMALL BUSINESS CREDIT MARKETS Working Paper No. 95 March 2008 Liang Han, Stuart Fraser and David J Storey

Warwick Business School s Small and Medium Sized Enterprise Centre Working Papers are produced in order to bring the results of research in progress to a wider audience and to facilitate discussion. They will normally be published in a revised form subsequently and the agreement of the authors should be obtained before referring to its contents in other published works. The Director of the CSME, Professor David Storey, is the Editor of the Series. Any enquiries concerning the research undertaken within the Centre should be addressed to: The Director CSME Warwick Business School University of Warwick Coventry CV4 7AL e-mail david.storey@wbs.ac.uk Tel. 024 76 522074 ISSN 0964-9328 CSME WORKING PAPERS Details of papers in this series may be requested from: The Publications Secretary CSME Warwick Business School, University of Warwick, Coventry CV4 7AL e-mail sharon.west@wbs.ac.uk Tel. 024 76 523692 1

ARE GOOD OR BAD BORROWERS DISCOURAGED FROM APPLYING FOR LOANS? EVIDENCE FROM US SMALL BUSINESS CREDIT MARKETS Liang Han (Hull University Business School) Stuart Fraser (Warwick Business School) David J. Storey (Warwick Business School) Abstract This paper conducts an empirical analysis on a U.S small business finance data-set to investigate which factors affect the likelihood of being a discouraged borrower. We find that riskier borrowers have higher probabilities of being discouraged. The results suggest that, in the US, discouragement is an efficient self-rationing mechanism, in that high risk borrowers are more likely to be discouraged than low risk borrowers, and the efficiency of this mechanism increases as information asymmetries are resolved. We also report that low risk borrowers are less likely to be discouraged in concentrated markets than in competitive markets; and, in concentrated markets, high risk borrowers are more likely to be discouraged the longer their financial relationships. These results suggest that discouragement is more efficient in concentrated markets than in competitive markets. There is little evidence to suggest that application costs discourage small businesses from borrowing. JEL: G14 G21 Keywords: Discouraged Borrowers; Self-rationing; Information Problem Corresponding author: L.Han@hull.ac.uk. 2

ARE GOOD OR BAD BORROWERS DISCOURAGED FROM APPLYING FOR LOANS? EVIDENCE FROM US SMALL BUSINESS CREDIT MARKETS 1. Introduction The problem of imperfect and asymmetric information lies at the heart of financing small businesses (e.g. Berger and Udell, 1998). This is because small firms are recognised as more informationally opaque than large firms, and the collection of private information, such as the risk type of small business borrowers, is costly (Ang, 1991). In providing finance for small firms, bank lenders are generally assumed to have poorer information about the individual small business than the borrower 1. Theoretically, faced by asymmetric information, banks could either ration credit (Stiglitz and Weiss, 1981) or offer a menu of contracts which act as a self-selection mechanism to distinguish good, from bad, borrowers (Bester, 1985). However, in conditions of imperfect information, amongst potential small business borrowers, some do not apply for bank loans even if they need capital. This is because they think their application will be rejected. Such borrowers are called Discouraged Borrowers. Kon and Storey (2003) define discouraged borrowers as those creditworthy (good) borrowers who do not apply because they feel they will be rejected. In this paper, we extend the scope of discouraged borrowers by including all types of borrowers (both good and bad) who experience discouragement. There are two principal reasons for doing this. Firstly, in purely practical terms, it is difficult to make a clear empirical distinction between good and bad borrowers since the definition of a creditworthy borrower may vary from lender to lender depending on risk tolerance levels. Secondly, and relating to the 1 Jovanovic (1982), Storey (1994) and De Meza and Southey (1996) argue that the information advantage of borrowers is created through a learning-by-doing process. New entrepreneurs are likely to be less well informed than established entrepreneurs. External lenders, e.g. banks, however have considerable experiential knowledge of new firms. 3

principal aim of the paper, in order to test whether discouragement is an efficient selfrationing mechanism, a sample which includes both good and bad borrowers in the discouraged group is required. According to the theory developed by Kon and Storey (2003) the reason discouraged borrowers exist is because of information asymmetries and positive application costs. They argue that banks do not know borrower types - either good (low risk) or bad (high risk). If they did, then they would charge an appropriate premium. But discouragement can also be viewed as a self-rationing mechanism by which potential borrowers make an application decision. This paper acknowledges that bad borrowers can also be discouraged so that discouragement is a good thing where the bad are discouraged. Banks do not care if bad borrowers are discouraged; but they do care if good borrowers are discouraged and/or if bad borrowers get into the loan pool. So, in terms of discouraged borrowers, we define self-rationing as efficient when it discourages bad borrowers but not good borrowers. Hence, with efficient self-rationing, good borrowers will be less likely to be discouraged than bad ones. Furthermore, in a fully informed market, all bad borrowers are discouraged from applying. This paper conducts both univariate and multivariate analysis on data from the 1998 U.S Survey of Small Business Finances (SSBF) and finds that, after controlling for the characteristics of both the business and entrepreneur, riskier borrowers are more likely to be discouraged. With the improvement of information quality, as financial relationships become longer, riskier borrowers have an increasing likelihood of being discouraged. These results suggest that discouragement is an efficient self-rationing mechanism. We also report that low risk borrowers are less likely to be discouraged in concentrated markets than in competitive markets and in concentrated market, high risk borrowers are 4

more likely to be discouraged over longer relationships, suggesting that discouragement is more efficient in concentrated markets than in competitive markets. However, we find little evidence that application costs are a key determinant of discouragement amongst small businesses. The rest of the paper is structured as follows. Section 2 reviews both theoretical and empirical literature related to discouraged borrowers. Section 3 introduces the data and the methodology employed in this paper. Section 4 reports the empirical findings and Section 5 concludes. 2. Discouraged Borrowers Theories and Empirics Discouraged borrowers have drawn insufficient attention from researchers until very recently. This is partly because they do not influence risk levels in banks loan portfolios and partly because information on them is difficult for banks to obtain. However, their importance is now increasingly recognised in both theoretical and empirical work for three reasons. Firstly, it has been found that discouragement leads to financial constraints for small businesses as they are more likely to report discouragement than rejection (Levenson and Willard, 2002). Secondly, discouragement may point to discrimination in the market for small business finance. In this context, it has been reported that the likelihood of being discouraged varies with the ethnic background of the entrepreneur both in the US (Cavalluzzo et al. 2002) and in the UK (Fraser, 2007) with discouragement being more likely amongst ethnic minority groups than white owned businesses. Thirdly, the examination of discouraged borrowers is actually a test of the lending efficiency of financial institutions in financing small businesses in terms of screening errors and application costs (Kon and Storey, 2003). 5

According to the theory proposed by Kon and Storey (2003), one of the most important determinants of discouragement, which is of crucial concern to lenders, is the unobservable quality of borrowers. Ideally, lenders would like to encourage good borrowers and discourage bad borrowers, but they do not know, or do not know exactly, the borrower s quality because of information asymmetries. In the empirical literature, borrower quality is measured in several ways. For example it is proxied by a Dun and Bradstreet Score (Cavalluzzo et al. 2002), or by the variance of returns to equity (Booth and Booth, 2006). Imperfect information therefore lies at the heart of the concept of discouraged borrowers and the acquisition of reliable information from informationally opaque small business borrowers is a concern to lenders. Petersen and Rajan (2002) use business credit card and credit lines to measure the information transparency of the business. Business credit card holders and lines of credit users are argued to be informationally transparent because their creditworthiness has been assessed in the external credit market. Relationship lending plays an important role in alleviating the adverse consequences of information asymmetries for small business borrowers. Indeed, relationships are argued to improve the availability (Petersen and Rajan, 1994) and reduce the cost (Berger and Udell, 1995) of small business finance. However, on the downside for borrowers, relationships may yield a rent to banks and as a result, loan rates may increase with the length of the relationship (Degryse and Ongena, 2005). Apart from the effects of the length of relationships on small business finance, the type of financial service supplied (e.g. Petersen and Rajan, 1994; Carey et al. 1998; Cowling, 1999) and the distance between banks and their customers (e.g. Petersen and Rajan, 2002; Degryse and Ongena, 2005) have previously been identified as influencing credit access and price. 6

Another important determinant in Kon and Storey s model of discouraged borrowers are screening errors made by lenders which again arise due to information asymmetries. In this regard, some types of lenders have an advantage in terms of private information acquisition. Banks, for instance, may collect information by monitoring transactions on current accounts held by borrowers; whilst, non-financial lenders and venture capitalists do not provide such services to their small business customers. Banks are the most important and comprehensive provider of financial services for small businesses (Bilter et al. 2001). Consequently, banks may have higher acceptance rates on loan applications from small businesses because of access to information on the businesses use of other financial services provided by the bank 2. Another possible reason is that small firms have lower application costs when they apply for credit from banks than from other types of institution. Thus, small firms are less likely to be discouraged if they expect to apply for finance from a bank. Physical distance between lenders and borrowers is another key factor in financing small businesses. On the one hand, for information reasons, external financiers prefer small business customers close to them as monitoring costs are lower and moral hazard problems are more easily identifiable. On the other hand, small businesses are discouraged partly because of higher application costs incurred by the physical distance to lenders (Kon and Storey, 2003). However, it is now well documented that small business financing behaviour has been significantly altered with the development of new information technologies. For example, Petersen and Rajan (2002) reported that in the U.S., the probability that a bank communicates with its small business borrowers in person, instead of using telephone or mail, declined from 59% in 1973 to 36% in 1993. 2 We find that 71% of applicants who were always approved borrowed from banks compared with only 29% of those that borrowed from non-bank institutions. Unfortunately, we don t know whether banks have a higher or a lower rejection rate on loan applications. 7

Meanwhile, the average physical distance from a bank to its typical small business borrowers had increased from 16 miles, in the period 1973-1979, to 68 miles in the period 1990-1993. Indeed, online banking now allows small businesses to lower or eliminate some transaction costs, such as those of transport. It also facilitates access to multiple lenders, located at a distance from the business, by lowering search costs. Han (2008) provides empirical support for this using a UK SME finance dataset. Equally, advanced information technologies also enable banks to develop electronic lending and credit decision making systems which substantially lower the costs of lending (Allen et al, 2002). As a result, both banks and their customers benefit from improvements in lending technologies and the quality and variety of banking services (Berger, 2003) to the extent that large banks have increased their share of small business loans (Frame et al, 2001). Overall, the evidence suggests that online banking has had a favourable impact on both the availability and cost of finance for small business borrowers. Another factor which may affect the likelihood of discouragement, and relates to both information issues and the cost of lending, is the extent to which financial relationships are concentrated with a single lender. Small business borrowers may self-select to borrow from a single lender or disperse their borrowing across multiple lenders by comparing the benefits and costs between these two financing strategies. Concentrated relationships may signal higher borrower quality (Bris and Welch, 2005) and help to overcome the free-rider problem (Holmstrom, 1982; Márquez, 2002). Concentration also reduces transaction and negotiation costs. On the other hand, multiple borrowing relationships provide the opportunity for competition between finance providers and reduce the problem of rent extraction which arises where a single lender alone has access to private information about the borrower (von Thadden, 1992). Indeed, it has been found that lower borrowing concentration has a significantly negative effect on the cost of borrowing (Repetto et al, 8

2004). This is because a single lender may charge above the market price on small business loans through a lock-in mechanism (Degryse and Cayseele, 2000). By borrowing from multiple sources, businesses can also insure themselves against liquidity shocks that hit banks so that liquidity risks are reduced (Detragiache et al, 2000). In summary, discouragement can be associated with the demographic profile of the entrepreneur, the quality of the borrower, information issues (including screening errors) and application costs. 3. Data and Methodology 3.1 Data This paper examines empirically the effects of demographic profiles, borrower quality, information issues and application costs on the likelihood of financial discouragement. The data used here is the 1998 U.S. Survey of Small Business Finances (SSBF98). The survey collects information on the use of credit by small firms and creates a generalpurpose database on the finances of such firms, with a target population of all for-profit, non-financial, non-farm, non-subsidiary businesses with fewer than 500 employees. The dataset contains 3,561 sample firms, representing 5.3 million small businesses operating in 1998 in the U.S. In this sample: 2,099 firms did not apply for external finance, in the three years before the survey, because they had no need for it; 962 firms applied for external finance; and 500 did not apply because of fear of rejection. The latter group are defined as discouraged borrowers. That is, around one third of the sample firms with demand for external finance were discouraged borrowers. Hence, as discussed earlier, our definition of discouraged borrowers encompasses all businesses (both high and low risk), with capital demands, but which did not apply because of fear of rejection. In contrast, recall that Kon 9

and Storey (2003) limit their definition of discouraged borrowers to good (low risk) potential borrowers who are discouraged from borrowing by information asymmetries and/or application costs. 3.2 Variables In order to estimate the effects of the determinants of discouragement, we use a logistic estimation procedure 3 on the sample firms with capital demands. The dependent variable is coded as one if the sample venture was discouraged from applying for external funds; zero otherwise. For the independent variables and following on from the previous section - we group these into four categories: (1) characteristics of the principal and venture; (2) borrower quality; (3) application costs; and (4) information issues and the nature of the primary lender, along with control variables representing industry and market concentration. The definitions of the variables, and summary statistics, are reported in Table 1. Table 1: here On average, a typical firm in the sample was family-owned, 11 years old and has 6 employees. It was owned by a 49 years old male owner who has a college degree, 16 years of experience in business and a total personal wealth of half million dollars. 41% of the sample had demand for external capital and 34% of them (14% of the total sample) were discouraged borrowers. One of the key determinants in examining discouraged borrowers is the quality or risk levels of the borrower. Borrower quality can be measured by a Dun and Bradstreet (D&B) Score (Cavalluzzo et al. 2002), or by the variance of returns to 3 Another possibility is to conduct a nested logistic model with the upper level modeling the demands for capital and the lower level modeling the choice between applying and being discouraged. The authors attempted to construct such a model but it failed to converge due to non-concavity in the likelihood function. Our interpretation is that this suggests the simpler models are more appropriate for the data. 10

equity (Booth and Booth, 2006). In this paper, we use instrumented D&B scores to measure the risk levels of borrowers in order to overcome the problem of endogeneity where the credit score is correlated with the error term or unobservables in the discouragement equation. Indeed, the instrumented measure is more appropriate than the real scores because unobserved entrepreneurial talent may affect both discouragement and credit scores, leading to endogeneity. Regarding information issues, we follow Petersen and Rajan (2002) and use business credit card and credit lines to measure the information transparency of the business. We also measure the severity of the information problem by the length of relationship with the primary financial service provider; this measure has been widely used in the existing literature (e.g. Berger and Udell, 1995). In addition, we measure the concentration of creditors by the number of financial service suppliers used by the sample firm. We expect that, as information improves, low risk small businesses are less likely to be discouraged and high risk borrowers are more likely to be discouraged. The third factor influencing discouraged borrowers are application costs. These are examined using a continuous variable, namely the physical distance to the primary lender, and a binary variable, whether the business applied for a loan online. We expect that discouragement is less likely when application costs are low. In Kon and Storey (2003), lenders screening errors are an important determinant of discouragement. However, we cannot directly measure this error because the information used in the lending decision is not available in the survey so we cannot test whether a loan application has been mistakenly rejected or mis-priced. Instead, we use a dummy variable for whether the primary lender is a bank to represent the nature of the institution. As argued earlier, banks may make fewer screening errors because of their advantage in 11

information collection resulting from their ability to monitor other services used by borrowers such as bank accounts. Therefore, we expect that small businesses are less likely to be discouraged when the primary lender is a bank due to the likelihood of fewer screening errors. We also include the demographics of the entrepreneur and business characteristics in regressions. These variables include, for example, the age, education background and personal wealth of the principal owner and the size, capital structure and use of financial products of the business. Finally, we include control variables such as industry and market concentration. Panels A and B (Table 2) report simple continuous and bivariate correlations on the independent variables, respectively. Table 2 shows that none is above 0.70 for continuous variables (Panel A) and apart from the high correlations between the binary instrumented D&B scores, all of others are under 0.30 (Panel B). As a result, we use continuous instrumented D&B scores (INST_DB2) in most of the following regressions and binary ones in a robustness check. This suggests that any problems in the regression models due to multicollinearity are minimal. Table 2: Here 4. Empirical Results 4.1 Univariate Tests We begin the analysis with univariate comparisons of the key determinants of discouragement between small businesses, which applied for loans, and those which were discouraged from borrowing despite having capital demands. The results are shown in Table 3. The first important finding is that discouraged borrowers are riskier than 12

applicants: discouraged borrowers have higher actual and instrumented D&B scores than applicants (p<0.01). The table also indicates that, on average, discouraged borrowers are younger and smaller (measured by the total number of employees) than applicants and are less likely to be incorporated. However, the capital structures of these two groups do not differ significantly in terms of the ratios of total liabilities to total assets. Regarding the demographic profile of the entrepreneur, discouraged borrowers are: lessexperienced (measured by the number of years of experience in business and management); poorer (measured by the total value of the owner s home equity and the net worth of other assets); less likely to have a college degree; and more likely to belong to an ethnic-minority group. Discouraged borrowers also have more concentrated financing relationships than applicants. However there may be an issue of causality here because concentrated relationships may be caused by being discouraged from approaching other lenders. Unfortunately, the analysis cannot fundamentally solve this causal problem because of the limited information available from the data. Table 3 also indicates that: the proportion applying for finance online amongst applicants is more than twice that amongst discouraged borrowers; and applicants are more likely to have a bank as their primary financial service provider. Finally, Table 3 shows that discouraged borrowers do not differ from applicants, at conventional levels of statistical significance, either in terms of the length of their financial relationships or the distances to their primary lender. Table 3: here 13

4.2 Business/Entrepreneur Characteristics and Application Costs We now run four logistic models to examine the determinants of discouraged borrowers and the results are shown in Table 4. In Model 1, we estimate the probability of being discouraged in terms of the characteristics of the entrepreneur and business. We add differently instrumented proxies for borrower s quality in Models 2 and 3 and application costs variables are added in Model 4. Examination of the percent concordant statistic, in the final row of the table, suggests that the percentage of correctly predicted cases of discouragement increases as explanatory variables are added to the model. In each model, dummy variables representing industry and the concentration of local banking markets are included. The effects of market concentration on discouragement are discussed at the end of this section. Table 4: here In Model 1, we include only the characteristics of the business and entrepreneur (in addition to industry and market concentration dummies). The results, reported in the first column, indicate that discouragement is less likely in larger firms and in those businesses which used external finance, such as business credit cards and business credit lines. The existing literature (e.g. Berger and Udell, 1998) suggests that larger businesses are less risky and more informationally transparent. In contrast, smaller firms are relatively informationally opaque which may increase the screening and monitoring costs of external lenders. These costs may make lending to small firms less profitable and hence increase discouragement amongst small business borrowers. As mentioned earlier, the use of some types of financial products, such as business credit cards and loans, may signal a good quality borrower as its creditworthiness has already been examined by existing lenders. Indeed, businesses which lack these good signals are more likely to be discouraged. The 14

negative coefficients on the use of external loans, supports this signalling hypothesis. Other business characteristics, such as age and capital structure/debt level of the business, do not have significant effects on the likelihood of discouragement. Model 1 also shows that the probability of being discouraged is positively associated with ethnic minority, older and poorer groups of entrepreneurs. The coefficient of the ethnic minority variable is positive and statistically significant at the 10% level in Model 1. It becomes insignificant in the following models in which we add risk variables. Indeed, inspection of the correlation between ethnic-minority and risk (see Table 2) shows a positive correlation which is significant at the 1% level suggesting that ethnic-minority owned businesses are riskier. Therefore, the ethnic variable affects the likelihood of discouragement through the risk variables. Lower confidence, amongst older entrepreneurs, may explain their higher rates of discouragement. We also find that an increase in the personal wealth of the entrepreneur significantly reduces the probability of being discouraged. This is because wealthier entrepreneurs are less likely to be discouraged by hard financial reasons, such as collateral/personal guaranty requirements or application costs. This highlights the importance of the personal wealth of the entrepreneur as a primary source of collateral and again emphasises that, for small businesses, the personal wealth of the owner and the assets of the business are difficult to separate (Ang et al. 1995). Models 2 and 3 are run using differently instrumented risk measures. The first measure is a probability weighted average of the risk levels (expected risk level) and is thus a continuous variable (INST_DB2 in Model 2). The other set of measures is a group of four dummy variables representing moderate (base group), average, significant and high risk levels respectively. The use of these measures addresses the endogeneity issue relating to 15

firms risk levels. While we cannot completely eliminate the problem of endogeneity, all our results are nevertheless robust to these alternative specifications. The principal result of these two models is that it is the riskier borrowers that are most likely to be discouraged, implying that discouragement is an efficient self-rationing mechanism. As emphasised earlier the definition of discouraged borrowers formulated by Kon and Storey (2003) is extended to include both good and bad borrowers so as to examine the efficiency of the self-rationing role of discouragement. Our result points to a low, but nevertheless positive, level of market imperfection in US small business financial markets. Only when there is no discouragement amongst good borrowers is the market deemed to be perfect 4. Even so, the opportunities for further improvements seem modest. Application costs are another important potential determinant of discouragement. These costs are proxied by the physical distance to the primary lender and whether loan applications were made online. Model 3 indicates that neither of these two variables have a statistically significant impact on the likelihood of being discouraged. This suggests that application costs are not a key reason for discouragement. 4.3 Information Issues and the Nature of the Primary Lender Table 5 presents the results of three further logit models which include variables representing information issues and the nature of the primary lender/financial service supplier. Apart from the independent variables examined earlier (Table 4), Model 5 (Table 5) considers the effects of the length of relationship between borrowers and the primary lender (LOGRELATION_PI), the concentration of creditors (N_FS) and the nature of the primary lender, i.e. whether or not it is a bank (PITYPE_BANK). Model 6 also includes 4 Kon and Storey also note that good (and bad) borrowers are not discouraged when lottery conditions exist. However this polar extreme is not relevant in an informed credit market such as the US in 1998. 16

an interaction term between risk and relationship lengths (DB2LNR) to examine how the effects of relationships vary over borrowers with different levels of risk. We also report a parsimonious version of Model 6 (Model 7) because only a subset of the explanatory variables in the full equation are statistically significant. For Model 7, marginal effects are also presented. Table 5: here Model 1 shows that small businesses are less likely to be discouraged where the primary lender is a bank. This is compatible with the view that banks have an advantage in private information collection compared to other lenders, due to their provision of other services to borrowers. It is this information advantage which reduces their screening errors. The significantly negative coefficient on the number of sources of financial services suggests that borrowers with dispersed financial relationships are less likely to be discouraged 5. This result is reasonable because a borrower can make multiple or repeated applications from different lenders. Another possible reason is that multiple financing relationships are positively associated with higher quality borrowers. This interpretation is compatible with existing theoretical models (e.g. Bolton and Scarfstein, 1996) and empirical analysis (e.g. Han et al. 2008). The existing literature has highlighted the importance of relationship lending in private information acquisition (e.g. Berger and Udell, 1995) and thus, we expect that borrower s quality would be more easily identified, and banks screening errors would be reduced, by developing a longer relationship. In other words, as relationship lengths increase, high 5 A causal problem may exist here because one can argue that small businesses may have dispersed creditors if they are not discouraged. The information collected in the data does not allow us to examine this causal relationship. 17

(low) quality borrowers would be less (more) likely to be discouraged. Indeed, the results of Models 6 and 7 support our expectation. The most significant difference between Models 5 and 6, in Table 5, is that after adding an interaction term between borrower s risk and relationship lengths, the coefficient of borrower s quality (INST_DB2) is no longer statistically significant; whilst the coefficients of the other independent variables change little. This result, along with the significance of the coefficient of the interaction term, implies that the relationship effects may depend on the borrower s level of risk. Based on the result of the parsimonious model (Model 7), we plot the estimated probabilities of being discouraged (Figure 1), for five hypothetical businesses, over the length of their relationships with the primary lender (holding other variables at either their mean or median). They differ in risk levels only and the curves plotted in Figure 1 are the estimated probabilities of being discouraged over length of relationship, for, from bottom to top, businesses with low risk (DB_SCORE=1), moderate risk (DB_SCORE=2), average risk (DB_SCORE=3), high risk (DB_SCORE=4) and significant risk (DB_SCORE=5), respectively. Figure 1 shows that high quality borrowers, i.e. those of low risk and moderate risks, are less likely to be discouraged by developing a longer relationship with the primary lender. In contrast, low quality borrowers, i.e. those of high and significant risks, are more likely to be discouraged from borrowing as the length of their relationships increase. Figure 1 illustrates that the relationship effects on discouragement depend on borrower risk levels and a longer relationship, between borrower and lender, improves the efficiency of the self-rationing mechanism by discouraging bad borrowers and encouraging good ones. Figure 1: here 18

4.4 Market Concentration Earlier results (Tables 4 and 5) also shows that the concentration of the banking market, where the borrower s headquarters are located, has a strong impact on the likelihood of discouragement. For example, the likelihood of discouragement of a borrower, which has its headquarters located in a moderately concentrated banking market, is nearly 17 percentage point lower than where the headquarters are located in a competitive market (see Model 7, Table 5). This result suggests that discouragement is affected by the degree of concentration in local banking markets. This is consistent with the existing literature which has identified the effects of market concentration on small business finance (see e.g., Cavalluzzo et al, 2002; and Petersen and Rajan, 1995) due essentially to the impact of competition on the market power of lenders (e.g. Degryse and Ongena, 2005; Berger 2004). In particular, Petersen and Rajan (1995) argued that the market condition effects work through relationship lending because the extent of market competition is important in determining the value of lending relationships indeed they found evidence to support the view that it is easier for lenders to internalise the benefits of lending in a more concentrated market. Following this idea, we further examine the effects of market conditions on discouragement by including interaction terms between borrower s quality, relationships and market concentration. This final set of results is reported in Table 6. Table 6: here Table 6 suggests that the likelihood of discouragement is jointly determined by borrower s quality, length of relationship and market concentration. Here, we would expect the probability of discouragement to decrease (resp. increase) amongst good (resp. bad) 19

borrowers as the length of their relationships increase (as shown in Figure 1). However, market concentration will affect the bank s returns from investing in relationships: in a competitive market there is less incentive to invest in relationships which would be expected to reduce the magnitude of the impact of relationships on discouragement. Accordingly, the joint effects of borrower s quality, length of relationship and market concentration are shown in Figures 2-1 to 2-5. These graphs plot the estimated probabilities of being discouraged, over the length of relationships, in competitive, moderately concentrated and highly concentrated banking markets, for borrowers of increasing risk levels. Figures 2-1 and 2-2 show three important results. Firstly, the likelihood of discouragement amongst good borrowers (with risk levels below average) reduces over the length of the relationship regardless of the degree of concentration in the local banking market. Hence, this result is consistent with our earlier findings as shown in Figure 1, suggesting that relationships improve information quality and reduce the likelihood of discouragement for good borrowers. Secondly, borrowers in competitive markets are more likely to be discouraged than in concentrated markets. Thirdly, the figures show that there is no big difference in market condition effects between moderate concentration and high concentration and the effects are not significant in magnitude in concentrated markets. The fact that relationships reduce discouragement amongst good borrowers in competitive markets suggests that banks are able to identify good borrowers, over the course of relationships, even where there is little investment in the relationship. However, the lower likelihood of discouragement, in concentrated banking markets, points to the higher value of relationships to lenders (and hence higher investment) in these markets and supports Petersen and Rajan (1995) s hypothesis of the beneficial effects of market concentration 20

for lenders. The similarity of the findings in moderately concentrated and highly concentrated markets suggests that the value of relationships is similar in both types of market. In other words, a moderately concentrated banking market is sufficient for lenders to invest enough in relationships such that good borrowers are unlikely to feel discouraged from applying for loans. Figure 2-1: here Figure 2-2: here Figure 2-3 shows how the estimated probability of being discouraged of an average risk borrower varies over the length of relationships. This graph indicates that, in a competitive market, relationships reduce the likelihood of discouragement but that relationships increase the likelihood of discouragement in concentrated markets. This suggests that banks in competitive markets invest an insufficient amount in relationships to be able to distinguish average/marginal borrowers from good borrowers (compare with Figures 2-1 and 2-2). However banks in concentrated markets have greater incentives to invest in relationships so that riskier marginal borrowers are increasingly likely to be identified and are discouraged by this. Figure 2-3: here Figures 2-4 and 2-5 present the estimated probabilities of bad borrowers (with risk levels above average). In competitive markets, the likelihood does not change significantly over relationships; while in concentrated markets, the likelihood increases significantly. These results suggest that concentrated markets are more efficient than competitive markets in 21

terms of discouraging bad borrowers from borrowing. Again this points to the higher value of relationships, and superior information production, in concentrated markets as an explanation for these findings. Figure 2-4: here Figure 2-5: here 5. Conclusions and Implications The importance of discouraged borrowers has been recognised in both the theoretical and empirical literature. Kon and Storey (2003) theory predicts that application costs, screening errors due to information problems and market rates will affect the likelihood of discouragement amongst good borrowers. This paper has conducted an empirical analysis, based on United States data, which encompasses both good and bad discouraged borrowers. In this manner we have been able to examine the determinants of discouragement and test whether discouragement is an efficient self-rationing mechanism 6. We find that both the demographics of the entrepreneur and business characteristics have a strong influence on discouragement. These characteristics include business size, the use of financial products, age and the personal wealth of the entrepreneur. Having controlled for these characteristics, we then find that riskier borrowers are more likely to be discouraged. Indeed, amongst high risk borrowers, this likelihood increases as information quality improves with longer financial relationships; while, the likelihood for low risk borrowers decreases. 6 One limitation of this paper is that we cannot examine the effects of interest rate identified in Kon and Storey s model because of the cross-sectional nature of the data. 22

The empirical results derived in this paper have three important implications. Firstly, they imply that discouragement is an efficient self-rationing mechanism because bad borrowers are more likely to be discouraged than good borrowers. Discouragement can therefore reduce banks costs in dealing with loan applications from risky borrowers. It can also help to overcome the problem of overinvestment resulting from the mistaken approval of applications from risky borrowers due to screening errors. Secondly, improvements in information quality increase the efficiency of discouragement as a self-rationing mechanism. With longer relationships, for instance, risky borrowers become more and more pessimistic about the outcome of their loan applications and therefore become increasingly likely to be discouraged from applying in the first place; while, less risky ones become less likely to be discouraged. Thirdly, our results suggest that the extent of market concentration has a strong impact on the discouragement of small business borrowers. Discouragement, as a self-rationing mechanism, is more efficient in concentrated banking markets because, in these markets, risky borrowers are more likely, and less risky ones are less likely, to be discouraged. It would, however, be unwise to assume these implications apply in small business financing markets in all countries. The US data, upon which these conclusions were derived, reflect a relatively sophisticated small business financing marketplace and yet even here we find some good borrowers are discouraged and some bad ones did apply. What we would expect is that in less sophisticated markets the nature and scale of discouragement would change and that discouragement itself could become a litmus test of market imperfections. Testing this view requires more empirical analysis of discouragement in the context of a diverse range of countries. 23

References Allen, F., McAndrews, J. and Strahan, P., 2002. E-Finance: an introduction. Journal of Financial Services Research 22, 5-27. Ang, J.S., 1991. Small Business Uniqueness and the Theory of Financial Management. Journal of Small Business Finance 1, 1-13. Ang, J.S., Lin, J.W. and Tyler, F., 1995. Evidence on the Lack of Separation between Business and Personal Risks among Small Businesses. Journal of Small Business Finance 4, 197-210. Berger, A.N., 2004. Potential Competitive Effects of Basel II on Banks in SME Credit Markets in the United States, Working Paper, Federal Reserve. Berger, A.N., 2003. The economic effects of technological progress: evidence from the banking industry. Journal of Money, Credit and Banking 35, 141-176. Berger, A.N. and Udell, G.F., 1998. The economics of small business finance: The roles of private equity and debt markets in the financial growth cycle. Journal of Banking & Finance 22, 613-673. Berger, A.N. and Udell, G.F., 1995. Relationship Lending and Lines of Credit in Small Firm Finance. Journal of Business 68, 351-381. Bester, H., 1985. Screening vs Rationing in Credit Markets with Imperfect Information. American Economic Review 75, 850-855. Bilter, P.P., Robb, A.A. and Wolken, J.D., 2001. Financial Services Used by Small Businesses: Evidence from the 1998 Survey of Small Business Finances. Federal Reserve Bulletin April, 183-205. 24

Bolton, P. and Scharfstein, D.S., 1996. Optimal Debt Structure and the Number of Creditors. Journal of Political Economy 104, 1-25. Booth, J.R. and Booth, L.C., 2006. Loan Collateral Decisions and Corporate Borrowing Costs. Journal of Money, Credit and Banking 38, 67-90. Bris, A. and Welch, I., 2005. The Optimal Concentration of Creditors. Journal of Finance 60, 2193-2221. Carey, M., Post, M. and Sharpe, S.A., 1998. Does Corporate Lending by Banks and Finance Companies Differ? Evidence on Specialization in Private Debt Contracting. Journal of Finance 53, 845-878. Cavalluzzo, K.S., Cavalluzzo, L.C. and Wolken, J.D., 2002. Competition, Small Business Finance, and Discrimination: Evidence from a New Survey. Journal of Business 75, 641-679. Cowling, M., 1999. The Determination of Bank Small Business Loan Premia in the UK. Applied Economics Letters 6, 252-253. de Meza, D. and Southey, C., 1996. The Borrower's Curse: Optimism, Finance and Entrepreneurship. The Economic Journal 106, 375-386. Degryse, H. and Ongena, S., 2005. Distance, Lending Relationships and Competition. Journal of Finance 60, 231-266. Degryse, H. and Von Cayseele, P., 2000. Relationship Lending within a Bank-Based System: Evidence from European Small Business Data. Journal of Financial Intermediation 9, 90-109. Detragiache, E., Garella, P. and Guiso, L., 2000. Multiple versus Single Banking 25

Relationships: Theory and Evidence. Journal of Finance 55, 1133-1161. Frame, W.S., Srinivasan, A. and Woosley, L., 2001. The effect of credit scoring on smallbusiness lending. Journal of Money, Credit and Banking 33, 813-825. Fraser, S., 2007. Finance for Small and Medium-Sized Enterprises: Comparisons of Ethnic Minority and White Owned Businesses. Department for Business Enterprise and Regulatory Reform Han, L., 2008. Bricks vs. Clicks: SME Online Banking Behaviour and Relationship Banking. International Journal of Entrepreneurial Behaviour and Research (forthcoming). Han, L., Storey, D.J. and Fraser, S., 2008. The Concentration of Creditors: Evidence from Small Businesses. Applied Financial Economics (forthcoming) Holmstrom, B.R., 1982. Moral Hazard in Teams. Bell Journal of Economics 13, 324-340. Jovanovic, B., 1982. Selection and the Evolution of Industry. Econometrica 50, 649-670. Kon, Y. and Storey, D.J., 2003. A Theory of Discouraged Borrowers. Small Business Economics 21, 37-49. Levenson, A.R. and Willard, K.L., 2000. Do Firms Get the Financing They Want? Measuring Credit Rationing Experienced by Small Businesses in the U.S. Small Business Economics 14, 83-94. Márquez, R., 2002. Competition, Adverse Selection, and Information Dispersion in the Banking Industry. Review of Financial Studies 15, 901-926. Petersen, M.A. and Rajan, R.G., 2002. Does Distance Still Matter? The Information 26

Revolution in Small Business Lending. Journal of Finance 57, 2533-2570. Petersen, M.A. and Rajan, R.G., 1994. The Benefits of Lending Relationships: Evidence from Small Business Data. Journal of Finance 49, 3-37. Petersen M.A. and Rajan, R.G., 1995. The Effect of Credit Market Competition on Lending Relationships. Quarterly Journal of Economics 110, 407-773. Repetto, A., Rodríguez, S., and Valdés, R., 2004. Do Bank Lending Relationships Benefit Borrowing Firms? Working Paper. University of Chile. Stiglitz, J. and Weiss, A., 1981. Credit Rationing in Markets with Imperfect Information. American Economic Review 71, 393-410. Storey, D.J., 1994. Understanding the Small Business Sector, London: Routledge. von Thadden, E., 1992. The Commitment of Finance, Duplicated Monitoring, and the Investment Horizon. CEPR, Network in Financial Market Working Paper 27. 27

Table 1: Descriptive Statistics Total number of observations is 3561. Not reported here, but available upon request from the authors, are the descriptive statistics of nine dummies representing the industry based on two-digit SIC code. Variable Definition Min Max Mean Median Std.Dev. MSA Business is in a Metropolitan Statistical Area (0,1) 0.0000 1.0000 0.7812 1.0000 0.4135 HHI3_B1 Headquartered in a competitive banking market (0,1) 1 0.0000 1.000 0.0514 0.000 0.2208 HHI3_B2 Headquartered in a moderately concentrated banking market (0,1) 1 0.0000 1.0000 0.4249 0.0000 0.4944 HHI3_B3 Headquartered in a highly concentrated banking market (0,1) 1 0.0000 1.0000 0.5234 1.0000 0.4995 Characteristics of the business and entrepreneur CAPNEEDS Demands for external capital (0,1) 0.0000 1.0000 0.4106 0.0000 0.4920 DB Discouraged borrower (0,1) 0.0000 1.0000 0.1404 0.0000 0.3475 C_CORPORATION Business is incorporated (0,1) 0.0000 1.0000 0.2491 0.0000 0.4325 FAMILY_OWNED Family owned (0,1) 0.0000 1.0000 0.8531 1.0000 0.3540 LOGFAGE Firm age (years) 2 0.0000 4.6540 2.4494 2.4849 0.7858 LOGTOTEMP Total employment number 2 0.0000 6.1779 1.8999 1.6094 1.5544 LGOGROWTH2 Sales growth: sale (year t) / sales (year t-1) 2 0.0000 5.3458 0.7913 0.7238 0.3582 LOGTLBTA Capital structure: total liability/ total business assets 2 0.0000 16.3805 0.5278 0.3499 0.9288 F4_BCC Business credit card (0,1) 0.0000 1.0000 0.3993 0.0000 0.4898 F7_CRL Credit lines (0,1) 0.0000 1.0000 0.3606 0.0000 0.4802 DEGREE Owner has a college degree or above (0,1) 0.0000 1.0000 0.5218 1.0000 0.4996 MALE Male owner (0,1) 0.0000 1.0000 0.7352 1.0000 0.4413 MINOR Minority owner (0,1) 0.0000 1.0000 0.1435 0.0000 0.3506 LOGOAGE Owner s age (years) 2 2.9957 4.5643 3.9216 3.9318 0.2221 LOGEXP Experience of owner in business (years) 2 0.0000 4.2905 2.8036 2.9444 0.6953 LNPW2 Personal assets of owner (million $) 2, 3 0.0000 4.7536 0.4296 0.2429 0.5130 Quality of borrowers DB_SCORE Dun and Bradstreet score: categorical 4 1.0000 5.0000 2.9700 3.0000 1.0402 28

Variable Definition Min INST_DB12 Instrumented D&B score: moderate risk (0,1) 5 0.0000 INST_DB13 Instrumented D&B score: average risk (0,1) 5 0.0000 Max Mean Median Std.Dev. 1.0000 0.2064 0.0000 0.4048 1.0000 0.6810 1.0000 0.4662 INST_DB14 Instrumented D&B score: high risk (0,1) 5 0.0000 1.0000 0.1098 0.0000 0.3127 INST_DB15 Instrumented D&B score: significant risk (0,1) 5 0.0000 1.0000 0.0028 0.0000 0.0529 INST_DB2 Instrumented D&B score: weighted average 5 1.4538 4.1491 2.6984 2.6503 0.4814 Application costs LOGDIST_PI Distance to the primary institution (miles) 2 0.0000 8.0876 1.3444 1.0986 1.1432 ONLINE Apply loans online (0,1) 0.0000 1.0000 0.0410 0.0000 0.1983 Information issues and nature of primary lender N_FS Number of sources of financial services 0.0000 8.0000 2.3002 2.0000 1.5238 PITYPE_BANK Primary institution is a bank (0,1) 0.0000 1.0000 0.8750 1.0000 0.3307 LOGRELATION_PI Length of relationship with primary institution (months) 2 0.0000 6.6606 4.1339 4.1109 1.0143 Interaction terms DB2LNR INST_DB2 * LOGRELATION_PI 0.0000 6.5806 0.9549 0.0000 1.8995 DB2LNRHB1 INST_DB2 * LOGRELATION_PI * HHI3_B1 0.0000 19.7290 0.5691 0.0000 2.5154 DB2LNRHB2 INST_DB2 * LOGRELATION_PI * HHI3_B2 0.0000 20.3723 4.6802 0.0000 5.8072 DB2LNRHB3 INST_DB2 * LOGRELATION_PI * HHI3_B3 0.0000 22.4352 5.7374 6.2213 5.8583 Note: 1. 2. 3. 4. 5. Concentration of banking market is measured by the Herfindahl Index which is a categorical variable and measures the degree of competition within the local financial market. It is equal to one when the index locates between 0 and 1000, meaning the financial market is less concentrated. It equals two when the index is between 1000 and 1800, suggesting a moderate concentrated financial market. It equals three when the index is larger than 1800, meaning that the financial market is highly concentrated. These variables are transformed into nature log value of one plus the real value. Personal wealth is defined as the total value of the owner s home equity and the net worth of other assets. Dun and Bradstreet score on risk is categorical with a range from 1 to 5, where 1 is low risk, 2 is moderate risk, 3 is average risk, 4 is high risk and 5 is significant risk. We estimate the instrumented D&B scores by conducting ordered logistic models on the business and entrepreneur variables and their credit history. One instrumented D&B score (INST_DB1) follows the categorical nature of the score, ranging from 2 (moderate risk) to 5 (high risk). The score is recoded to a specific category where the sample firm has the highest probability of falling into this category. For comparable purpose, the other instrumented D&B score (INST_DB2) is continuous and a weighted average of the possible categories where the weight is the probability of falling into a specific category estimated from an ordered logistic model. Detailed instrumentation process is available from the authors on request. 29