Debt Financing, Survival, and Growth of Start-Up Firms

Transcription

1 Debt Financing, Survival, and Growth of Start-Up Firms Rebel Cole DePaul University Chicago, IL phone: (312) fax: (312) Tatyana Sokolyk Brock University St. Catharines, ON phone: (905) ext fax: (905) Abstract: We analyze how the start-up capital structure of a firm affects subsequent firm outcomes. We find that, after three years, firms using debt at start-up in particular, business debt but not personal debt are significantly more likely to survive, and to achieve higher levels of revenues than are other firms. We attribute this to the incentives and expertise of the lenders, which, for business debt, almost always are banks. We also examine a start-up s decision to use credit; we find that better-quality start-ups are more likely to use credit, and are more likely to use business rather than personal credit. Keywords: banking/financial intermediation, entrepreneurial finance, credit, debt financing, capital structure, Kauffman, KFS, small business, start-up, survival JEL Classifications: G21, G32, J71, L11, M13

2 According to most corporate finance textbooks, the capital-structure decision how to finance a firm s fixed assets is one of the three most important decisions facing a financial manager; second in importance only to the capital-budgeting choice. 1 Yet we know very little about how the owner/manager of a start-up firm decides which of various types of debt and equity to use to finance the firm s fixed assets and whether that decision affects subsequent firm performance. 2 Because of the high degree of uncertainty about the future performance of a start-up firm and large information asymmetry between the provider of capital and the firm s owner, the question of what funding mix improves the performance prospects of a newly founded firm is of great interest to financiers, researchers, and policymakers. We argue that traditional theories of capital structure, such as the pecking-order theory, are largely irrelevant for start-up firms because start-ups have no retained earnings, have no access to public debt or equity markets, and have only extremely limited access to the private equity market. Rather, the sources of capital available to a start-up firm are primarily limited to owner s equity and bank debt. 3 Existing theories on capital markets (see, e.g., Stiglitz and Weiss, 1981; Berger and Udell, 1998) posit that large information asymmetries lead to credit rationing and, thus, limited access to credit financing. Relative to public firms, and even to mature privately held firms, start-ups are the most opaque; consequently, they should find it most difficult to obtain external financing. Yet we document that almost half of all entrepreneurial firms successfully obtain business loans, primarily from banks, at the firm s start-up with no record of the firm s operations to inform their lenders. 1 See, for example, Brealey, Myers, and Allen (2011) and Ross, Westerfield, and Jaffe (2012), two of the leading corporate finance textbooks used in U.S. MBA programs. 2 Most existing empirical studies of capital structure have focused on large publicly traded corporations, for which data are readily available. Four notable exceptions are Brav (2009), who analyzes data on privately held U.K. companies; Ang, Cole, and Lawson (2010) and Cole (2013), who analyze data on privately held U.S. companies from the Federal Reserve s Surveys of Small Business Finances; and Robb and Robinson (2014), who, as we do, analyze data on U.S. start-up companies from the Kauffman Firm Surveys. 3 Conventional wisdom posits that start-up firms rely heavily on financial capital from friends and family, but Robb and Robinson (2014) document that this is not the case.

3 In addition, we find that 55% of all start-ups report using personal loans to the owners to finance the business start-up, but we argue that these types of loans are fundamentally different from loans directly to the business. When evaluating a personal loan application, a lender is evaluating the creditworthiness of the person making the application, with no commitment to monitor the firm should the lender extend credit. In fact, the lender often does not know that the loan proceeds will be funnelled by the borrower into financing the start-up firm. In this respect, personal loans used to finance a start-up are really equity injections by the owner rather than private debt. Consequently, we hypothesize that start-up firms that obtain business loans benefit from the selection and monitoring by bank lenders, whereas start-up firms that obtain financing from personal loans to their owners do not. We further hypothesize that this selection and monitoring leads to superior outcomes for entrepreneurial firms that obtain business loans at the firm s start-up. We test these hypotheses using data from the Kauffman Firm Surveys (KFS) on U.S. start-up firms, and find that (1) better-quality start-ups are more likely to obtain credit, and (2) start-ups who obtain credit during their first year of operation grow faster and survive longer than their counterparts without credit financing. Notably, business debt, as opposed to personal debt, carries positive benefits for a firm s performance and survival. Furthermore, of all credit users, better-quality start-up firms are more likely to obtain business credit and are less likely to obtain personal credit. This new evidence on the differences between business and personal credit is consistent with our reasoning outlined above that personal loans are fundamentally different from loans made directly to the business. The results are robust to controls for the initial firm and owner characteristics, for the sample selection bias, and for the availability of credit financing to small businesses at the state level. Furthermore, our results hold true when we employ propensity-score-matching analysis, again showing that start-up firms that use business debt in their initial capital structures outperform otherwise similar firms that do not. Finally, the result that start-ups who obtain credit, and in particular business credit, show a higher level of revenue is robust when we control for the potential endogeneity between the probability of receiving debt financing and performance outcomes of start-up firms. In the endogeneity - 2 -

4 analysis, we use an exogeneous variable that estimates the probability of obtaining credit, and utilize data from another survey of small businesses, the Federal Reserve s Survey of Small Business Finances (SSBF). Our study expands the capital-structure literature and, in particular, the literature on the special role of financial intermediaries in reducing information asymmetry about a borrowing firm through screening and monitoring (see Diamond, 1984, 1991; Fama, 1985; Ramakrishan and Thakor, 1984). In contrast to the prediction of credit-rationing theory, we document that banks participate quite actively in the most informationally opaque market the credit market for start-up businesses. Furthermore, we show that banks are successful in screening and monitoring brand-new businesses. This evidence complements the existing literature on pre-ipo (initial public offering) firms and newly public firms (Schenone, 2004; Gonzalez and James, 2007; Benzoni and Schenone, 2010) and extends earlier studies examining the special role of banks in public markets (e.g., Billett, Flannery, and Garfinkel, 1995; Slovin, Sushka, and Poloncheck, 1993; Hoshi, Kashyap, and Scharfstein, 1990; James, 1987; James and Wier, 1990; Lummer and McConnell, 1989). Once again, we emphasize that bank lending is unique in the setting of startup firms that have extremely limited access to other sources of funds and are characterized by large information asymmetry. In addition, we contribute to the literature on debt financing of privately held firms (see, e.g., Ang, 1992; Berger and Udell, 1998; Brav, 2009; Ang, Cole, and Lawson, 2010; Robb and Robinson, 2014; Cole, 2013) by documenting that the start-up s decision to obtain credit in particular, business credit has important consequences on subsequent firm outcomes of survival and growth. Furthermore, by documenting the importance of start-up credit financing for subsequent firm outcomes, we contribute to the strand of literature that seeks to determine what initial factors affect the survival and growth of new entrepreneurial firms (Cassar, 2004; Cooper, Gimeno-Gascon, and Woo, 1994; Schwienbacher, 2013). Finally, we contribute to the growing literature that analyzes start-up firms using data from the Kauffman Firm Surveys (see, e.g., Coleman and Robb, 2009; Fairlie and Robb, 2009; Coleman and Robb, 2012; Robb and Watson, 2012; Cassar, 2014; Robb and Robinson, 2014)

5 Our study has several important implications for economic policymakers and investors interested in entrepreneurial start-ups. Without a doubt, small firms are vital to the U.S. and world economies. According to the U.S. Small Business Administration, small firms are responsible for almost two out of three net new private sector jobs, and almost half of both private-sector employment and private sector output. 4 A better understanding of what types of firms obtain credit and the sources of credit finance at the initial stage of business formation can help policymakers take actions to increase the availability of credit to start-up firms that will potentially lead to the creation of more jobs and faster economic growth. 5 Furthermore, a better understanding of the importance of credit use at the firm s start-up stage should provide policymakers with guidance on how to tailor economic and tax policies to help start-up businesses obtain credit when they need credit, thereby increasing both employment and productivity. This information has especially important implications for tax reform proposals that would limit the deductibility of interest on business debt, which would increase the cost of credit to small firms. Additionally, the results of our study provide some guidance to investors considering new entrepreneurial firms. We show that entrepreneurial firms that are able to secure business credit at the firm s start-up tend to outperform other start-up firms. Finally, our results underscore the critical role that banks play in the success of start-up firms. In the United States, the number of banks has plummeted by almost half over the past three decades, primarily by the consolidation of community banks that specialize in lending to small businesses. Policymakers need to rethink burdensome regulations that are driving many bankers to leave the business. Our study provides important new evidence and contributes to the existing literature; however, it has several limitations that we are not able to address. We leave these for future research. For example, the KFS data do not identify start-up firms that apply for business credit and get rejected. With data on loan-application details and outcomes, researchers could analyze 4 See frequently asked questions at 5 Such policies recently have included the Small Business Lending Fund, which provides capital to community banks and community development loan funds as an incentive to increase their lending to small firms, and the 7(a) Loan Program of the U.S. Small Business Administration, which provides loans to qualifying small firms

6 the selection mechanisms and processes used by lenders in the case of start-up firms. Also, the KFS data do not provide information on the terms and structure of the loans or the characteristics of the lenders. That information would enable researchers to better analyze how banks select and monitor start-up firms, the role of relationship lending in start-up financing, and policy implications regarding the types of lenders that are most successful in selecting and nurturing start-up firms. Nevertheless, our study provides new evidence on the importance of bank lending for firms with no record of business operations and very limited access to other types of financing. The remainder of our study is structured as follows. In section I, we review related studies that analyze (A) credit financing of privately held firms; (B) banks roles in screening and monitoring borrowers; and (C) performance of young entrepreneurial firms. In addition, we develop hypotheses in section I. In section II, we describe our data and methodology. In section III, we present our results, followed by a summary and conclusions in section IV. I. Literature Review and Hypotheses A. Credit Financing of Privately Held Firms A large number of articles, dating back almost 70 years to Wendt (1946), have examined the issue of availability of credit to small businesses. Theory on capital rationing due to information asymmetry suggests that even good-quality firms may have limited access to debt financing (Stiglitz and Weiss, 1981). Extending this theory, Berger and Udell (1998) discuss small business financing through a financial growth cycle paradigm in which different capital structures are optimal at different points in the firm s life cycle. They suggest that firms use internal funds in the firm s early years of operation and external funds as they grow older and larger. Petersen and Rajan (1994) find that close ties with creditors lead to greater availability of credit at lower rates of interest and highlight the importance of firm-lender relationships in the allocation of credit. Berger and Udell (1998) extend the findings of Petersen and Rajan (1994) by analyzing the lines of credit, a type of lending where relationships should be especially important. They document that loan rates are lower for firms with longer firm-lender - 5 -

7 relationships. Cole (1998), in contrast, focuses on the lender s decision whether to extend credit and finds that the existence rather than the length of the firm-lender relationship affects the likelihood of a lender extending credit to the firm. Cole (2013) examines the capital structure decisions of small, privately held firms in the United States and documents that leverage is negatively related to firm size, age, profitability and credit quality; and is positively related to tangibility and limited liability. In a related paper, Cole (2009) finds that firms using no credit are significantly smaller, more profitable, more liquid, and of better credit quality than firms using credit, and concludes that this evidence is generally consistent with the pecking-order theory of firm capital structure. 6 Robb and Robinson (2014) examine the capital structure decisions of start-up firms and document that, in contrast to the conventional wisdom and theory implications, young entrepreneurial firms rely heavily on formal debt financing. These authors document that more than 40% of initial capital comes from external debt financing. Several studies analyze whether the business owner s race and gender influence the availability of credit for small businesses and find significant differences in availability of credit by owner s race (Cavalluzzo and Cavalluzzo, 1998; Blanchflower, Levine, and Zimmerman. 2003; Cavalluzzo and Wolken, 2005). Furthermore, Coleman (2002) documents that small businesses with Black owners are more likely to be denied borrowers and are more likely to be discouraged borrowers. Blanchard, Zhao, and Yinger (2008) find that both Black-owned and Hispanic-owned firms are more likely to be denied credit. Robb, Fairlie, and Robinson (2009) document that Black-owned start-up firms face greater difficulty in obtaining financial capital than do White-owned firms. B. Banks Role in Screening and Monitoring Borrowers Theories of financial intermediation posit that banks, through screening and monitoring, play a special role in reducing information asymmetry about the borrowing firm (see Berlin and 6 Harris and Raviv (1991) review the theories of capital structure. The pecking-order theory of financing demonstrates that capital structure is driven by firms' desires to finance new investments, first internally, then with low-risk debt, and finally with equity (Myers, 1984)

8 Loeys, 1988; Diamond 1984, 1991; Fama, 1985; Ramakrishan and Thakor, 1984). Banks identify successful firms ex ante, prior to providing a loan (screening); and influence borrowing firms to achieve better performance ex post, while a loan is outstanding (monitoring). Several empirical studies, that examine publicly traded firms provide evidence of a bank s special role by documenting positive abnormal returns to the announcements of bank loan agreements (e.g., James, 1987; Lummer and McConnell, 1989; Mikkelson and Partch, 1986). These studies conclude that banking relationships generate valuable private information about the financial prospects of the borrowing firms. Several more recent studies, however, have questioned the special role of bank lending. Billett, Flannery, and Garfinkel (2006) document that firms announcing bank loans experience negative abnormal returns over the subsequent three years. Fields, Fraser, Berry, and Byers (2006) argue that the advantages of bank lending have disappeared since the 1980s due to financial system changes and greater availability of (and less costly) financial information. However, recent studies examining pre-ipo and young publicly traded firms present evidence of positive effects of bank lending on the borrowing firm. Schenone (2004) documents that firms with a pre-ipo banking relationships with a prospective underwriter have a lower level of IPO underpricing. Gonzalez and James (2007) find that the existence and size of pre-ipo banking relations positively affect the firm s post-ipo performance and suggest that firms with the best prospects establish banking relationships prior to IPO. Shaffer and Sokolyk (2013) document that bank lending is associated with positive long-term outcomes of newly public firms. Newly public firms that borrow experience significantly smaller decreases in operating performance and better long-term stock performance than non-borrowers. These studies highlight the importance of banks screening and monitoring in an environment characterized by high information asymmetry. C. Performance of Young Entrepreneurial Firms Most studies of firm performance and survival have focused on public firms or on older, more established privately held firms. Our focus on privately held start-up firms provides an important contribution to this literature. From the perspective of economic theory, informational - 7 -

9 asymmetry is an important framework for the analysis of firms financing decisions and outcomes (see Myers, 1984; Jensen and Meckling, 1976). The analysis of privately held start-up firms provides new insights to the literature because start-up firms have extremely high levels of information asymmetry due to the opacity of their operations, their lack of a performance track record, their small size, and the central role of their founders in designing and managing firm operations. Several studies from the entrepreneurship literature analyze the performance prospects of young entrepreneurial firms and argue that decisions made at the initial stage of business formation can have permanent and significant effects on subsequent firm outcomes. Cassar (2004) examines the determinants of initial capital structure, and Schwienbacher (2013) develops a theory of the effects of early stage financing decisions on subsequent firm growth and development. Cooper, Gimeno-Gascon, and Woo (1994) develop a theoretical model of the initial financial capital and entrepreneur s characteristics and their effect on the performance of newly established firms. These authors argue that factors observable at the firm s start-up provide a highly predictive power in firm performance analysis. D. Hypotheses Development Our primary hypotheses relate to the differences between start-up firms that obtain credit ( use-credit firms ) and start-up firms that do not obtain credit ( no-credit firms ). As outlined above, a start-up firm has a high degree of informational asymmetry due to the lack of a performance track record and a high degree of uncertainty about future performance. We hypothesize that new entrepreneurial firms with better performance prospects are more likely to obtain credit financing at the firm s start-up. Formally, our first hypothesis states: H 1 : Lenders are more likely to extend credit to start-up firms with better performance prospects. Whereas our first hypothesis relates to the screening role performed by financial intermediaries when evaluating performance prospects of potential borrowers, our second hypothesis emphasizes the monitoring function performed by credit providers. We hypothesize that firms that obtain credit financing at the firm s initial stage exhibit better future performance - 8 -

10 than do firms that do not obtain credit financing at the firm s start-up. This hypothesis is also consistent with the notion that availability of credit at the initial stage of business formation helps firms grow and develop. Formally, our second hypothesis states: H 2 : Monitoring by lenders enables firms that obtain credit financing at start-up to achieve better subsequent performance. Our second set of hypotheses relate to the differences between firms that, at start-up, obtain business credit versus personal credit. Robb and Robinson (2014) document that debt financing provides an important source of capital for start-up firms. They further argue that it is important to distinguish between insider debt (obtained from the owner, family, and friends) and outsider debt (obtained from sources such as banks, nonbanks, government, and so forth). Robb and Robinson (2014) document that the portion of total capital financed through outsider debt is positively related to firm performance three years after the firm s start-up. They conclude that outsider debt is most important for successful firm outcomes because of the superior enforcement technologies available to outsiders. In so doing, they conflate the effects of personal and business loans, especially those loans from banks. 7 We argue that a much superior taxonomy is based upon the entity (i.e., the firm versus the firm s owner) being evaluated, and subsequently monitored, by the prospective lender. When the entity is the firm s owner, the prospective lender is only tangentially concerned with the firm s creditworthiness, finances, and prospects for success; instead, the lender is focused on the creditworthiness of the firm s owner. In fact, the prospective lender often does not know that the borrowed funds will be used to finance a start-up, as in the case of the firm s owner using proceeds from a home-equity line of credit or personal credit card to finance the start-ups assets. H 3 : The ability of a start-up firm to obtain financing by personal credit is unrelated to the firm's performance prospects. 7 Whereas Robb and Robinson (2014) presume that personal guarantees and personal collateral must often be posted to secure financing for startups, Avery, Bostic, and Samolyk (1998) report that 60% of small business lending is not personally guaranteed. In addition, Avery, Bostic, and Samolyk (1998) exclude from their analysis credit card loans, which account for a large portion of KFS business loans and are typically an unsecured form of credit

11 In contrast, when the entity being evaluated is the firm itself, the prospective lender is highly focused on the firm s creditworthiness, finances, and prospects for success. H 4 : Lenders are more likely to extend business credit to start-up firms with better performance prospects. Moreover, when the firm is successful in obtaining financing, it then benefits from the lender s monitoring throughout the life of the loan. Consequently, our taxonomy categorizes debt based upon the entity being evaluated by the lender the firm or the firm s owner in marked contrast to the categorization by Robb and Robinson (2014). We hypothesize that firms obtaining loans in the name of the firm, rather than in the name of the firm s owner, are more likely to succeed in terms of both survival and revenues because of the banks monitoring. Formally, we formulate our fifth and six hypotheses as follows: H 5 : Monitoring by lenders enables firms that obtain business credit financing at start-up to achieve better subsequent performance. H 6 : Start-up firms that obtain personal credit do not benefit from bank monitoring, so its subsequent performance is unrelated to obtaining personal credit. In the next section, we describe the data and methodology used to test these hypotheses. II. A. Data and Sample Description Data and Methodology We use data from the Kauffman Firm Survey to examine what types of firms use credit at the firm s start-up and the effect of the initial credit use on subsequent firm performance and survival. This annual survey follows 4,928 privately held firms that were established in Currently, the survey results are available for the baseline year (2004) and seven follow-up years ( ), with the final 2012 results to be released during The KFS represents the 8 The KFS identified start-up firms by using a random sample from Dun & Bradstreet s database list of new businesses established during 2004, excluding wholly owned subsidiaries of existing businesses, businesses inherited from someone else, not-for-profit organizations, and firms that had any kind of business activity prior to For more detailed information about the KFS data, see Ballou et al. (2008) and DesRoche et al. (2012)

12 largest and the most comprehensive data on U.S. start-up firms. Along with detailed information on the firm s use of credit, the KFS provides data on various firm and owner characteristics. The richness of the KFS data allows us to identify what types of firms use credit as well as the different types of credit at the firm s start-up and to explore the relation between the use of different types of credit at the firm s start-up and subsequent firm performance. Whereas Robb and Robinson (2014) analyze the amounts of financial capital used by start-up firms, we focus on the incidence of credit use by such firms. The analysis of the incidence of credit use is important because of the mass of points at zero for most of the amounts of capital and highly skewed distributions for those firms with non-zero amounts of capital, especially for credit sub-types. This means that the median amount can be zero while the mean amount is large and positive for many debt categories. Table I provides a summary and definitions of KFS variables used in the study; Table II presents the descriptive statistics for these variables. Table II shows that 76% of firms use some type of credit (either business, personal, or trade) at the firm s start-up. This is quite similar to 80% of closely held U.S. firms of any age documented by Cole (2010) and suggests that entrepreneurial firms try to secure credit financing at the earliest stage of their formation. Furthermore, 55% of firms use personal credit at the firm s start-up, suggesting that more than half of start-ups owners explicitly take on personal liability beyond their equity investment in the firm. This evidence is consistent with the notion that many entrepreneurs hold levered equity stakes in their ventures (see Robb and Robinson, 2014). However, a sizable percentage (44%) of start-up firms secure financing on the name of the business itself. While it is possible that business credit is also backed by personal guarantees and personal collateral, the fact that a majority of start-ups explicitly use personal credit and a large minority use business credit suggests that there are important differences in these firms and in their capital-structure decisions. 9 9 The KFS does not provide data on personal guarantees or personal wealth of the entrepreneurs in 2004 so we cannot directly test this proposition

13 Table II also reports summary statistics on the firm s and primary owner s characteristics. On average, start-up firms generated about $230,000 in revenues, a surprising $486 in net income, and had $347,000 in total assets by the end of the first year of operations. The average level of cash and tangible assets was $38,000 and $187,000, respectively. The financial characteristics of a median firm, however, are quite different from the characteristics of an average firm. The median firm generated only $7,500 in revenues, reported a net loss of $300, had $20,000 in total assets, $2,000 in cash, and $5,000 in tangible assets by the end of the startup year. The 25 th percentile start-up firm had values of zero for revenue, cash, current assets, and tangible assets, and reported a net loss of $10,000. The 75 th percentile start-up firm had values for revenues and assets that are substantially smaller than the sample averages, indicative of the highly skewed distributions for these financial variables. In addition, the standard errors for firm variables are very large. These observations demonstrate that the distribution of start-up firms is highly skewed with the potential presence of significant outliers. To address the skewness of the distributions, we take the natural logarithms of firm financial characteristics variables in our subsequent analysis. Table II also shows that about two in three start-up firms are organized as corporations and one in three has more than one owner. At the firm s start-up, the primary owner, on average, is 45 years old, has almost 13 years of prior experience in the same industry, and has one prior start-up experience. 10 One in four start-up firms has a female as its primary owner. Four out of five firms have a White primary owner; but less than one in ten have a Black primary owner; one in twenty has a primary owner who is Asian; and one in twenty has a primary owner who is 10 Similar to prior studies using the KFS data, we define a firm s primary owner as the firm s owner who has the highest percentage ownership (see, e.g., Ballou et al. 2008; Robb, Fairlie, and Robinson, 2009). In cases where two or more owners have the same percentage ownership, the owner who works the most hours in the firm is defined as the primary owner. In cases where two or more owners report the same ownership and the same number of work hours, a series of other variables (i.e., owner s education, age, work experience, amount of initial equity invested, and other start-up experience) is used to create a ranking of owners in order to define the primary owner

14 Hispanic. With respect to educational attainment, two out of three primary owners have some college education or hold a college (bachelor s) degree, and one in five holds a graduate degree. B. Methodologies In order to test our first hypothesis H 1, we use a multivariate probit model to analyze what types of firms are more likely to use credit at the firm s start-up. To test our third and fourth hypotheses, H 3 and H 4, we examine what types of firms are more likely to use personal credit and what types of firms are more likely to use business credit, conditional on the use of any type of credit; for this analysis, we employ the multivariate bivariate probit model. We describe the methodologies used to test these hypotheses in section II.B.1. To test our second hypothesis H 2, we explore whether the use of credit at the firm s startup is associated with firm outcomes specifically, firm survival and firm growth in terms of revenues. For survival analysis, we use the Cox (1972) proportional hazard model. For growth analysis, we employ Weighted-Least Squares (WLS) regression to examine the effect of credit use on the firm s revenues three years after the start-up. Since this analysis includes only firms that have survived three years from the firm s start-up, we use the Heckman (1979) two-step model to control for sample selection bias. To test hypotheses H 5 and H 6, we use the same two models to examine the relation between the use of business credit and personal credit and the firm s revenue and survival three years after the firm s start-up. 11 The methodologies for testing these hypotheses are described in section II.B.2. In all our analyses, we incorporate the survey sampling weights because the KFS sample is not a random sample, but instead is a stratified random sample where high technology firms are over-represented relative to firms in other industries. B.1. Types of Firms that Use Credit at Start-Up To examine whether better-quality firms are more likely to obtain credit at the firm s start-up (H 1 ) and are more likely to obtain business credit (H 4 ), we estimate a multivariate 11 We analyze revenues for and survival up to 2007, 2008, 2009, and 2010, and find that our results are qualitatively similar across all these years. The results for 2007 are most consistent across various specifications, and we focus on these results in this paper

15 bivariate probit selection model. The first/selection equation estimates the probability of using any credit (test of H 1 ), with the dependent variable equal to one if the firm used business, personal, or trade credit at the firm s start-up. The second/outcome equation estimates the probability of using business credit, conditional upon using any type of credit (test of H 4 ). To test H 3, we use the same bivariate probit selection model with the second/outcome equation estimating the probability of using personal credit. The bivariate probit selection model accounts for a non-random selection mechanism operating on those firms that decide to use credit and choose whether to use business or personal credit. The two equations in the bivariate probit model are as follows: y * = ' X + є, y = sign (y* ) (1) 1 and y * = ' X + є, y = sign (y* ) (2) 2 where: є, є ~ Bivariate Normal (0, 0, 1, 1, ρ) 1 2 In the bivariate probit selection model, (y, x ) are observed only when y is equal to one, so the error terms in eqs. (1) and (2) must be re-specified as є j = exp(γ j, z j) u j, where (u 1, u 2 ) have the bivariate standard normal distribution. The estimated correlation coefficient ρ (the correlation between error terms є 1 and є 2 ) can be used to test for selection bias. If ρ is statistically significant, then we can reject the null hypothesis that selection bias is not present. X 1 and X 2 are the vectors of independent variables (defined in Table I), which include Firm Characteristics, Owner Characteristics, Other Sources of Capital, and Industry Classifications. The probability of using credit is likely to be influenced by the availability of credit, so we include a proxy for the supply of credit, measured as the amount of 2004 state-level small-business bank lending per small business enterprise. 12 To make sure that the amount of state-level small-business bank 12 Data on bank lending to small businesses are from the FFIEC s Consolidated Reports of Condition and Income, which is a quarterly regulatory report filed by all U.S. commercial banks

16 lending is exogenous to the use of credit by start-up firms, we instrument it with the amount of state-level homestead bankruptcy exemption and use the predicted value from that regression as a proxy for the supply of credit to small businesses. 13 Finally, the coefficients and in eqs. (1) 1 2 and (2) are estimated by the Maximum Likelihood Estimation Method. 14 B.2. Does Initial Credit Explain Firm Outcomes? After analyzing what types of start-ups are more likely to use credit, we explore whether the use of credit at the firm s start-up is associated with firm outcomes (test of H 2, H 5, and H 6 ). We first examine the relation between credit use and the probability of survival during the first three critical years of firm operations. For surviving firms, we explore the relation between the use of credit at the firm s start-up and the level of revenues three years after the firm s establishment. With firm s survival and revenue measures, we are capturing both the stability of the firm and its growth during the first critical years of the firm s operations. B.2.1. Survival Analysis We use the Cox (1972) proportional hazard model to estimate the effect of initial credit conditions on the probability of survival of young entrepreneurial firms. The advantage of using a survival model, such as the Cox proportional hazard model, is that the survival functions follow the firm through time and observe at which point in time it experiences an event of (see Part II of Schedule RC-C Loans to Small Businesses and Small Farms ). Business loans with original amounts less than $1 million are reported on this schedule. We then scale the total amount of small business lending by the number of small business enterprises in each state, using the number of enterprises in each state as reported in the 2004 County Business Patterns, which summarizes results from a U.S. Census survey of businesses. Information on the amount of statelevel homestead bankruptcy exemption as of 2003 is taken from Table 1 in Berger et al. (2011). 13 These results, which show that the exemption amount is a valid instrument, are available from the authors upon request. Berkowitz and White (2004) show that small firms are more likely to be turned down for loans if they are located in states that have higher bankruptcy exemptions. We argue that there is no reason for the bankruptcy exemption to be correlated with the use of bank credit by entrepreneurial firms except through its impact on the supply of credit by lenders. 14 We operationalize this methodology by using the heckprobit procedure of the Stata 11 software

17 interest (see Shumway, 2001). Also, survival models incorporate data truncation; that is, if some events are unobserved because they occur beyond the end of the sample period, they are taken into consideration through right censoring. The Cox proportional hazard model is a semiparametric model that employs a maximum partial likelihood estimation method and has the following form (Cox, 1972): h ( t) h ( t)exp( X ), (3) i 0 i where h i (t) is the time-t hazard of firm i (t = in our analysis), which is the probability that firm i will be out of business at the end of year t, conditioning on firm i surviving up to time t; h 0 (t) is the baseline hazard function that is left unspecified and corresponds to the probability of an event when all explanatory variables are zero; X i is a vector of explanatory and control variables, corresponding to firm i; ß is a vector of coefficients to be estimated. 15 Variables are defined in Table I. The explanatory variables are: Credit, which equals one if the firm used business, personal, or trade credit at the firm s start-up, and Business (Personal) Credit, which equals one if the firm used business (personal) credit at the firm s start-up. In some specifications, we also control for Firm Characteristics, Owner Characteristics, Other Sources of Capital, Industry Classifications, and for state-level credit supply to small businesses. B.2.2. Revenue Analysis We then examine whether young entrepreneurial firms that use credit at the firm s startup perform better in terms of revenue growth. We estimate a multivariate Weighted-Least- Squares regression: Revenue i = X i α + µ i (4) where the dependent variable Revenue i is measured as the natural logarithm of one plus the level of revenue of firm i three years after the firm s start-up (2007 KFS); 16 X i is a vector of 15 We operationalize the Cox proportional hazard model using the stcox procedure in the Stata 11 software package. 16 For firms that do not report the amount of 2007 revenues, we use the midpoint value of the reported 2007 revenue range if it is available. Our results do not qualitatively change if we

18 explanatory and control variables, corresponding to firm i; and µ i is a normally distributed error term. Explanatory variables include Credit, which indicates whether the firm used any type of credit at the firm s start-up; in the analysis that differentiates between the use of personal and business credit, Credit is replaced with the indicators Business Credit and Personal Credit. Independent variables also include controls for Firm Characteristics, Owner Characteristics, Other Sources of Capital, Industry Classifications, and a proxy for credit supply. The variables are defined in Table I. Since this analysis includes only firms that have survived three years from the firm s start-up, we use the Heckman (1979) two-step model to control for sample selection bias. This methodology involves first estimating a single-equation multivariate probit model to explain survival through 2007 (rather than the Cox proportional hazard model explained above in section II.B.2.1). y * = X i ' + є, y = sign (y * ) (5) i i i i where y i is equal to one if the firm survives through 2007 and is equal to zero otherwise, є i ~ Normal (0, 1); and corr (µ i; є ) = ρ. In the probit model, we use the same set of regressors X i i as in the Cox proportional hazard model discussed above. We then use the results of this probit estimation to calculate the Inverse Mills Ratio (IMR), and then include IMR as an additional regressor in the WLS model of revenues shown in eq. (4). 17 III. A. Types of Firms that Use Credit at Start-Up Data Analysis and Empirical Results This section presents the analysis on the use of credit by start-up firms. We examine what types of firms are more likely to use credit at the firm s start-up and, conditional upon credit use, what types of firms use business credit and what types of firms use personal credit. Table III exclude these firms from our analysis. We add one to the level of revenue before taking the natural logarithm to avoid creating a missing value for a firm with zero level of revenues. 17 We operationalize the two-step Heckman sample-selection model using the heckman procedure in the Stata 11 software package

19 presents the results of this analysis, utilizing a weighted bivariate probit selection model. In the first selection stage (results reported in column 1 of Table III), the dependent variable takes a value of one if the firm uses any type of credit. The second stage estimates the probability of using business/personal credit, conditional upon using any type of credit estimated in the first stage; the results of this analysis appear in columns 2 and 3 of Table III, respectively. In each of the columns, for ease of economic interpretation, we report odds ratios, which are calculated as e, where is a vector of estimated coefficients. t-statistics are reported in parentheses. As shown in column 1 of Table III, firms that use credit at the firm s start-up are significantly larger, as measured by both revenues and assets, than firms that use no credit. The economic significance of these variables is also large: 4% higher odds of using credit for a 1% increase in revenues, and an almost 6% higher odds of using credit for a 1% increase in assets. 18 In general, corporations are 16% more likely to use credit than other forms of business organizations combined. Furthermore, firms with greater credit risk are less likely to borrow at the firm s start-up. A one-step decrease in the firm s credit category is associated with 10% lower odds of obtaining credit. Firms with a competitive advantage are 16% more likely to use credit. Firms that provide product only are 35% more likely to use credit, and firms that provide product and services are 18% less likely to use credit. Overall, the results provide support for the hypothesis (H 1 ) that better-quality firms are more likely to obtain credit financing at the firm s start-up. These results are different from prior findings on more mature privately held firms (see, e.g., Cole, 2009, 2013) and provide new evidence on the use of credit by start-ups. Furthermore, as shown in columns 2 and 3 of Table III, there are notable differences between firms that use business credit and firms that use personal credit. Firms that are larger in terms of revenue and total assets are more likely to use business credit. In contrast, larger firms are less likely to use personal credit. Corporations are 35% more likely to use business credit, but being incorporated does not affect the probability of personal credit use. Firms that have multiple owners are almost 15% less likely to use personal credit. Firms with lower credit scores are 18 We add one to cash and revenue in order to avoid creation of missing values when we perform logarithmic transformations on those variables

20 significantly less likely to use business credit but are significantly more likely to use personal credit. These findings suggest that, out of all credit users, better-quality firms are more likely to obtain business credit, whereas lesser-quality firms are more likely to obtain personal credit. These results support H 4 that lenders provide business credit to start-ups with better performance prospects. However, we did not expect firm characteristics to affect the probability of obtaining personal credit financing. One possible explanation is that firms with worse performance prospects are either denied or are discouraged from applying for business credit and they apply for personal credit instead. Whereas we do not have data on the loans application process to test this prediction, it is clear from Table III that there are important differences between firms that obtain business credit and firms that obtain personal credit at the firm s start-up. Prior studies examine the effect of owner characteristics on the probability of obtaining credit for privately held firms. Table III demonstrates that several owner characteristics significantly relate to the probability of using credit at the firm s start-up. Firms with owners who are college educated are 18% more likely to borrow, and firms with owners who have a graduate degree are 52% more likely to borrow. However, when differentiating between the use of personal credit and the use of business credit, the results show that owners with graduate degrees are 20% more likely to use business credit, but the owner s education does not affect the probability of personal credit use. Furthermore, consistent with prior studies documenting lower availability of credit to minorities, our results show that firms with Black owners are 45% less likely to use credit at start-up; this result is also negative and significant for business credit use, but not significant for personal credit use. Firms that are owned by females are 17% more likely to use personal credit, but are no less likely to use business credit. Finally, firms with equity contributions by insiders (i.e., friends and family of the owners) are 47% more likely to use credit at the firm s start-up. However, insider equity is not significant in business and personal credit regressions; instead, the insider debt variable is significant, but in opposite ways. Firms with insiders lending money to the firm are less likely to use business credit and are more likely to use personal credit

21 Overall, the results presented in Table III support H 1 and H 4 and suggest that betterquality firms are more likely to obtain credit, specifically business credit, at the firm s start-up. These findings are different from the predictions of the pecking-order theory of financing developed for more mature firms and provide new evidence and contribution to the existing literature on young entrepreneurial firms. We find that better-quality entrepreneurial firms secure credit financing at the earliest stage of business formation. Furthermore, while there is a common notion that entrepreneurs are often personally liable for business debts of the firm and that the distinction between personal and business debt is negligible, the results presented in Table III suggest that it is important to differentiate between the use of business versus personal credit by start-up firms. Our results suggest that, of the credit users, better-quality firms are more likely to use business credit and are less likely to use personal credit. B. Credit Use and Firm Performance In this section, we examine whether the firm s initial capital-structure decisions regarding the use of credit affect the firm s outcome during the first critical years of operation. We test the hypotheses that firms obtaining credit (H 2 ), particularly business credit (H 5 ), at the firm s startup benefit from lenders monitoring. We also test H 6 that start-up firms obtaining personal credit do not receive such benefits. In section III.B.1, we examine the relation between credit use at the firm s start-up and firm survival three years after the start-up; in section III.B.2, we examine the relation between credit use and the level of revenue three years after the start-up. In sections III.B.3 and III.B.4, we perform robustness tests by examining the relation between credit use and revenue and survival in propensity-score-matched sample and by controlling for the endogeneity between the probability of credit use and performance outcomes

22 B.1. Survival Table IV presents the results from the survival analysis and alludes to the idea that firms that obtain credit at the firm s start-up are more stable, as measured by the hazard rate of going out of business. The table presents hazard ratios, which equal 100*(e ß - 1) and summarize the economic significance of a given variable. In column 1 of Table IV, the coefficient for Credit is a statistically significant 0.848, which indicates that a firm that uses credit at start-up has about a 15% lower hazard rate of going out of business; or, equivalently, a 15% greater survival rate. Furthermore, as shown in column 2, which differentiates between the use of business and personal credit, this relation is driven by the use of business credit. The coefficient for Business Credit also is a statistically significant 0.848, whereas the coefficient for Personal Credit is not statistically significant in explaining the firm s survival rate. In the next columns, we sequentially add the vectors of Firm Characteristics, Owner Characteristics, Industry Classifications, Other Sources of Capital, and State-Level SME Lending, which proxies for the credit supply to small entrepreneurial firms. Thus, our analysis controls for a large range of firm, owner, and industry characteristics. For brevity, the results for control variables are not shown in Table IV, but are reported in Appendix Table A.1. Columns 3, 4, and 5 show that, after controlling for differences in various firm, owner, industry, and other financial capital characteristics, the coefficient of Business Credit remains statistically significant at better than the 0.01 level, and the magnitude of the coefficient falls to , indicating that a firm obtaining business credit at start-up has a 20% to 23% higher survival rate than does a firm unable to obtain such financing. B.2. Revenues Table V presents the results of the regressions examining the relation between credit use at the firm s start-up and the level of revenue three years after the firm s establishment (KFS 2007). In each specification, we control for sample selection (i.e., we observe only the firms that have survived to 2007) by including the Inverse Mills Ratio estimated by using the probit model

23 shown in eq. (5). 19 Column 1 shows that the use of credit at the firm s start-up is associated with a higher level of revenue three years after the firm s start-up. When Credit is decomposed into Business Credit and Personal Credit (column 2), business credit drives this positive relation between credit use at the firm s start-up and the level of revenue three years after the firm s establishment. In columns 3 5 of Table V, we sequentially add groups of control variables: vectors of Firm Characteristics, Owner Characteristics, Industry Classifications, Other Sources of Capital, and State-Level SME Lending. For brevity, the results for these control variables are not reported in Table V, but are presented in Appendix Table A.2. The coefficient on Business Credit remains positive and highly significant; while the coefficient on Personal Credit is negative and significant. For example, the results in column 5, which control for industry, other sources of capital, and variety of firm and owner characteristics, show that firms that use business credit at start-up achieve the level of revenue three years after the start-up, which is 104% higher than the level of revenue of firms that do not use business credit at the firm s start-up. In contrast, firms that use personal credit at start-up have a 46% lower level of revenue three years after the firm s formation. 20 Overall, the results presented in Tables IV and V indicate that firms that obtain credit at the firm s start-up, in particular business credit, show better subsequent performance than do firms that do not obtain credit financing at start-up. The traditional explanation for these findings is that the ability of a firm to secure credit financing at the firm s start-up makes the firm less capital constrained, and hence allows it to grow faster than firms that cannot obtain such financing (see, e.g., King and Levine, 1993a, 1993b; Rajan and Zingales, 1998). However, the 19 Initially, we estimated the simultaneous-equations version of the Heckman correction for sample selection, but were unable to obtain convergence. 20 Halvorsen and Palmquist (1980) explain the correct interpretation of coefficients of dummy variables in logarithmic equations. The percentage effect of the dummy variable equals 100*(e ß - 1)

24 fact that business credit positively relates to firm performance while personal credit does not suggests that the availability of financial capital per se does not improve the performance of young entrepreneurial firms. The evidence is consistent with our reasoning that personal loans are fundamentally different from loans directly to the business. With business loans, creditors evaluate, select, and monitor the business venture; with personal loans, creditors are looking at the individual borrower, but not necessarily at the business venture. This analysis, however, does not indicate whether the superior performance of business credit-users is due to the lender s ability to choose more successful firms, or is due to the lender s monitoring, which helps firms to be more successful, or both. The next sections present the results of robustness tests for our findings using propensity score matching and endogeneity analysis. The purpose of these tests is to determine whether lenders help young entrepreneurial ventures to perform better through nurturing and monitoring once the loan is granted or whether firms obtaining business credit at start-up are just betterquality firms, and the use of business credit is simply an indicator of a higher-quality firm. With propensity-score-matching analysis, presented in section III.B.3, we examine the differences in subsequent performance between firms that are otherwise similar but differ in terms of credit use at the firm s start-up. The endogeneity analysis presented in section III.B.4 examines whether the positive relation between credit use and performance remains significant once we control for the potential endogeneity between the probability of obtaining credit at the firm s start-up and the firm s quality. B.3. Propensity-Score-Matching Analysis In this section, we utilize propensity-score-matching analysis to determine whether firms that are different in terms of credit use but are similar across other firm characteristics perform differently three years after the firm s start-up. By performing one-to-one matching, we identify firms that are similar, at start-up, in terms of credit risk, revenue, and total assets, except that one

25 group uses credit at the firm s start-up and the other does not. 21 We compare the survival rates of credit users to the survival rates of comparable firms that do not use credit; and we compare the level of revenue of credit users to the level of revenue of credit non-users. To adjust for possible sample selection bias of being included in the propensity-score-matched sample, we include the Inverse Mills Ratio estimated from a probit model in which the dependent variable is equal to one if the firm is in the matched sample and is equal to zero otherwise; other explanatory variables are the same as included in the survival model shown in eq. (5). Results of propensity-score-matching analysis are presented in Table VI. Panel A reports the results for survival analysis; Panel B reports the results for revenue analysis. For each type of analysis, we run a regression with Credit as a variable and a set of controls for firm, owner, industry, other sources of capital, and state-level SME lending and a regression with Business Credit and Personal Credit variables. For brevity, the results for the control variables are not reported in Table VI, but are presented in Appendix Table A.3. The results show that the positive effect of the use of business credit on the firm s survival and the level of revenue remains strong and significant in both economic and statistical terms in the matched sample of firms. This suggests that, for otherwise similar firms, the ones that obtain business credit at the firm s startup, exhibit better subsequent performance than those that do not obtain business credit. Results imply that the availability of credit financing helps young entrepreneurial firms to grow and become more stable. Furthermore, the results provide support for the hypothesis that lenders, who are mainly banks in our sample, provide an important monitoring role of borrowed firms. Start-ups obtaining business loans benefit from bank monitoring, whereas start-ups obtaining personal loans to finance the firm s operations do not show such benefits. B.4. Endogeneity Analysis In this section, we address the possibility of the endogenous relation between the probability of receiving debt financing and firm performance outcomes using instrumental variable estimation. Despite the fact that we control for a wide range of firm and owner 21 We operationalize the propensity-score matching using the psmatch2 procedure in the Stata 11 software package, choosing the nearest neighbor without replacement

26 characteristics in our regressions, some unobservable firm and owner characteristics may jointly affect the probability of obtaining debt financing at the firm s start-up and subsequent firm performance outcomes. In other words, it may appear that firms using credit, in particular business credit, at the firm s start-up perform better than firms not using credit, whereas, in fact, the types of firms that obtain credit, and in particular business credit, at the firm s start-up are also the types of firms that are more likely to survive and achieve higher levels of revenue. We employ the instrumental variable (IV) method to test whether the positive relation between credit use and performance outcomes remains once we control for this potential endogeneity issue. To perform this analysis, we need a variable that is correlated with the probability of obtaining credit at the firm s start-up but is not correlated with the performance outcomes of KFS firms. To create such a variable, we utilize a different survey of privately held firms the 2003 Survey of Small Business Finance (SSBF). 22 Using data exogeneous to KFS firms, we estimate the probability of receiving credit financing at the firm s start-up. Specifically, utilizing common variables between 2003 SSBF data and 2004 KFS data, we estimate the probability of using credit by analyzing the sample of SSBF firms that reported they are less than ten years old. The dependent variable in this regression equals one if the 2003 SSBF firm reports that the firm had outstanding loans or used trade credit during the survey year, and it equals zero otherwise. We then use the coefficients (reported in Appendix Table A.4, Panel A, model 1) from the SSBF regression to estimate the probability of obtaining credit at the firm s start-up for the sample of KFS firms. The predicted value from the regression of the probability of obtaining credit at the firm s start-up Use Credit # serves as the instrumental variable for the variable Credit used in the main regressions. In a similar manner, we create IVs for Business Credit and Personal Credit. To create an IV for the use of business credit at the firm s start-up, Use Bank Credit #, we estimate the 22 See Cole (2009, 2010, 2013) for detailed information about the 2003 SSBF, which is a nationally representative survey of U.S. small businesses operating in 2003 conducted by the Federal Reserve Board. The KFS was modelled after the SSBF, but with a focus on entrepreneurial firms that began operation during Cole (2010) provides details on the SSBF variables that we use to create our instrumental variables

27 probability that a 2003 SSBF firm reported that it had an outstanding loan from a commercial bank using the sample of SSBF firms that reported they are less than ten years old. The dependent variable in this estimation equals one if the 2003 SSBF reports that the firm had an outstanding loan from a commercial bank during the survey year, and it equals zero otherwise. We reason that, since most of business lending is done by banks, this should be a valid instrument. The coefficients from this regression are reported in Appendix Table A.4, Panel A, model 2. To create an IV for the use of personal credit at the firm s start-up, No Need for Credit #, we estimate the probability that the 2003 SSBF firm reported that it did not need credit during the previous three years using the sample of 2003 SSBF firms that reported they are less than ten years old. The coefficients from this regression are reported in Appendix Table A.4, Panel A, model 3. The dependent variable in this estimation equals one if the 2003 SSBF reports that the firm did not need credit during the previous three years, and it equals to zero otherwise. We reason that the firms that do not need credit will not be using personal credit to finance the firm s operation and expect a negative relation between the probability of using personal credit at startup and the predicted value from the regression of probability that the firm did not need credit. Panel B in Appendix Table A.4 shows that Use Credit #, Use Bank Credit #, and No Need for Credit # are valid instruments for Credit, Business Credit, and Personal Credit, respectively. The coefficient on each of the three IVs is statistically significant at better than the 0.05 level, and has the predicted sign. Table VII presents the results of survival (Panel A) and revenue (Panel B) analysis with instrumental variables for Credit, Business Credit, and Personal Credit. The regressions include controls for various Firm, Owner, Industry Characteristics, Other Sources of Capital, and State- Level SME Lending. For brevity, the results for control variables are not reported in Table VII, but are presented in Appendix Table A.5. Results show that, after controlling for potential endogeneity between the probability of obtaining credit at the firm s start-up and firm performance outcomes, the relation between credit use, in particular business credit, remains positive and significant in the revenue analysis, and is positive but not statistically significant in

28 the survival analysis. Overall, the results (with the exception of statistically insignificant effects in the survival regressions) are consistent with the main results presented in Tables IV and V. IV. Summary and Conclusions In this study, we use data from the Kauffman Firm Survey to provide new evidence on the use of debt financing by U.S. start-up firms. Start-up firms can be characterized by an extremely high level of informational asymmetry and great uncertainty about future performance. Theories of information asymmetries in the capital markets (e.g., Stiglitz and Weiss, 1981) imply that start-up firms will have limited access to credit financing. Yet, we find that 75% of firms use some type of credit at start-up; almost half of all entrepreneurial firms obtain business loans and 55% use personal loans to the owners to finance the firm s operations. We argue that business loans to the firms are fundamentally different from personal loans to the firm s owners. When evaluating a personal loan application, a lender is evaluating the creditworthiness of the person making the application and makes no commitment to monitor the firm if the loan is approved. In contrast, when the lender extends business credit to the start-up firm, it evaluates the creditworthiness and performance prospects of the firm and provides monitoring while the loan is outstanding. We document that this distinction, which has not been addressed in prior literature, is important. Better-quality start-up firms are more likely to obtain business loans and are less likely to obtain personal loans. Notably, the initial capital structure decision is important because it influences future outcomes in terms of survival and growth. Firms that use debt in their initial capital structure in particular, firms that use business debt are significantly more likely to survive their first three years of operations. If they survive, they also have significantly higher revenues three years after the firm s formation. This study documents the importance of credit use, and in particular, bank lending, at the initial stage of business formation. We document that banks perform valuable screening and monitoring roles in the environment, which is likely to have the highest degree of information asymmetry. Furthermore, our finding on the positive impact of the use of business credit at the firm s start-up provides important guidelines to investors interested in contributing financial

29 capital to start-up firms and to policymakers concerned with entrepreneurship growth and development. References Ang, J., 1992, On the theory of finance for privately held firms, Journal of Small Business Finance 1, Ang, J., Cole, R., and Lawson, J., 2010, The role of owner in capital structure decisions: An analysis of single-owner corporations, Journal of Entrepreneurial Finance 14, Avery, R., Bostic, R., and Samolyk, K., 1998, The role of personal wealth in small business finance, Journal of Banking and Finance 22, Ballou, J., Barton, T., Des Rouches, D., Potter, F., Reedy, E.J., Robb, A., Shane, S., and Zhao, Z., 2008, Kauffman Firm Survey: Results from the baseline and first follow-up surveys. Available at: Benzoni, L., and Schenone, C., 2010, Conflict of interest and certification: Long-term performance and valuation of U.S. IPOs underwritten by relationship banks, Journal of Financial Intermediation 19, Berger, A., and Udell, G., 1998, The economics of small business finance: The roles of private equity and debt markets in the financial growth cycle, Journal of Banking and Finance 22, Berger, A., Cerqueiro, G., and Penas, M., 2011, Does debtor protection really protect debtors? Evidence from the small business credit market, Journal of Banking & Finance 35, Berkowitz, J., and White, M.J., 2004, Bankruptcy and small firm s access to credit, The RAND Journal of Economics 35, Berlin, M., and Loeys, J., 1988, Bond covenants and delegated monitoring, Journal of Finance 43, Billett, M., Flannery, M., and Garfinkel, J., 1995, The effect of lender identify on a borrowing firm's equity return, Journal of Finance 50,

30 Billett, M. T., Flannery, M. J., and Garfinkel, J., 2006, Are bank loans special? Evidence on the post announcement performance of bank borrowers, Journal of Financial and Quantitative Analysis 41, Blanchard L., Zhao B., and Yinger J., 2008, Do lenders discriminate against minority and woman entrepreneurs? Journal of Urban Economics 63, Blanchflower, D., Levine, P., and Zimmerman, D., 2003, Discrimination in the small business credit market, Review of Economics and Statistics 84, Brav, O., 2009, Access to capital, capital structure, and the funding of the firm, Journal of Finance 64, Brealey, R., Myers, S., and Allen, F., 2011, Principles of Corporate Finance. McGraw-Hill. ISBN Cassar, G., 2004, The financing of business start-ups, Journal of Business Venturing 19, Cassar, G., 2014, Industry and startup experience on entrepreneur forecast performance in new firms, Journal of Business Venturing 29, Cavalluzzo, K., and Cavalluzzo, L., 1998, Market structure and discrimination: The case of small businesses, Journal of Money, Credit and Banking 30, Cavalluzzo, K., and Wolken, J., 2005, Small business loan turndowns, personal wealth and discrimination, Journal of Business 78, Cole, R., 1998, The importance of relationships to the availability of credit, Journal of Banking and Finance 22, Cole, R., 2009, Who needs credit and who gets credit? Evidence from the Surveys of Small Business Finances, In Small Business in Focus: Finance. A Compendium of Research by the Small Business Administration Office of Advocacy, July, Cole, R., 2010, Bank credit, trade credit or no credit? Evidence from the Surveys of Small Business Finances, U.S. Small Business Administration Research Study No Cole, R., 2013, What do we know about the capital structure of privately held firms? Evidence from the Surveys of Small Business Finances, Financial Management 42,

31 Coleman, S., 2002, The borrowing experience of Black and Hispanic-owned small firms: Evidence from the 1998 Survey of Small Business Finances, The Academy of Entrepreneurship Journal 8, Coleman, S. and Robb, A., 2009, A comparison of new firm financing by gender: Evidence from the Kauffman Firm Survey data, Small Business Economics 33, Coleman, S. and Robb, A., 2012, Capital structure theory and new technology firms: Is there a match? Management Research Review 35, Cooper, A. C., Gimeno-Gascon J., and Woo, C. W., 1994, Initial human and financial capital as predictors of new venture performance, Journal of Business Venturing 9, Cox, D. R., 1972, Regression models and life tables, Journal of the Royal Statistical Society, Series B (Methodology) 34, DesRoches, D., Potter, F, Santos, B, Sengmavong, A, and Zheng, Y., 2012, Kauffman Firm Survey (KFS) sixth follow up methodology report. Available at: Diamond, D.W., 1984, Financial intermediation and delegated monitoring, Review of Economic Studies 51, Diamond, D.W., 1991, Monitoring and reputation: The choice between bank loans and directly placed debt, Journal of Political Economy 99, Fairlie, R. and Robb, A., 2009, Gender differences in business performance: Evidence from the characteristics of business owners survey, Small Business Economics 33, Fama, E.F., 1985, What s different about banks? Journal of Monetary Economics 10, Fields, L. P., Fraser, D.R., Berry, T.L., and Byers, S., 2006, Do bank loan relationships still matter? Journal of Money, Credit, and Banking 38, Gonzalez, L., and James, C., 2007, Banks and bubbles: How good are bankers at spotting winners? Journal of Financial Economics 86, Halvorsen, R. and Palmquist, R., 1980, The interpretation of dummy variables in semilogarithmic equations, The American Economic Review 70, Harris, M. and Raviv, A., 1991, The theory of capital structure, Journal of Finance 46,

32 Heckman, J., 1979, Sample selection bias as a specification error, Econometrica 47, Hoshi, T., Kashyap, A., and Scharfstein, D., 1990, The role of banks in reducing the costs of financial distress in Japan, Journal of Financial Economics 27, James, C., 1987, Some evidence on the uniqueness of bank loans, Journal of Financial Economics 19, James, C., and Wier, P., 1990, Borrowing relationships, intermediation, and the cost of issuing public securities, Journal of Financial Economics 28, Jensen, M., and Meckling, W., 1976, Theory of the firm: Managerial behavior agency costs and capital structure, Journal of Financial Economics 3, King, R. G., and Levine, R., 1993a, Finance and growth: Schumpeter might be right, Quarterly Journal of Economics 108, King, R. G., and Levine, R., 1993b, Finance, entrepreneurship, and growth: Theory and evidence, Journal of Monetary Economics 32, Lummer, S., and McConnell, J., 1989, Further evidence on the bank lending process and the capital market response to bank loan agreements, Journal of Financial Economics 25, Mikkelson, W. H., and Partch, M. M., 1986, Valuation effects of securities offerings and the issuance process, Journal of Financial Economics 15, Myers, S., 1984, The capital structure puzzle, Journal of Finance 39, Petersen, M. and Rajan, R., 1994, The benefits of lending relationships: Evidence from small business data, Journal of Finance 46, Rajan, R.G, and Zingales, L, 1998, Which capitalism? Lessons from the East Asian crisis, Journal of Applied Corporate Finance 11, Ramakrishan, R.T.S., and Thakor, A.V., 1984, Information reliability and a theory of financial intermediation, Review of Economic Studies 51, Robb, A.M., R. Fairlie, and D.T. Robinson, 2009, Financial capital injections among new black and White business ventures: Evidence from the Kauffman Firm Survey, working paper

33 Robb, A.M. and D.T. Robinson, 2014, The capital structure decisions of new firms, Review of Financial Studies 27, Robb, A.M., and J. Watson, 2012, Gender differences in firm performance: Evidence from new ventures in the United States, Journal of Business Venturing 27, Ross, S., Westerfield, R., and J. Jaffe., 2012, Corporate Finance, 10 th edition. McGraw Hill. ISBN Schenone, C., 2004, The effect of banking relationships on the firm's cost of equity in its IPO, Journal of Finance 59, Schwienbacher, A., 2013, The entrepreneur's investor choice: The impact on later-stage firm development, Journal of Business Venturing 28, Shaffer, S. and T. Sokolyk, 2013, Bank loans to newly public firms, Journal of Entrepreneurial Finance 16, Shumway, T., 2001, Forecasting bankruptcy more accurately: A simple hazard model, Journal of Business 74, Slovin, M.B., Sushka, M.E., and Poloncheck, J. A., 1993, The value of bank durability borrowers as bank stakeholders, Journal of Finance 48, Stiglitz, J., and Weiss, A., 1981, Credit rationing in markets with imperfect information. American Economic Review 71, Wendt, P., 1947, The Availability of Capital to Small Businesses in California, Journal of Finance 2,

34 Table I KFS Variable Definitions This table presents variable definitions. All variables are from the 2004 Kauffman Firm Survey. Credit Variables: Credit Business Credit Personal Credit Trade Credit Other Sources of Capital: Insider Equity Insider Debt Outsider Equity Dummy variable, equals 1 if firm reports that it used either business credit, personal credit, or trade credit. Dummy variable, equals 1 if firm reports that it used business credit. Business credit includes any of the following categories: business bank loan, business credit line, business loan from nonbank institutions, business credit card, business credit card issued on owner s name, business loan from the government, business loan from other businesses, business loan from other sources. Dummy variable, equals 1 if firm reports that it used personal credit. Personal credit includes any of the following categories: personal bank loan by the primary owner, personal bank loan by other owners, the primary owner s personal credit card used for business purposes, and the other owners personal credit cards used for business purposes. Dummy variable, equals 1 if firm reports that it used trade credit. Dummy variable, equals 1 if either spouse or parent provided equity financing. Dummy variable, equals 1 if either family, employee, or any firm owner loaned money to the firm, primary owner, or other owners. Dummy variable, equals 1 if informal investors (i.e., angel investors), businesses, government, venture capitalists, or other entities provided equity financing

35 Table I (continued) Outsider Debt/Totcap Outsider debt, divided by total financial capital (debt + equity). Outsider debt is as defined in Robb and Robinson (2014) as the sum of balances of the following sources of capital: personal bank loan by the primary owner, personal bank loan by other owners, business bank loan, business credit line, business loan from nonbank institutions, business credit card, business credit card issued on owner s name, business loan from the government, business loan from other businesses, business loan from other sources, other individual loans. Owners Equity/Totcap Firm Characteristics: Revenue Total Assets Net Income ROA Cash Current Assets Tangible Assets Credit Risk Corp Non-Corp Multiown Total equity invested by all owners divided by total financial capital (debt + equity). Annual revenue from sales of product or service. Total assets (sum of cash, current assets, and tangible assets) Annual profit or loss (profit positive, loss negative) Net income divided by total assets Cash Sum of accounts receivable and inventory Sum of equipment, land/building, vehicles, other business property, and other assets such as intangibles Categorical variable (1 to 5) based on the credit score of the firm derived from Dunn and Bradstreet U.S. Ratings and Scores. A firm with a credit risk of 1 has the highest credit quality; a firm with a credit risk of 5 has the lowest credit quality. Firm is organized as an S-corporation, C-corporation, or Limited Liability Company/Partnership (LLC/LLP) Firm is organized as a Sole-proprietorship or Partnership Firm has more than one owner

36 Table I (continued) Intell Property Comp Advantage Product Product & Service Urban Owner Characteristics: Primary Owner Ownership Owner Age Female Asian Black Hispanic White Other Race High School College Education Graduate Degree Prior Experience Prior Start-ups Hours worked Dummy variable, equals 1 if firm reports that it has trademarks, patents, or copyrights. Dummy variable, equals 1 if firm reports that it has a comparative advantage. Dummy variable, equals 1 if firm only sells product. Dummy variable, equals 1 if firm sells product and service Dummy variable, equals 1 if business location is in a metropolitan statistical area (MSA). Owner with the highest percentage of firm ownership Firm ownership (in %) by primary owner Age of primary owner (in years) Primary owner is female Primary owner is Asian Primary owner is Black Primary owner is Hispanic Primary owner is White Primary owner is other than White, Asian, Hispanic, or Black Primary owner is either a high school graduate, has some high school education but no diploma, or has less than ninth-grade education Primary owner has either attended some college, has a bachelor s degree, or may have attended a graduate school but has no graduate degree Primary owner has a graduate degree (master s, professional school, or doctorate) Prior work experience (in years) of the primary owner in the same industry Number of prior business start-ups by the primary owner Number of hours worked per week by the primary owner

37 Table I (continued) Industry Classifications: Two-Digit NAICS Code Agriculture, Forestry, Fishing and Hunting 11 Mining and Utilities 21, 22 Construction 23 Manufacturing Wholesale Trade 42 Retail Trade Transportation and Warehousing Information 51 Finance and Insurance 52 Real Estate and Rental and Leasing 53 Professional, Management, and Educational Services 54, 55, 61 Administrative and Support and Waste Management and Remediation Services 56 Health Care and Social Assistance 62 Arts, Entertainment, and Recreation 71 Accommodation and Food Services 72 Other Services, including Public Administration 81,

38 Table II Descriptive Statistics of Start-up Firms This table reports descriptive statistics of Kauffman Firm Survey 2004 start-up firms. Variable definitions are provided in Table I. Survey weights are applied. Variable 25 th Percentile Mean Median 75 th Percentile Standard Error Number of Observations Sources of Capital: Credit (any type) ,928 Personal Credit ,928 Business Credit ,928 Trade Credit ,928 Insider Equity ,928 Insider Debt ,928 Outsider Equity ,928 Outsider Debt/Totcap ,928 Owners Equity/Totcap ,919 Firm Characteristics: Revenue ($) 0 229,789 7,500 62,500 80,263 4,741 Net Income ($) -10, ,000 6,455 4,586 Total Assets ($) 3, ,177 20,000 76,000 81,292 4,818 ROA ,123 Cash ($) 0 37,682 2,000 10,000 6,047 4,680 Current Assets ($) 0 125,983 1,000 15,000 65,263 4,755 Tangible Assets ($) 0 186,548 5,000 29,000 39,190 4,810 Corp ,928 Credit Risk ,606 Rural ,928 37

39 Table II (continued) Variable 25 th Percentile Mean Median 75 th Percentile Standard Error Number of Observations Multiown ,923 Ownership ,880 Comp Advantage ,858 Product ,928 Product & Service ,928 Primary Owner Characteristics: Owner Age ,860 Hours Worked ,825 Prior Experience ,907 Prior Start-ups ,893 Female ,920 Asian ,888 Black ,888 Hispanic ,888 White ,888 Other Race ,888 High School ,895 College Education ,895 Graduate Degree ,895 38

40 Table III Factors Explaining the Use of Credit at the Firm s Start-Up This table reports odds ratios from a weighted bivariate probit selection model. The sample includes Kauffman Firm Survey 2004 start-up firms. Column 1 presents the results from the first-stage probit model, examining the determinants of credit use. The dependent variable, Credit, equals 1 if the firm reports that it uses trade, business, or personal credit, and equals 0 otherwise. Column 2 presents the results from the second-stage regression, where the dependent variable, Business Credit, equals 1 if the firm reports it uses business credit, and equals 0 otherwise. Column 3 presents the results from the second-stage regression, where the dependent variable, Personal Credit, equals 1 if the firm reports it uses personal credit, and equals 0 otherwise. t-statistics are in parentheses. Variable definitions appear in Table I. State-Level SME Lending is a predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption. Industry dummies (based on two-digit NAICS codes) are included but omitted from the table for the sake of brevity. Survey weights are applied. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively Variable First stage: Second Stage: Second Stage: Credit Business Credit Personal Credit Firm Characteristics: Ln (Revenue+1) (6.25)*** (5.27)*** (-4.39)*** Ln (Total Assets+1) (5.79)*** (7.46)*** (-3.54)*** Corp (2.12)** (4.70)*** (-0.63) Multiown (-0.64) (1.54) (-2.49)** Credit Risk (-2.40)** (-2.98)*** (2.45)** Intell Property (-1.13) (1.38) (0.86) Comp Advantage (2.39)** (-0.16) (-1.33) Product (2.73)*** (0.00) (-0.65) Product & Service (-1.81)* (0.14) (0.08) Rural (-0.85) (-0.29) (0.02) 39

41 Table III (continued) Variable First stage: Second Stage: Second Stage: Credit Business Credit Personal Credit Owner Characteristics: Female (-0.84) (-1.38) (2.18)** Hours Worked (2.71)*** (2.25)** (0.34) Prior Experience (-3.10)*** (-1.36) (-2.15)** Prior Start-Ups (-0.47) (-1.25) (-1.33) Owner Age (0.31) (1.14) (-0.71) Owner Age (0.36) (1.23) (0.96) Asian (-0.16) (0.78) (0.18) Black (-5.76)*** (-3.37)*** (-0.27) Hispanic (-0.22) (-1.09) (-0.20) Other (0.14) (1.36) (0.89) College Education (1.80)* (0.95) (0.73) Graduate Degree (3.68)*** (1.81)* (-0.35) 40

42 Table III (continued) Other Sources of Capital: Log(Owners' Equity+1) (1.93)* (0.81) (1.03) Insider Equity (2.42)** (1.63) (-0.58) Insider Debt (-0.31) (-2.19)** (2.30)** Outsider Equity (-0.57) (-0.63) (0.34) State-Level SME Lending (1.36) F-statistic 7.80*** 3.66*** p-value Number of Observations 3,243 3,243 3,243 41

43 Table IV Credit Use and Firm Performance: Survival Analysis This table reports hazard ratios from the Cox proportional hazard model. The dependent variable is the probability that firm i will be out of business by year t, conditioning on firm i surviving up to time t (= ). Independent variables are from KFS 2004 and are described in Table I. Where indicated, regressions include controls for Firm Characteristics, Owner Characteristics, Other Sources of Capital, State-Level SME Lending, and Industry Classifications. State-Level SME Lending is a predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption. Survey weights are applied. z-statistics are reported in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Variable Credit (-2.26)** Business Credit (-2.51)** (-3.29)*** (-2.72)*** (-2.63)*** Personal Credit (0.64) (1.24) (0.92) (0.51) Firm Characteristics No No Yes Yes Yes Owner Characteristics No No No Yes Yes Other Sources of Capital No No No No Yes State-Level SME Lending No No No No Yes Industry Classifications No No Yes Yes Yes Number of Observations 4,301 4,301 3,010 2,904 2,869 F-statistic 5.09** 3.24** 3.43*** 3.16*** 2.86*** 42

44 Table V Credit Use and Firm Performance: Revenue Analysis This table reports Weighted Least Squares regressions of revenue, measured as the natural logarithm of one plus the level of revenue three years after the firm s start-up (KFS 2007). Independent variables are from KFS 2004 and are described in Table I. Inverse Mills Ratio is estimated from the probit regression of the firm s survival to Where indicated, regressions include controls for Firm Characteristics, Owner Characteristics, Other Sources of Capital, State-Level SME Lending, and Industry Classifications. State-Level SME Lending is a predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption. Survey weights are applied. z-statistics are reported in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Variable Credit (5.54)*** Business Credit (6.45)*** (3.39)*** (2.85)*** (2.60)*** Personal Credit (-0.90) (-2.05)** (-2.28)** (-2.35)** Inverse Mills Ratio (3.41)*** (3.06)*** (3.25)*** (0.95) (-0.26) Firm Characteristics No No Yes Yes Yes Owner Characteristics No No No Yes Yes Other Sources of Capital No No No No Yes State-Level SME Lending No No No No Yes Industry Classifications No No Yes Yes Yes Number of Observations 1,990 1,990 1,953 1,953 1,953 R

45 Table VI Credit Use and Firm Performance: Propensity-Score-Matched Sample This table reports the results of survival (Panel A) and revenue (Panel B) analysis from the propensityscore-matched sample. Propensity-score matching is performed on three firm characteristics from KFS 2004: credit score, revenue, and total assets. For survival analysis, the table reports hazard ratios from the Cox proportional hazard model, where the dependent variable is the probability that firm i will be out of business by year t, conditioning on firm i surviving up to time t (= ). For revenue analysis, the table reports Weighted Least Squares regressions of revenue, measured as the natural logarithm of 1 plus the level of revenue three years after the firm s start-up (KFS 2007). Independent variables are from KFS 2004 and are described in Table I. IMR Matched is estimated from a probit model, where the dependent variable is equal to 1 if the firm is in the matched sample and is equal to 0 otherwise, and the explanatory variables are the same as included in the survival model shown in eq. (5). Regressions include controls for Firm Characteristics, Owner Characteristics, Other Sources of Capital, State-Level SME Lending, and Industry Classifications. State-Level SME Lending is a predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption. Survey weights are applied. z- and t-statistics are reported in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Panel A: Survival Analysis Panel B: Revenue Analysis Variable Credit (-1.75)* (2.01)** Business Credit (-2.37)** (2.00)** Personal Credit (0.02) (0.57) IMR Matched (1.49) (1.73)* (0.70) (0.59) Firm Characteristics Yes Yes Yes Yes Owner Characteristics Yes Yes Yes Yes Other Sources of Capital Yes Yes Yes Yes State-Level SME Lending Yes Yes Yes Yes Industry Classifications Yes Yes Yes Yes Number of Observations 1,253 1, F-statistic / R *** 44.28***

46 Table VII Credit Use and Firm Performance: Endogeneity Analysis This table reports the results of survival (Panel A) and revenue (Panel B) analysis controlling for the endogeneity between the probability of obtaining credit and firm performance outcomes. For survival analysis, the table reports hazard ratios from the Cox proportional hazard model, where the dependent variable is the probability that firm i will be out of business by year t, conditioning on firm i surviving up to time t (= ). For revenue analysis, the table reports Weighted Least Squares regressions of revenue, measured as the natural logarithm of 1 plus the level of revenue three years after the firm s startup (KFS 2007). Use Credit #, Use Bank Credit #, and No Need for Credit # are instrumental variables for Credit, Business Credit, and Personal Credit, respectively, estimated using data on credit financing from SSBF (see section III.B.4 in the paper and Appendix Table A.4). Independent variables are from KFS 2004 and are described in Table I. Regressions include controls for Firm Characteristics, Owner Characteristics, Other Sources of Capital, State-Level SME Lending, and Industry Classifications. State- Level SME Lending is a predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption (see Section II.B.1 in the paper). Survey weights are applied. z- and t-statistics are reported in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Panel A: Survival Analysis Panel B: Revenue Analysis Variable Use Credit # (-0.23) (3.01)*** Use Bank Credit # (-0.13) (2.23)** No Need for Credit # (-0.58) (-0.27) Firm Characteristics Yes Yes Yes Yes Owner Characteristics Yes Yes Yes Yes Other Sources of Capital Yes Yes Yes Yes State-Level SME Lending Yes Yes Yes Yes Industry Classifications Yes Yes Yes Yes Number of Observations 2,869 2,869 1,953 1,953 F-statistic / R

47 Appendix Tables Table A.1 Credit Use and Firm Performance: Complete Set of Results from Survival Analysis This table reports hazard ratios for explanatory and control variables from the Cox proportional hazard model. A condensed version of this table is presented in Table IV. The dependent variable is the probability that firm i will be out of business by year t, conditioning on firm i surviving up to time t (= ). Independent variables are from KFS 2004 and are described in Table I. State-Level SME Lending is a predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption. Industry dummies are included in regressions for columns 3-5 but are unreported. Survey weights are applied. z-statistics are reported in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. (continues) 46

48 Table A.1 (continued) Variable Credit (-2.26)** Business Credit (-2.51)** (-3.29)*** (-2.72)*** (-2.63)*** Personal Credit (0.64) (1.24) (0.92) (0.51) Firm Characteristics: Ln (Revenue+1) (-0.89) (-0.14) (-0.42) Corp (1.16) (2.24)** (2.02)** Multiown (-0.01) (-0.69) (-0.76) Credit Risk (1.56) (1.34) (1.35) Intell Property (-1.99)** (-1.66)* (-1.73)* Comp Advantage (-0.84) (-0.49) (-0.53) Product (2.20)** (1.98)** (1.55) Product & Service (-3.07)*** (-3.24)*** (-3.00)*** Rural (0.51) (-0.13) (-0.06) Owner Characteristics: Female (1.79)* (1.55) Hours Worked (-1.30) (-1.80)* Prior Experience (-3.53)*** (-3.17)*** Prior Start-Ups (2.56)** (2.40)** 47

49 Table A.1 (continued) Variable Owner Characteristics: Owner Age (-2.12)** (-1.90)* Owner Age (1.96)* (1.75)* Asian (-2.34)** (-2.36)** Black (0.74) (0.55) Hispanic (1.36) (0.67) Other (1.01) (0.74) College Education (-2.93)*** (-2.99)*** Graduate Degree (-3.05)*** (-3.14)*** Other Sources of Capital: Trade Credit (0.67) Owners' Equity/Totcap (0.36) Outsiders' Debt/Totcap (0.31) Insider Equity (1.56) Insider Debt (1.56) Outsider Equity (-0.11) State-Level SME Lending (-2.38)** Number of Observations 4,293 4,301 3,010 2,904 2,869 F-statistic 5.09** 3.24** 3.43*** 3.16*** 2.86*** 48

50 Table A.2 Credit Use and Firm Performance: Complete Set of Results from Revenue Analysis This table reports the results of Weighted Least Squares regressions of revenue on credit, firm, financing, industry, and owner characteristics. A condensed version of this table is presented in Table V. The dependent variable, Revenue, is the natural logarithm of 1 plus the level of revenue three years after the firm s start-up (KFS 2007). Independent variables are from KFS 2004 and are described in Table I. Inverse Mills Ratio is estimated from the probit regression of the firm s survival to State-Level SME Lending is the predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption. Industry dummies are included in the regressions for columns 3-5 but are unreported. t-statistics are reported in parentheses. Survey weights are applied. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Variable Credit (5.54)*** Business Credit (6.45)*** (3.39)*** (2.85)*** (2.60)*** Personal Credit (-0.90) (-2.05)** (-2.28)** (-2.35)** Firm Characteristics: Inverse Mills Ratio (3.41)*** (3.06)*** (3.25)*** (0.95) (-0.26) Ln (Revenue+1) (9.52)*** (8.11)*** (7.59)*** Corp (6.14)*** (4.50)*** (0.73) Multiown (0.15) (0.33) (0.40) Credit Risk (-0.18) (-0.42) (-0.42) Intell Property (-0.18) (-0.48) (-0.52) Comp Advantage (1.82)* (1.32) (0.79) Product (0.10) (0.13) (-0.36) Product & Service (1.17) (0.74) (0.53) Rural (-0.43) (-0.78) (-0.71) 49

51 Table A.2 (continued) Variable Owner Characteristics: Female (-2.32)** (-0.97) Hours Worked (3.09)*** (0.85) Prior Experience (-0.20) (0.33) Prior Start-Ups (1.15) (-0.11) Owner Age (-0.69) (0.22) Owner Age (0.38) (-0.27) Asian (-1.71)* (-0.29) Black (-2.80)*** (-1.49) Hispanic (-1.19) (-0.65) Other (-1.30) (-1.31) College Education (-0.88) (0.21) Graduate Degree (-0.52) (0.28) 50

52 Table A.2 (continued) Variable Other Sources of Capital Trade Credit (2.60)*** Owners' Equity/Totcap (0.12) Outsiders' Debt/Totcap (0.23) Insider Equity (-1.05) Insider Debt (-0.24) Outsider Equity (2.09)** State Level SME Lending (-0.74) Constant (4.62)*** (5.96)*** (-0.59) (0.97) (0.61) Number of Observations 1,990 1,990 1,953 1,953 1,953 R

53 Table A.3 Credit Use and Firm Performance: Complete Set of Results from Propensity-Score-Matched Sample This table reports the results of survival (Panel A) and revenue (Panel B) analyses from the propensityscore-matched sample. Propensity-score matching is performed on three firm characteristics from KFS 2004: credit score, revenue, and total assets. For survival analysis, the table reports hazard ratios from the Cox proportional hazard model, where the dependent variable is the probability that firm i will be out of business by year t, conditioning on firm i surviving up to time t (= ). For revenue analysis, the table reports Weighted Least Squares regressions of revenue, measured as the natural logarithm of 1 plus the level of revenue three years after the firm s start-up (KFS 2007). Independent variables are from KFS 2004 and are described in Table I. IMR Matched is estimated from a probit model, where the dependent variable is equal to 1 if the firm is in the matched sample and is equal to 0 otherwise, and the explanatory variables are the same as included in the survival model shown in eq. (5). State-Level SME Lending is a predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption. Survey weights are applied. z- and t-statistics are reported in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Panel A: Survival Analysis Panel B: Revenue Analysis Credit (-1.75)* (2.01)** Business Credit (-2.37)** (2.00)** Personal Credit (0.02) (0.57) Firm Characteristics: IMR Matched (1.49) (1.73)* (0.70) (0.59) Ln (Revenue+1) (-1.43) (-1.61) (-0.26) (-0.16) Corp (-0.08) (0.11) (1.38) (1.36) Multiown (0.44) (0.40) (-0.09) (-0.06) Credit Risk (3.03)*** (3.17)*** (0.04) (-0.01) Intell Property (-0.29) (-0.14) (0.06) (-0.01) Comp Advantage (-1.22) (-1.33) (1.62) (1.81)* 52

54 Table A.3 (continued) Panel A: Survival Panel B: Revenue Product (-0.79) (-0.81) (-1.03) (-0.95) Product & Service (-1.41) (-1.39) (1.46) (1.39) Rural (-0.39) (-0.35) (1.51) (1.55) Owner Characteristics: Female (1.14) (1.21) (-1.88)* (-1.87)* Hours Worked (-1.85)* (-1.96)* (1.90)* (1.93)* Prior Experience (-2.06)** (-2.02)** (0.92) (0.81) Prior Start-Ups (0.08) (0.04) (1.05) (1.10) Owner Age (-1.95)* (-1.97)** (1.17) (1.13) Owner Age (1.81)* (1.80)* (-1.53) (-1.46) Asian (-1.51) (-1.57) (-0.93) (-0.89) Black (-0.20) (-0.09) (-1.28) (-1.37) Hispanic (0.02) (-0.19) (-1.17) (-1.02) Other (0.13) (0.21) (-0.02) (-0.11) College Education (-2.41)** (-2.57)** (-1.30) (-1.35) Graduate Degree (-2.27)** (-2.60)*** (-1.37) (-1.26) 53

55 Table A.3 (continued) Panel A: Survival Panel B: Revenue 1 2 Other Sources of Capital Trade Credit (-1.03) (-1.42) (-0.34) (-0.13) Owners' Equity/Totcap (-0.58) (0.15) (1.60) (1.34) Outsiders' Debt/Totcap (-1.33) (-1.00) (-0.35) (-0.48) Insider Equity (1.00) (1.02) (-1.82)* (-1.74)* Insider Debt (-0.34) (0.06) (1.88)* (1.60) Outsider Equity (0.63) (0.91) (1.95)* (1.71)* State Level SME Lending (-0.41) (-0.42) (-0.03) (-0.15) Constant (0.73) (0.60) Number of Observations 1,253 1, F-statistic / R *** 44.28***

56 Table A.4 Instrumental Variables for Credit Use The table reports the coefficients used to create instrumental variables (Panel A) and the regressions of credit use variables on instrumental variables (Panel B). The process for instrumental variables estimation is discussed in section III.B.4. Explanatory variables are defined in Table 1. Coefficients are estimated using data from the Federal Reserve s 2003 Survey of Small Business Finances (SSBF), and then applied to data from the Kauffman Firm Survey (KFS) to create the instrumental variables. Use Credit is an SSBF indicator variable that is equal to 1 if the firm reported that it had one or more outstanding loans or used trade credit, and equal to 0 otherwise; it is used as an instrument for the KFS variable Credit. Use Bank Credit is an SSBF indicator variable that is equal to 1 if the firm reported that it had one or more outstanding loans from a commercial bank, and equal to 0 otherwise; it is used as an instrument for the KFS variable Business Credit. No Need for Credit is an SSBF indicator variable that is equal to 1 if the firm reported that it had no need for credit during the previous three years, and equal to 0 otherwise; it is used as an instrumental variable for the KFS variable Personal Credit. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Panel A: Coefficient Estimates from SSBF data Model 1 Use Credit Model 2 Use Bank Credit Model 3 No Need for Credit Intercept (-2.44)** (3.93)*** (0.03) Ln(Revenue+1) (8.96)*** (7.23)*** (-4.60)*** Corp (2.19)** (4.18)*** (-3.61)*** Rural (2.02)** (0.80) (-0.97) Credit Score (1.15) (-0.40) (3.15)*** Owner Age (-2.53)** (-1.45) (4.24)*** Graduate Degree (-2.12)** (-1.65)* (3.03)*** College Education (-2.64)*** (-1.73)* (4.80)*** Female (-3.38)*** (-3.78)*** (-1.21) 55

57 Table A.4 (continued) Panel B: Regressions of Credit Use Variables on Instrumental Variables Dependent Variable: Variable Credit Business Credit Personal Credit Use Credit # (11.66)*** Use Bank Credit # (13.76)*** No Need for Credit # (-4.61)*** Constant (14.14)*** (-13.05)*** (7.33)*** Number of observations 3,423 3,423 3,423 Pseudo R

58 Table A.5 Credit Use and Firm Performance: Complete Set of Results from Endogeneity Analysis The table reports the results of survival (Panel A) and revenue (Panel B) analysis controlling for the endogeneity between the probability of obtaining credit and firm performance outcomes. For survival analysis, the table reports hazard ratios from the Cox proportional hazard model, where the dependent variable is the probability that firm i will be out of business by year t, conditioning on firm i surviving up to time t (= ). For revenue analysis, the table reports Weighted Least Squares regressions of revenue, measured as the natural logarithm of 1 plus the level of revenue three years after the firm s startup (KFS 2007). Use Credit #, Use Bank Credit #, and No Need for Credit # are instrumental variables for Credit, Business Credit, and Personal Credit, respectively, estimated using data on credit financing from SSBF (see section III.B.4 in the paper). Independent variables are from KFS 2004 and are described in Table I. State-Level SME Lending is a predicted value from the regression of the amount of state-level bank lending to small and medium enterprises, scaled by the number of small/medium firms in the state, on the amount of state-level homestead bankruptcy exemption. Survey weights are applied. z- and t- statistics are reported in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Panel A: Survival Analysis Panel B: Revenue Analysis Variable Use Credit # (-0.23) (3.01)*** Use Bank Credit # (-0.13) (2.23)** No Need for Credit # (-0.58) (-0.27) Firm Characteristics: Ln (Revenue+1) (-0.05) (0.26) (1.22) (1.25) Corp (1.74)* (1.45) (3.53)*** (2.37)** Multiown (-0.84) (-0.89) (0.55) (0.59) Credit Risk (1.37) (1.46) (-0.54) (-0.82) Intell Property (-1.78)* (-1.78)* (-0.39) (-0.43) Comp Advantage (-0.56) (-0.57) (1.84)* (1.83)* Product (1.55) (1.56) (-0.23) (-0.22) 57

59 Table A.5 (continued) Product & Service (-3.00)*** (-3.00)*** (1.67)* (1.66)* Rural (0.10) (0.15) (-1.63) (-0.93) Female (1.38) (1.58) (-1.45) (-1.64) Hours Worked (-1.83)* (-1.87)* (4.41)*** (4.61)*** Prior Experience (-3.18)*** (-3.20)*** (0.77) (0.74) Prior Start-Ups (2.46)** (2.46)** (0.13) (0.17) Owner Age (-1.91)* (-1.91)* (0.00) (-0.10) Owner Age (1.78)* (1.65)* (0.28) (0.32) Asian (-2.38)** (-2.38)** (-1.52) (-1.55) Black (0.74) (0.71) (3.20)*** (3.20)*** Hispanic (0.74) (0.77) (-1.92)* (-1.90)* Other (0.68) (0.69) (-1.28) (-1.30) College Education (-2.91)*** (-2.54)** (0.46) (-0.11) Graduate Degree (-3.18)*** (-3.03)*** (0.79) (0.39) Other Sources of Capital Trade Credit (0.03) (0.04) (0.27) (0.22) Owners' Equity/Totcap (-0.57) (-0.58) (0.80) (0.77) Outsiders' Debt/Totcap (-1.04) (-1.04) (0.85) (0.84) Insider Equity (1.43) (1.42) (-1.88)* (-1.90)* 58

60 Table A.5 (continued) Insider Debt (1.48) (1.48) (0.22) (0.20) Outsider Equity (-0.13) (-0.15) (2.29)** (2.31)** State Level SME Lending (-2.38)** (-2.38)** Constant (-1.00) (0.38) Number of Observations 2,869 2,869 1,953 1,953 F-statistic / R