WHY DO SMALL FIRMS CHANGE BANKS? - Temple University

WHY DO SMALL FIRMS CHANGE BANKS?

Jonathan A. Scott Fox School of Business Temple University Philadelphia, PA 19122 USA

July 2006

Key words: small businesses, banking relationships, community banks JEL Classification Codes: G21, G28

Telephone: 215.204.7605 Email: [email protected] Thanks are due to Ken Cyree and Karyen Chu for their helpful comments on an earlier draft.

WHY DO SMALL FIRMS CHANGE BANKS?

Abstract This paper investigates why small firms change banks using survey data obtained from a sample of U.S. small businesses. These data are unique because they include the reasons for changing banks (e.g., mergers, loan officer turnover, inadequate credit line) and owner ratings of characteristics important to conducting their banking business (e.g., the bank’s knowledge of their business, location, range of services offered). The primary findings of the paper are: 1) information-opaque small firms unlikely to be held-up by their current bank; 2) the reasons for changing banks are related to how owners rate the relative importance of their relationship versus transactions banking needs; and 3) owners changing banks are located further from their new bank with the least information-opaque firms moving the furthest away.

WHY DO SMALL FIRMS CHANGE BANKS? 1. Introduction and motivation The length of a bank-borrower relation – or duration of a relation – and its effect on small firm credit outcomes is well documented in the literature (e.g. Berger and Udell, 1995, and Boot, 2000). Private information gathered by lenders during the course of a banking relationship can help resolve the information asymmetry problems commonly faced by small information opaque firms (Boot, 2000). This private information may be “hard”, consisting of financial statements, tax returns, or usage of banking services or “soft”, consisting of qualitative assessments of the owner’s ability to repay their debt (Berger and Udell, 2002). A longer bank relation is generally associated with improved banking outcomes for a small firm as a result of private information accumulation over time. Elsas (2005) and Ongena and Smith (2001) note, however, that relation length is also a function of switching costs and potential “hold-up” problems (e.g., Sharpe, 1990, Rajan, 1992, or Greenbaum, Kanatas and Venezia, 1989). Thus, a long bank relation could be unfavorable to an information opaque borrower because they are held-up by the lender and cannot change banks to improve availability or loans terms. Not surprisingly, the empirical evidence is mixed for the association between the length of relation and credit market outcomes. A number of papers find that the length of relation is associated with improved access to credit and improved terms (e.g., Berger and Udell, 1995, Cole, 1998, Degryse and Van Cayseele, 2000, Petersen and Rajan, 1994). But there are exceptions (Angelini et al , 1998, Harhoff and Korting, 1998, and Cole et al, 2004, Degryse and Ongena, 2005 and Elsas, 2005), especially in regards to loan terms.

1

These mixed results could be due to differences in the net advantage of the private information accumulation through a longer relation versus the hold-up and switching costs (Ongena and Smith, 2001). The only paper that directly addresses the benefits and costs of switching banks is Ongena and Smith who used a sample of Norwegian firm-level data to estimate the probability of ending a bank relation. They find that the probability of ending a bank relation increases with duration and that the most information-opaque borrowers maintain the shortest relationships. They interpret their findings as a lack of evidence to support the presence of hold-up by lenders. Ongena and Smith note that their data do not allow them to analyze why these firms changed their primary bank. Nor were they able to address the location of the new bank – an important component of the switching costs involved in the change. The dramatic change in bank structure in the U.S. and the use of technology in the delivery bank services during the past 15 years raise a number of questions that motivate a need for additional analysis of why small firms change banks. Consolidation has resulted in fewer banking organizations, which could lead to less competition and an increased chance of hold-up. On the other hand, deregulation and financial innovation has created more competition in many product lines compressing other lending margins. Has this trend resulted in more competition for small firm lending business where there has been less innovation, mitigating hold-up problems? Improvements in information technology have facilitated the development of credit scoring for small firms as well as improvements in monitoring by banks (Akhavein et al, 2005). Have these developments made it easier for small firms to change banks, especially those more interested in

2

transactions versus relationship banking? If so, what are the characteristics of these firms? Finally, the community bank market share of U.S. deposits has been declining as documented by DeYoung et al (2004). Has this decline increased the distance and therefore switching costs for small firms that want a focus on relationship banking at a community bank? A better understanding of the tradeoffs small firms make between their reported reasons for switching, the size of the new bank chosen, and the distance to the new bank can help provide a more informed context of how recent changes in U.S. bank structure may have affected small firm access to credit. Similar to Ongena and Smith, a unique data set is used in the analysis that is comprised of survey data obtained in 2001 from the membership of the National Federation of Independent Business, a large U.S.-based small business trade organization. The survey asks firms when they most recently changed their primary bank, not how long they have been with their bank. Most importantly, owners report the reasons for the change that are sorted into four categories: relationship, loan term, merger and other. The owners also rate a set of characteristics for their importance to the conduct of their banking business such as location, the bank’s knowledge of their business and market, accessibility to loan officer, speed of decisions, etc. If the owner’s current bank is not meeting needs considered important for the conduct of their banking business, and the owner changes banks to get these needs met, the ratings can serve as a proxy for the benefits of changing banks. Also included in the survey is information on a set of banking and transactions services used by small firms that can serve as a proxy for the scope of the banking relation. And finally, the survey reports the distance to the owner’s current bank measured in time. Other firm characteristics are included such as years in

3

business and employment that can serve as proxies for the degree of information asymmetry. Three questions are addressed in the paper. First, is there any evidence that small U.S. firms are held-up by their banks after controlling for the degree of information asymmetry, scope of relationship, and U.S. market structure? Despite the many papers that have addressed the effect of the duration of bank relations on small firm banking outcomes, no work has directly examined the effect of duration of relation on potential hold-up of U.S. small firms. Given the importance of banks to small firm finance in the U.S. documented by Berger and Udell (2002), it is important to understand whether holdup is a pervasive problem among small firms. A finding similar to Ongena and Smith’s would suggest that researchers do not need to worry much about this problem in analyses of the costs and benefits of relationship banking to small firms. Second, does the profile of small firms differ by the reason given for the change? For example, are the firms that change for relationship reasons more or less informationopaque than those changing for loan term reasons? A related question is whether the reasons given by owners for changing banks related to the ratings of characteristics that they deem central to the conduct their banking business? Assuming that owners change banks because of some unmet need, the rating of the importance of these characteristics can identify the potential benefits of changing banks – a factor in the switching decision that has not been considered in previous work on this subject. In particular, differences in the ratings of these characteristics can provide additional insight into the profile of small firms that value relationship versus transaction oriented banking.

4

And finally, do owners that report a recent change in their primary bank locate further away and is distance related to the reason for changing? Petersen and Rajan (2002) provided the first documentation of the increasing distance between small firms and their primary bank in the U.S., but they only examined firms with loans outstanding from a synthetic panel. Since that time (early 1990s) mergers and acquisitions have continued with the number of banking organizations falling by over 2,800 between 1994 and 2004 (2.9% per year). Yet, the number of branches increased by almost 15,000 (2.7 percent per year) over the same period, which has the potential to put offices closer to their small firm customers. Although Peterson and Rajan found that firm age had no effect on distance from the owner’s bank, distance could still matter to small firms with severe asymmetric information problems that seek relationship bankers. These firms may change to a closer bank, while those less information-opaque firms may seek better transactions outcomes (e.g. lower prices) at a bank that is further away (e.g. Degryse and Ongena, 2005).1 A preview of the primary findings of the paper is as follows. First, no evidence is found to support the idea that U.S. small firms are held-up by their current bank, confirming the findings of Ongena and Smith for Norwegian firms. Second, owners changing for relationship reasons rate soft information aspects of banking more important to the conduct of their financial business (such as how well their bank knows their business, access to and social contact with their loan officer) compared to those changing for transactions reasons. Third, owners that changed banks were located further from their primary bank than those firms that did not switch - a result that corroborates the 1

For example, Degryse and Ongena (2005) have found evidence of spatial price discrimination in bank

lending with a negative relationship between loan rate and distance.

5

findings of Petersen and Rajan (2002) – with the least information opaque firms moving the furthest away. Survey research faces some unique problems associated with potential biases and measurement errors in the data. The data were obtained from the membership of the National Federation of Independent Business (“NFIB”) membership which may not be representative of overall small firm experience in the United States. However, as noted in Berger et al (2005), the hypotheses tested in the paper apply to all observations equally. Weighting would be important if the goal was to make population prediction about a specific probability (or log odds) of the incidence of changing banks with respect to various firm characteristics. Nonetheless, as part of the sensitivity tests, the data are weighted by region, industry and employment using Census Bureau enterprise survey data. Survey responses often reflect beliefs and not actions and therefore are sometimes considered less reliable than primary economic data or market prices. However, the focus of this analysis is on an action – a change of the owner’s primary bank. While the NFIB respondents may have misreported when they changed banks, there is no reason to believe that they systematically erred in reporting a bank change or distance in a way that the error is correlated with any of the independent variables. Thus, without any other specification problems, the coefficient estimates should be unbiased and consistent. The remainder of the paper is organized as follows. The survey data used in the analysis is described in section 2. Section 3 presents the methodology and empirical results, while concluding comments are offered in section 4.

6

2. Data 2.1 Survey description The data are obtained from the Credit, Banks and Small Business (“CBSB”) survey conducted by NFIB in late 2001. This survey, the sixth in a series that extends back to 1980, collects information about the credit market experiences of a random sample of the NFIB’s 600,000 members. Detail about the survey can be found at the NFIB’s web site (www.nfib.org/research). Questionnaires were mailed to 12,500 firms and responses were received from 2,223, a response rate of 18 percent. The data are similar to those gathered by the Board of Governors of the Federal Reserve System in the Survey of Small Business Finance (“SSBF”) through a commissioned telephone survey of small firms drawn from the Dun & Bradstreet files. Neither the NFIB nor SSBF surveys attempt an independent verification of the accuracy of the self-reported data.2 Frequency distributions for key firm characteristics are presented in the Appendix. The median years in business of all respondents is 17, median sales was $600,000, and median total employment is 6; the mean values are 19 years in business, $2.7 million average sales and 17 years in business. Retail trade is the most frequently represented sector (20 percent), followed by construction (15 percent), and business services (15 percent). About 20 percent of the respondents are from the Northeast, 22 percent from the Midwest, 33 percent from the South and 27 percent from the West, while 62 percent of the firms are located in a Metropolitan Statistical Area as defined by the U.S. Census Bureau. A full listing of the variables used in the analysis is presented in Table 1. 2

The Fed collects a significant amount of income statement and balance sheet data that is cross-checked

and adjusted for consistency and omissions using internally developed algorithms.

7

When compared to the enterprise data compiled by the U.S. Census Bureau in 1997, the NFIB sample tends to be somewhat larger in terms of employment per firm, under-represents the service sector, and over-represents agriculture and manufacturing.3 In addition, NFIB firms are under-represented along the Eastern seaboard and overrepresented in Midwest and Plains states. The effect of weighting, based on employment, 1-digit SIC classification and U.S. Bureau of Census geographic regions is also shown in the Appendix. Weighting places more emphasis on smaller, younger firms, those organized as proprietorships, those doing business in retail and services, and those located in the northeast and south compared to the original sample. The median firm age (years in business) for the weighted sample is now 15; the median total employment is 4.5; and the median gross sales are $400 thousand. 2.2 Reasons for changing banks A summary of the responses to the question “When was the last time you changed your primary financial institution?” for the last three NFIB surveys is presented in Table 2, Panel A. A slight downward trend in the number of firms changing banks is apparent, perhaps reflecting the absolute decline in the number of banks resulting from industry consolidation over this between 1987 and 2001. In 1987, 19 percent of the firms reported changing banks within the past three years, decreasing to 17 percent in the 1995 survey, and 14 percent in the 2001 survey.

3

Enterprise data refers to business owners versus establishments that count all locations opened by a single

owner.

8

Fifty-three percent (1,188 respondents) reported changing financial institutions at some point in time in the company’s history (Table 2, Panel B).4 “Mergers” (labeled Merger change in Table 1) was the most frequently reported single reason for those firms changing banks in 2001, 2000, and 1998 accounting for 18 percent of all reasons cited for changing banks. “Other” (included in Other change in Table 1) was the overall most frequently reported single reason and increased in importance over time.5 “Loan officer turnover,” “followed loan officer,” and “treated as a stranger” are combined in a category labeled Relationship Change in Table 1, which constitutes 28 percent of the reasons given. “Inadequate credit line,” “loan terms too difficult,” and “couldn’t get desired loan size” are combined in a category labeled Loan Term Change, which comprises 25 percent of the responses 6 2.3 Characteristics central to the conduct of banking business The survey includes 11 characteristics that are rated by the owners on a scale from 1 (not important) to 5 (very important) for importance in the conduct of their financial business at their primary financial institution. These characteristics have been included in every CBSB survey since 1980. These ratings can serve as a proxy for the benefits of changing banks if owners change banks because of some unmet banking need important to the conduct of their financial business.

4

It is possible that owners using multiple banks dropped a bank but still reported no change here because

no action was taken with its primary bank. 5

This high frequency may be due to memory lapses by the owners who forgot why they changed,

especially if the change has not been within the past several years. 6

Ideally a category labeled ‘Transaction Changes’ that includes loan terms, service offerings and service

quality would have been the first choice had the survey included these reasons.

9

The top three characteristics based on the mean ratings shown in the Table 1 are knows my business, reliable source of credit, and speed of decisions. The three least important characteristics are social contact with loan officer, provides helpful suggestions, and knows my industry. The standard deviation of the ratings tends to vary inversely with the rating, i.e. the highest rated characteristics have the lowest standard deviation.7 This outcome suggests that the most important characteristics are more widely shared by the owners with less variance than those that are less important. All of the importance ratings are significantly correlated at the .01 level (Table 3). However, the correlations are generally not high, ranging from a low of 0.096 between location important and knows my industry to a high of 0.620 between speed of decisions and easy access to loan officer. The correlation coefficients between the importance characteristics and the incidence of changing banks, labeled Changed banks (1=’yes’ and 0=’no’) are also presented in Table 3. Four of the characteristics are positively correlated with the incidence of changing banks at the five percent level: knows my business, convenient location, speed of decisions and easy access to loan officer. These characteristics reflect a mix of relationship attributes (knows my business and easy access to loan officer) and transaction attributes (speed of decisions and convenient location). The negative correlation of location suggests that owners valuing proximity to their current bank are less likely to be changing banks. 2.4 Distance to bank The survey asks owners to report the approximate time it takes to travel from their place of business to their principal financial institution. The average (median) reported

7

F-tests for equality of variance, though not reported here, confirm this conclusion.

10

time is 9.5 (5.0) minutes, ranging from one to 90 minutes (see Table 1). Firms that changed banks are located 13 percent (10 minutes versus 8.8 minutes) further away from their bank compared to firms that did not change banks (Table 2, Panel C). These distances are longer than those reported by Degryse and Ongena (2005) who compute a mean distance in time of 6.9 minutes and median of 4.3 minutes. This difference is not surprising given the size and corresponding population density differences between the U.S. and Belgium. The mean distance in time is not directly comparable to Petersen and Rajan (2002) who found a mean (median) distance 43 (4) miles between the firm and its bank lender using data from 1993 and 1988. If an average speed of 25 to 40 miles per hour is used, a comparable median distance in miles is between 2.1 to 3.3 miles. Distance to their primary bank is also positively correlated with the incidence of changing (Table 3). This result suggests that owners changing to new banks are locating further away, consistent with Petersen and Rajan (2002). Although Petersen and Rajan (2002) documented an increase in distance over time, the mean distances for firms that changed banks shown in Table 2, Panel C suggests little change over time. 3. Methodology and empirical results 3.1 Incidence of changing banks 3.1.1 Empirical strategy The null hypothesis evaluated in this section is that more information opaque owners are less likely to report changing banks. The framework for testing this hypothesis is that the incidence of changing banks is a function of its benefits and costs. The benefits of changing should be related to the better service or lending terms that an

11

owner would receive at the new bank, while the costs include both holdup costs and switching costs. Under the null hypothesis, the costs of switching should exceed the benefits for the most information-opaque firms. The proxy for the benefits of switching is the set of 11 importance ratings of the characteristics central to the conduct of the owner’s banking business. It is unclear how the importance of each of the individual characteristics should differ for those changing banks. As a set, however, they should be significant if owners change banks because the current bank does not satisfy one or more aspects of service that is important to the owner. Several firm-specific characteristics serve as proxies for hold-up costs that should be associated with the degree of small firm information asymmetry. The first is an adjusted years in business that subtracts the years since the owner last changed banks from the reported years in business in the survey.8 The second is firm size based on the number of full-time equivalent employees (FTE). This measure of size is used instead of gross sales because of the high “no answer” response (above 20 percent) in the survey. Both adjusted firm age and firm size variables are entered in log form in the analysis, i.e. Ln (age-adjusted) and Ln (FTE). Owners also reported their average annual sales growth for the past three years in five categories ranging from ”declined more than five percent” to “grew more than 20 percent.” This variable, Sales growth (which takes value from 1 to 5) is a rough proxy for growth opportunities and should be positively correlated with

8

This process follows the method used by Ongena and Smith (2001). There were several cases where this

adjustment resulted in a negative number, in which case these observations were set equal to 1. The baseline estimates of the model were run with these recoded observations and without them with no effect on the coefficients of the other variables of interest.

12

the degree of information asymmetry. If information opaque borrowers are held-up by their current banks, then the coefficient on Ln (age-adjusted) and Ln (FTE) should be positive and the coefficient on Sales growth should be negative. Other firm characteristics included in the model as control variables are industry classification and form of business. Another set of variables related to hold-up costs and the degree of information asymmetry of small firms is the outcome of the last loan and whether the owner is a nonborrower who only uses transactions services. A 1/0 variable, Denied in last loan try, is created for the outcome of the last loan application and takes a value of 1 if the firm was denied on its last loan application (mean = 11 percent). Firms turned down on their most recent loan are likely to apply to a different bank. These firms must be controlled for to ensure that their experience is not creating an upward bias on the potential association between the incidence of changing banks and any potential hold-up effect. A related 1/0 variable is created for the borrowing status of the owner, where the variable, nonborrower, takes a value of 1 if the firm does not report any recent application for a loan (mean = 22 percent). Owners who only use banks for transactions services should be less likely to face a hold-up situation if the rents banks might extract as result of their investment in private information are related to credit underwriting. The coefficients on both these variables should be positive. Several variables are used as proxies for switching costs. The first is the scope of the banking relation. Firms that use more services may have a greater chance of being locked in, especially if there is an underwriting aspect to providing the service, or if the set up costs are high. The survey includes a set of six transactions services (night

13

depository, cash management, credit card, bill payment, receivables collection, business checking account and international trade financing) and three credit services (seasonal financing, fixed term financing over one year, revolving line of credit).9 There is substantial variation in the usage of these services, ranging from 90 percent of the sample using business checking accounts to only one percent using international trade financing (Table 1). Two variables are created to capture the scope of relationship based on this service usage. The first variable, Ln (transaction services), is the log of the number of transactions services used, excluding business checking because almost all the sample uses this service; the second, Ln (credit services), is the log of the number of credit services used. Three other proxies for switching costs are included in the analysis: the number of banks used, location, and concentration of bank deposits. The number of banks used ranges from one to eight, with an average of 1.5 banks used. This variable, Ln (banks used), is entered in log form in the multivariate analysis. The expectation is that the more banks used reduces switching costs and thus should be positively related to the incidence of changing. The location proxies are Metropolitan Statistical Area and U.S. Census region. Sixty-two percent of the firms are located in a metropolitan statistical area as defined by the U.S. Bureau of the Census. The switching costs should be lower in metropolitan areas where it is easier to find new banks, as well as in more populated U.S. Census regions where there are more banking organizations and branches. The bank concentration variable is based on a Herfindahl-Hirshmann index of deposits and is

9

These classifications are somewhat arbitrary. Credit cards and international trade financing could be

included as credit services. Alternate classifications of these variables had no effect on the conclusions in the paper.

14

computed at the MSA level or county for non-MSA locations (mean= 0.24). In more concentrated markets (those with a higher HHI), there may be less competition and thus a lower incidence of changing. 3.1.2 Empirical results The dependent variable for first hypothesis is Changed banks. The model to be estimated is: Changed banksi = a0 + b1Importance ratings + b2Ln (age) + b3Ln (FTE) + b4Sales growth + b5Denied in last loan try + b6Non-borrower + b7Ln (transactions services) + b8Ln (loan services) +b9Ln (number of banks used) +b10MSA location + b11HHI + b12Region + b13Other control variables + u The proxy for the benefits of switching, Importance ratings, should be positively related to Changed banks. The proxies for information asymmetry should be negatively related, i.e. b2>0, b3>0, b40, b6>0. Increased switching costs should be negatively related to Changed banks, i.e. b70 for market size, b110 in more densely populated regions. The other control variables include industry and form of business. The model is estimated using logistic regression and the results are presented in Table 4. The baseline estimates that include all of the proxies for the benefits and costs of switching are shown in column 1.10 Additional sensitivity tests of the model are presented in columns 2 to 4. None of the individual importance ratings serving as proxies for the benefits of changing banks are significant in column 1. It is possible that the importance ratings are 10

Industry, form of business, and regional control variables are included in the estimates but not reported in

the tables for ease of presentation.

15

significant as a group because of the high correlation among the characteristics (see Table 3). At the bottom of the table, a chi-square test statistic and significance level is presented for the null hypothesis that all the importance characteristics equal zero. The importance characteristics as a group are not significant in explaining the decision to change and thus do not differ between those changing and those not changing. The proxies for asymmetric information do not support the idea that firms are locked into their banking relations. Firm age is negatively related to the incidence of changing, while firm size is not (Table 4, column 1). These results are similar to Ongena and Smith where a negative relationship with age was found, but differ because firm size is not significant in Table 4. In addition, sales growth is positively associated, not negatively, with the incidence of changing banks. Also, the coefficient on Non-borrower is negative, not positive, which means that borrowers are more likely (and non-borrower less likely) to change banks. If hold-up was present, the reverse would be expected information opaque borrowers are bound to their bank because of the bank’s nontransferable private information (Rajan, 1992). To further investigate the effect of firm age and size on the incidence of changing, the coefficients are allowed to vary non-linearly in column 2. In this specification, the coefficient on Ln (age-adjusted) takes on different values for owners in the first quartile (under 8 years), the second and third quartile (8-24 years), and the fourth quartile (over 24 years). Adjusted firm age has no effect on the incidence of changing except for the oldest firms (over 24 years of age) that are less likely to change banks than younger firms. More importantly, the youngest firms are not less likely to change banks. One interpretation of this outcome is that older firms may be more satisfied with their primary

16

bank and have fewer incentives to switch. Younger firms, by contrast, are still searching for the bank that best meets their financial needs as their business model becomes more fully developed. The coefficient on firm size is also allowed to vary non-linearly in column 2 following the procedure used with adjusted firm age. The first quartile of firm size includes firms with 2.5 of fewer FTEs, the second and third quartile includes firms with between 3 and 14 FTEs, and the fourth quartile includes firms with of 15 or more FTEs. The results in column 2 show that only the largest firms are more likely to change banks, although the economic significance is small. For example, an increase of 10 FTEs (e.g. from 15 to 25) results in only a 0.007 increase in the chance of changing banks at the mean of the dependent variable. Similar to youngest firms, the smallest firms are not less likely to change banks. In other words, firm size has no effect on the incidence of changing banks for the smallest firms further supporting the contention that most information opaque firms are not held up by their banks. Another perspective on the strength of any hold-up effect can be seen from the coefficient on Denied on the last loan try, which is positive as expected and economically significant. A firm reporting that it was denied on the last loan application increases the log odds of Changed banks by 32 percent (at the mean of the dependent variable). Thus, the cost of any information capture must be less than the switching costs for these owners. It is possible that firms denied on their most recent loan application are responsible for the association between firm age and the incidence of changing banks, especially if they are also the youngest firms. As can be seen in Table 4, column 3, omission of firms denied in their last loan request has no effect on either the significance

17

or magnitude of the adjusted firm age and firm size coefficients in the baseline model in column 1. Although not shown, the exclusion of these firms also did not affect the conclusions about firm size and age when the coefficients on these variables were allowed to vary non-linearly. The proxies for switching costs have limited significance for small firms, but not in a way that suggests hold-up. Product usage varies positively, not negatively, with the decision to switch banks in column 1 and the association is significant for both the number of loan services and transactions services used. Although the incidence of switching is positively related to the number of banks used, the association is not significant. And finally, market size is positively related to the decision to switch as expected. Firms located in a MSA increase the log odds of changing banks by 0.13 at the mean of the dependent variable. Ongena and Smith found longer relationships associated with larger banks, which they conjecture as a result of the size of the firm’s borrowing needs. A 1/0 variable is included for bank size in column 7 that takes a value of 1 if the owner’s primary bank has greater than $20 billion in assets at the time of the survey. Bank size is found to be negatively related to the incidence of changing banks, which is broadly consistent with Ongena and Smith’s finding. Further tests (not shown) reveal that the negative effect is only significant for larger firms, whose borrowing needs which may exceed the regulatory limit for small banks in the U.S.11 While this association may suggest a holdup effect for the largest firms, competition among large banks in the U.S. is quite fierce

11

Banks are generally prohibited from lending more than 15 percent of their capital to any one borrower.

The data do not permit a direct test of whether the incidence of changing banks is caused by owners outgrowing their primary bank’s lending capacity.

18

and, as Ongena and Smith note, the lack of switching may just reflect satisfaction with the current level of service. The last sensitivity test is shown in column 4 where the U.S. Census-weighted data are used to estimate the model. The results are generally the same as the baseline model in regards to the signs of the coefficients with a few differences in statistical significance. Sales growth and loan services used, while still positively related, are no longer significant. Even with the weighting, the proxies for the benefits of switching – the importance ratings of services central to the conduct of the owner’s financial business – are still not significant. Overall, the multivariate results in Table 4 provide little support for the null hypothesis that small firms are held-up by their current banks. Neither the youngest firms nor the smallest firms are less likely to change banks, or those with more growth opportunities. In addition, there is no evidence that the scope of relationship locks small firms into their bank or that switching costs are high enough to prevent a bank change. While the finding of no hold-up does not mean that relationship banking always provides favorable outcomes to small firms, it does suggest that an analysis of the net benefits of relationship banking to small firms can be limited to whether the private information accumulation results in better banking outcomes for small firms.12 3.2. Reasons for changing banks 3.2.1 Empirical strategy The null hypothesis tested in this section is that the costs or benefits of changing are unrelated to the reason for changing. The framework for testing this hypothesis is the 12

Soft information accumulation may not always benefit small firms. For example, a loan officer that

learns of a serious illness of an owner without any succession plan may be reluctant to extend more credit.

19

same as that in Section 3.1.1, i.e., the incidence of changing (or probability of changing) banks is a function of its benefits and costs. The dependent variable takes five values: Relationship change, Loan term change, Merger change, Other change, and Did not change. Multinomial logit is used to estimate this model because the dependent variable is an unordered list of outcomes. By permitting the dependent variable to vary by the reason for changing, sharper predictions can be made for some of the independent variables, especially for a Relationship change versus a Loan term change. For example, information asymmetry may matter more for owners that are concerned about their banking relationship than for those owners concerned about their loan terms. Thus, smaller, younger firms may be more likely to report a Relationship change, while larger, older firms may be more likely to report a Loan term change if their primary lender has a smaller information advantage. There should also be differences in the importance characteristic ratings that serve as a proxy for the benefits of changing. Owners reporting Relationship change should rate characteristics related to soft information production such as knows my business, provides helpful suggestions, easy access to loan officer, and social contact with loan officer higher than those reporting a Loan term change. More transactions-related characteristics such as wants cheapest money, location important and wide range of services should be rated higher by owners reporting a Loan term change. And finally, switching costs may differ by reason. For example, owners reporting a Loan term change may be less sensitive to distance as described in DeGryse and Ongena (2005). For these firms, the lower switching costs associated with more densely

20

populated markets (banks or people) would be less important resulting in no association with a MSA location. 3.2.2 Empirical results The multinomial logit estimates are presented in Table 5. The baseline model is presented in Panel A; the non-linear firm age and size effects in Panel B; and the baseline model excluding firms denied on their most recent loan attempt in Panel C.13 Each of the possible outcomes for the dependent variable is shown in the columns, with the exception of the omitted category, Did not change. The significance of coefficients should be interpreted relative to Did not change. The proxy for the benefits for switching, the importance ratings, now show some significant differences between reasons for switching in Panel A. As shown at the bottom of Table 5, Panel A, the null hypothesis of equivalence of these coefficients between Relationship change and the other three reasons is rejected. Owners reporting a Relationship change are more likely to rate knows my business, provide and social contact with a loan officer higher; and wants cheapest money and wide range of services lower compared to those that did not change. Only one of the importance characteristics, speed of decisions, is significant Loan term change, but it is not significantly different between reasons. The firm characteristics serving as proxies for information opacity show some differences by reason for the change. Neither adjusted firm age nor firms size are related owners reporting a Relationship change, but sales growth still shows a positive association. If these owners are the more information opaque firms that rely on banks

13

Estimates with the U.S. Census-weighted data were also made, but not presented, because the results were virtually the same as in Panel A – an outcome similar to Table 4, column 5.

21

using soft information to meet their credit needs, then the lack of significance reinforces the conclusions from Table 4 that these owners are no likely held-up by their banks. Firm age is negatively related to owners reporting a Loan term change and Other change. Firm size continues to have no association with the incidence of changing regardless of the reason reported. Firm age and size are permitted to have a non-linear effect in Table 5, Panel B. Adjusted firm age and firm size still have no effect on those reporting a Relationship change. However, the oldest firms (fourth quartile) are least likely and the largest firms (fourth quartile) are more likely to report a Loan term change. The association of firm size with Loan term change may reflect the size of the credit line needed, but this information not available in the survey. More importantly, the youngest firms that do want better loan terms are no less likely to change banks compared to older firms. These results, combined with those in Table 4, again reaffirm the conclusion that more informationally opaque firms do not appear to be held-up by their primary banks. Several other differences between owners reporting a Loan term change and Relationship change are shown in Table 5, Panel A. The incidence of changing increases with the number of loan services only for firms reporting a Loan term change and Merger change; but increases with both loan and transactions services for owners reporting a Relationship change. However, the coefficient on Ln (loan services) is significantly higher for owners reporting a Loan term change versus a Relationship change. In addition, location in an MSA is not significant for a Loan term change as expected, which suggests that this element of switching costs is less important for these owners compared to those reporting a Relationship change.

22

Similar to Table 4, those owners denied in their last loan attempt were significantly more likely to change regardless of the reason reported, although it is not significant for Merger change (Table 5, Panel A). A linear restriction test of equality of magnitude of the coefficient across reasons cannot be rejected. The effect of excluding the firms denied in their last loan attempt has virtually no effect on the baseline coefficient estimates (Table 5, Panel C). While not shown, multinomial logit estimates that include non-linear firm and age effects for the sample in Panel C did not change any of the conclusions based on the Table 5, Panel A coefficient estimates. Owners reporting a Merger change (Table 5, Panel A, column 3) show no significant association with firm size, age, growth, or borrowing status. The only significant variable is location in a metropolitan area and the number of banks used. Thus, owners motivated to switch banks because of problems associated with a merger are randomly distributed with respect to firm characteristics, with no adverse effect on the most information opaque firms. Overall the results reported in Table 5 reject the null hypothesis that the costs and benefits of changing are unrelated to the reason for changing. Owners who value soft information production more highly and those located in MSAs are more likely to report changing for relationship reasons. Firms that change for loan term reasons put more value on transaction-related characteristics important to the conduct of their business and are less likely to be older firms. Finally, the results reinforce the conclusion from Table 4 that scope and switching costs appear to have a limited effect for small firms.

23

3.3. Distance to the primary bank 3.3.1 Empirical strategy The null hypothesis examined in this section is that distance to the owner’s primary bank is unrelated to the decision to change banks. The natural log of time to the bank, Ln (time), is the dependent variable, which is related to Changed banks and other predictors discussed below. This variable is different from the year the relationship started used by Petersen and Rajan (2002) and Ongena and Smith. The survey only asks the last time the principal bank was changed, not how long the owner has been with their principal bank with all changes since 1997 are grouped into one category. Thus, the coefficient will reflect the average change in distance over the four year period (1997 to 2001). If the trend documented by Petersen and Rajan (2002) has continued, then the coefficient on Changed banks should be positive, rejecting the null hypothesis. If distance matters to small firms, it should be because the degree of information opacity creates a need to be closer to their relationship lenders. Thus, firm age and size should be positively related to distance, while sales growth should be negatively related. Empirically it may be difficult to isolate the firm age effect given the association between changed banks and firm age shown in Table 4. Industry classification (with retail firms as the omitted category) is also included as a control variable. Petersen and Rajan (2002) found significant industry effects and they are expected here as well. Similar to Petersen and Rajan, control variables are also included for the presence of a business or personal credit card and checking account. They found that distance was negatively related to the presence of a checking account but no association with a credit card. Because the sample includes borrowers and non-

24

borrowers (a broader sample than Petersen and Rajan, 2002, who limited their analysis to borrowing firms only), a 1/0 variable for borrowing status is enclosed that takes a value of 1 if the owner is a non-borrower. A set of variables for market density and spatial distribution of banks is included. Firms located in metropolitan areas would presumably be closer to their bank because of the increased density of banks; thus the coefficient on this variable should be negative as was found by Petersen and Rajan.14 In addition, a Herfindahl-Hirshmann index of deposit concentration is also included as a measure of market density. Higher levels of the HHI should be associated with fewer banks in a market area. The coefficient should be positive where bank density is lower, and presumably distance between banks is greater. Finally, a 1/0 variable for current bank size is included that takes a value of 1 if the owner currently banks at a CFI (community financial institution). Community banks may be unique in their ability to produce soft information and their organizational architecture may create a comparative advantage in serving small firms (Berger and Udell, 2002, Brickley, Linck and Smith, 2003). If their advantage is based on proximity to the borrower, then the coefficient on the CFI variable should be negative (Berger et al, 2005). 3.3.2 Empirical results Ordinary least squares regression is used to estimate this model and the results are shown in Table 6. Owners that change banks are located about 17 percent further away based on time reported to get to their primary bank, a result that rejects the null 14

The significant positive correlation of MSA with the incidence of changing banks, like that with firm

age, may make it difficult to empirically capture this effect.

25

hypothesis and corroborates the findings of Petersen and Rajan for an earlier period. In column 2 a new dependent variable, Year the Relation Started, is used that is similar to the Petersen and Rajan dependent variable. Despite the truncation of this variable at six years prior to the survey, the increase in distance per year, 4.6 percent, is within the range found by Petersen and Rajan. Although not shown, these results (column 1 or 2) are virtually unchanged when the U.S. Census weighted data are used. The baseline results in column 1 as show that firms with CFIs as their primary bank are located closer than those at large financial institutions (“LFI”) as expected. This result is consistent with the idea that small firms choose to locate closer to CFIs because of their comparative advantage in the delivery of banking services (e.g. Berger et al, 2005). Deposit concentration is positively associated with distance: owners in less concentrated markets with more banks are located closer to their primary bank. The significance of firm characteristics is mixed. Neither years in business nor firm size has an affect on distance. Strong industry effects are present with only FIRE and non-professional services not significantly different from retail firms. The presence of a credit card has no effect on distance, but firms with checking accounts (91 percent of the sample reports having a business checking account) are located closer – findings similar to Petersen and Rajan. Finally, there is no difference in distance between nonborrowers and borrowers. Several other specifications for Changed banks are shown in columns 3 to 6 in Table 6. In column 3 the coefficient on Changed banks is allowed to vary by the reason for the change. Firms reporting a Relationship change may locate relatively closer to their new bank than those seeking better loan terms, especially if the production of soft

26

information requires increased closeness. Although the coefficient on Loan term change interaction variable is greater than the coefficient for the Relationship change interaction variable, the difference is not significant. Additional insight into the relation between distance and firm age is shown in column 4, where the coefficient on Changed banks is allowed to vary by three age categories based on the quartiles of the adjusted years in business distribution. “Young” takes a value of 1 if years in business are within the first quartile (under 8 years in business), “mid-age” takes a value of 1 if years in business falls within the second or third quartile (8-24 years in business), and “old” takes a value of 1 if years in business falls within the fourth quartile (over 24 years in business). The results show that older firms are moving further away to their new bank, but younger firms are no further from their new bank than firms that did not change. This difference could reflect the lower cost of monitoring older firms, which reduces the information cost advantage of a local lender (Inderst and Mueller, 2006). In column 5, Changed banks take a different coefficient based on the status of the most recent loan outcome. Owners turned down on their last loan are located significantly further from their new bank compared to those not turned down. This difference is economically significant as well, with firms turned down locating 38 percent further away versus 14 percent for those not turned down. Not only are these owners more likely to change banks (see Table 4), but they also locate further away when they do. The coefficient on the Changed banks is allowed to vary by size of the new bank (CFI or LFI) in column 6. Although owners at CFIs, on average, are located closer to

27

their bank, at the margin (i.e. when they change) do they also locate closer? Surprisingly, owners changing to CFIs are located further from their new bank than if they changed to LFIs.15 Even after considering the standalone effect having a CFI as the owners primary bank, the combined effect (0.243 - 0.151) is still positive. This result might be explained by the owner switching to a de novo bank located further away than the current bank. Another explanation may be that owners are willing to incur the (small) additional travel costs to derive the benefits from conducting their financial business at community banks. Perhaps technology has reduced some of the need of physical closeness for both the borrower and the lender. Whatever the cause, the data suggest that the benefits received by both the owner and the bank exceed the costs of further distance to the CFI. In summary, proximity to their lender does not appear to be a concern of the most information opaque small firms that have recently changed banks. They are further away from their primary bank than those that have not changed, although the youngest firms (less than five years in business), are no further away than those that did not change. The outcome of the most recent loan had a big effect on distance to the new lender, with those turned down locating two and one-half times further away. And finally, while owners at CFIs are closer to their bank, those changing to CFIs are locating further away. 4. Summary and Conclusions This paper empirically examines the extent to which information opacity hinders the decision of a small firm to change banks. Although Ongena and Smith (2001) provide evidence that hold-up is not experienced by a sample of small Norwegian firms, no evidence have been directly provided for U.S. small firms. A firm that changes banks 15

Alternative breakpoints were considered to define a CFI (e.g. $500 million) and had no effect on the

results shown in column 6.

28

also makes a decision about the location of the new bank that should be included in its assessment of the switching costs. Even though Petersen and Rajan (2002) concluded that distance appears to matter less to small firms, it could still matter more to small firms with severe asymmetric information problems that need to be closer to a relationship lender. This paper uses recent survey data of small firms to further analyze why these firms change banks. A unique feature of the survey is that it asks the respondents to report why they changed their principal financial institutions. The reasons are sorted into four categories: relationship, loan term, merger and other. In addition, the survey asks owners to rank a set of characteristics for their importance in conducting the firm’s banking business. These characteristics serve as proxies for the potential benefits of changing banks. The principal findings of the paper are as follows. First, no evidence is found to support hold-up, confirming the findings of Ongena and Smith (2001) for small Norwegian firms. The youngest and smallest firms are no less likely to change banks, while the oldest firms are less likely to change banks. Although the largest firms are more likely to change banks, the economic significance is small. Next, owners changing for relationship reasons are found to place a higher value on the soft-information aspects of banking (such as how well their bank knows their business, access to and social contact with their loan officer) compared to those changing for loan term reasons. Those changing for loan term reasons are likely to be larger, more intensive users of loan services, and less likely to be constrained to banks in their immediate market area.

29

And finally, owners changing banks are more likely to be further away from their bank – about 17 percent in terms of time - regardless of the reason given for the change. This result confirms the earlier findings of Petersen and Rajan (2002) although the units and base of reference are not exactly the same. This aspect of switching costs is less important to older firms that move further to their new bank than younger firms. In addition, firms changing to CFIs tend to locate further away than those at non-CFIs, suggesting that the costs of being further away are outweighed by the benefits of conducting their financial business at these banks. Several related questions remain unanswered. From a public policy perspective, it would be helpful to know how many of these firms switched to new banks that were new (do novo) charters. With a better identification of the old versus new bank, a better analysis can be made of how bank size affects the decision to change bank. The identification of the old bank could also help strengthen the conclusions about distance and the new bank chosen, especially the linkage between distance, the degree of information asymmetry, market structure, and the reason for changing banks. These issues remain for future research.

30

References Akhavein, J., Frame, W.S., and White, L.J., 2005.“The Diffusion of Financial Innovations: An Examination of the Adoption of Small Business Credit Scoring by Large Banking Organizations.” Journal of Business, 78, 577–96. Angilini, P., Di Salvo, R., and Ferri, G., 1998. Availability and cost of credit for small businesses: Customer relationships and credit cooperatives. Journal of Banking and Finance 22, 925-954. Berger, A.N., Miller, N.H., Petersen, M.A., Rajan, R.G., Stein, J.C., 2005. Does function follow organizational form? Evidence from the lending practices of large and small banks. Journal of Financial Economics 76,237-269. Berger, A. N., and Udell, G.F., 1995. relationship lending and lines of credit in small firm finance. Journal of Business 68, 351-382. Berger, A. N., and Udell, G.F., 2002. Small business credit availability and relationship lending: the importance of bank organisational structure. Economic Journal 112, F32F53. Berger, A.N., Frame, S., Miller, N., 2005. Credit scoring and the availability, price and risk of small business credit. Journal of Money, Credit, and Banking 37, 191–222. Boot, A.W.A. 2000. Relationship banking: what do we know? Journal of Financial Intermediation 9, 7-25. Brickley, J.A., Linck, J.S., Smith, C.W. “Boundaries of the firm: Evidence from the banking industry.” Journal of Financial Economics 70 (2003), 351-383. Cole, R.A., 1998. The importance of relationships to the availability of credit. Journal of Banking and Finance 22, 959-977. Cole, R.A., Goldberg, L.G., White, L.J., 2004. Cookie-cutter versus character: The micro structure of small business lending by large and small banks. Journal of Financial and Quantitative Analysis 39, 227-251. Degryse, H., and Ongena, S., 2005. Distance, lending relationships, and competition. Journal of Finance 60, 231-266. Degryse, H., and Van Cayseele, P., 2000. Relationship lending within a bank-based system: Evidence from European small business data, Journal of Financial Intermediation 9, 90-109. DeYoung, R., Hunter, W.C., Udell, G.F., 2004. The past, present, and possible future for community banks? Journal of Financial Services Research 25, 85-133.

31

Elsas, R., 2005. Empirical determinants of relationship lending. Journal of Financial Intermediation 14, 32-57. Greenbaum, S. I., Kanatas, G., and Venezia, I., 1989. Equilibrium loan pricing under the bank-client relationship. Journal of Banking and Finance 13, 221-235. Harhoff, D., and Korting, T., 1998. Lending relationships in Germany – Empirical evidence from survey data. Journal of Banking and Finance 22, 1317-1353. Inderst, R. and Mueller, H.M., 2006. A lender-based theory of collateral. Unpublished manuscript. Ongena, S. and Smith, D., 2001. The duration of bank relationships. Journal of Financial Economics 61, 449-475. Petersen, M.A. and Rajan, R.G., 1994. The benefits of firm-creditor relationships: Evidence from small business data. Journal of Finance 49, 3-37. Petersen, M.A. and R. G. Rajan, 2002. Does distance still matter? The information revolution in small business lending. Journal of Finance 57, 2533-2570. Rajan, R.G., 1992. Insiders and outsiders: the choice between informed and arm’s-length debt. Journal of Finance 47, 1367-1399. Sharpe, S.A., 1990. Asymmetric information, bank lending, and implicit contracts: a stylized model of customer relationships. Journal of Finance 45, 1069-1087.

32

Appendix Unweighted and Weighted Distribution of Survey Respondent Characteristics Frequency distributions are presented of selected firm characteristics from the 2001 National Federation of Independent Business (NFIB) Credit, Banks and Small Business Survey. Raw survey responses are used in the column labeled "Unweighted." In the column labeled "Weighted," the 1997 U.S. Bureau of the Census Enterprise Survey data are used to weight the raw survey data by employment, 1-digit SIC, and Census regions. UnUnweighted Weighted Years in Business weighted Weighted Form of Business Proprietorship 26 32 0-4 11 12 Partnership 8 8 5-9 14 16 Corporation 40 35 10-14 16 17 S-Corporation 24 23 15-19 14 14 No answer 2 2 20-24 13 11 100 100 25 or more 28 26 Full Time No answer 5 5 Equivalent 100 100 One 7 9 Industry 2-4 29 38 Construction/mining 15 10 5-9 25 27 Manufacturing 11 6 10-19 17 11 Transportation 4 3 20-49 12 7 Wholesale 10 10 50-99 4 2 Retail 20 24 100-499 2 2 FIRE 6 6 No answer 5 4 Business services 17 23 100 100 Professional services 8 11 Agriculture 5 3 Gross sales ($000) Under 25 3 3 No answer 4 4 25-49 1 2 100 100 50-99 4 5 Region 100-249 12 16 Northeast 15 18 250-499 14 15 Midwest 39 30 500-999 15 15 South 24 31 1,000-2,499 14 10 West 22 21 2,500-4,999 7 5 No answer * * 5,000-9,999 4 3 100 100 10,000 or more 6 5 Sales Growth (past 3 years) No answer 21 21 Decline 15 16 100 100 No change (-5 to 5%) 29 27 Urban Location Grew 6-10% 28 27 Yes 62 62 Grew 11-20% 12 12 No 38 38 Grew > 20% 9 9 No answer * * Too new to tell 2 3 100 100 No answer 6 7 100 100 * less than 1 percent

Table 1 Variable Definitions Definitions and summary statistics are computed from the 2001 Credit, Banks and Small Business survey conducted by the National Federation of Independent Business. No answer responses are excluded in the computation of the summary statistics. Variable description Max Min Mean Std Dev Variable names Changed banks 1 if owner reported ever changing their principal financial institution and 0 otherwise 1 0 0.57 0.49 Relationship change 1 if owner reports changing banks and gives a relationship explanation as the reason 0 1 0.28 0.45 Loan term change 1 if owner reports changing banks and gives a loan term explanation as the reason 0 1 0.25 0.43 Merger change 1 if owner reports changing banks and gives merger as the reason 0 1 0.18 0.38 Other change 1 if owner reports changing banks and gives another reason or no answer 0 1 0.30 0.46 Firm characteristics Importance ratings Knows my business 1 if 'Not Important' to 5 'Very Important' 1 5 4.45 0.89 Provides helpful suggestions 1 if 'Not Important' to 5 'Very Important' 1 5 3.33 1.21 Wants cheapest money 1 if 'Not Important' to 5 'Very Important' 1 5 4.15 1.05 Location important 1 if 'Not Important' to 5 'Very Important' 1 5 4.25 0.90 Reliable credit source 1 if 'Not Important' to 5 'Very Important' 1 5 4.36 0.95 Knows my industry 1 if 'Not Important' to 5 'Very Important' 1 5 3.65 1.17 Speed of decisions 1 if 'Not Important' to 5 'Very Important' 1 5 4.30 0.84 Easy access to loan officer 1 if 'Not Important' to 5 'Very Important' 1 5 4.28 0.94 Wide range of services 1 if 'Not Important' to 5 'Very Important' 1 5 3.78 0.99 Knows local market 1 if 'Not Important' to 5 'Very Important' 1 5 3.84 1.08 Social contact with loan officer 1 if 'Not Important' to 5 'Very Important' 1 5 2.93 1.37 Denied in last loan try 1 if 'Yes' and 0 otherwise for those that applied for a loan 1 0 0.11 0.31 Non-borrower 1 if 'Yes' and 0 otherwise 1 0 0.22 0.41 Number of banks used Number of bank and non-bank financial institutions used 1 8 1.57 0.89 Sales growth 1 if declined more than 5%; 2 if no change (-5% to 5%); 3 if 6% -10%; 4 if 11% to 20%; and 5 if > 20% 1 5 2.77 1.26 Age (adjusted) Reported years in business less the years since the firm last changed banks 1 99 18.1 14.0 FTE Full-time equivalent employees 0.5 989 16.7 48.9 Industry Agriculture 1 if 'Yes' and 0 otherwise 0 1 0.07 0.26 Manufacturing 1 if 'Yes' and 0 otherwise 0 1 0.12 0.32 Construction 1 if 'Yes' and 0 otherwise 0 1 0.15 0.36 Transportation 1 if 'Yes' and 0 otherwise 0 1 0.04 0.19 Wholesale 1 if 'Yes' and 0 otherwise 0 1 0.10 0.30 Retail 1 if 'Yes' and 0 otherwise 0 1 0.20 0.40 FIRE 1 if 'Yes' and 0 otherwise 0 1 0.06 0.24 Services 1 if 'Yes' and 0 otherwise 0 1 0.15 0.36 Professional 1 if 'Yes' and 0 otherwise 0 1 0.07 0.25

Table 1 (continued) Variable names Form of business Proprietorship Partnership (incl. LLC) S-corporation Corporation Product/service usage Seasonal financing Fixed term financing (>1 year) Revolving line of credit No. of credit services Night depository Business credit card Cash management/sweep account Bill payment services Business checking account Receivables collection service International trade financing No. of transaction services Market/location MSA HHI Northeast South Midwest Plains West Distance to bank CFI Very large bank

Variable description

Max

Min

Mean Std Dev

1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise

0 0 0 0

1 1 1 1

0.26 0.08 0.24 0.40

0.44 0.27 0.43 0.49

1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise Sum of all credit services reported 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise 1 if 'Yes' and 0 otherwise Sum of all transactions services reported

0 0 0 0 0 0 0 0 0 0 0 0

1 1 1 3 1 1 1 1 1 1 1 6

0.23 0.43 0.48 1.13 0.23 0.37 0.19 0.11 0.90 0.02 0.01 1.85

0.42 0.49 0.50 1.04 0.42 0.48 0.40 0.32 0.30 0.14 0.12 1.02

1 if the firm is located in a Metropolitan Statistical Area Herfindahl-Hirshmann index of deposit concentration for the MSA where the business is located, or county if in a non-MSA 1 if the firm is located in New England or Mid-Atlantic Census regions 1 if the firm is located in South Atlantic or East South Central Census regions 1 if the firm is located in East North Central Census region 1 if the firm is located in West North or West South Central Census regions 1 if the firm is located in Pacific or Mountain Census regions Distance to bank in minutes to primary financial institution 1 if the current bank size is $1billion or less 1 if the current bank size exceeds $20 billion

0

1

0.62

0.49

0 0 0 0 0 0 1 0 0

1 1 1 1 1 1 90 1 1

0.24 0.14 0.24 0.39 0.22 0.22 9.52 0.47 0.27

0.14 0.36 0.43 0.49 0.42 0.41 9.39 0.50 0.44

Table 2 Last Time Owners Changed Their Principal Financial Institution Panel A presents the responses to the question "When was the last time you changed principal financial institutions?" using the 1987, 1995, and 2001 Credit, Banks and Small Business surveys. In Panel B the reasons are given for changing banks in the 2001 survey. Panel C presents the distance to the owner's bank for those that change (by year of change) and those that did not change. a

1995 6 5 6 6 10 64 4 100

1987 6 7 6 6 7 65 3 100

2001 23 16 9 5 6 6 18 15 2 100

2000 30 9 9 7 2 9 10 19 5 100

1999 13 7 11 7 7 10 17 28 1 100

1997 or 1998 earlier 28 14 7 7 8 6 3 6 13 7 6 14 9 13 25 27 2 6 100 100

Total Observations Percent of total

82 7

106 9

112 9

120 10

768 64

1,188 100

C. Distance to bank (minutes) Changed banks Did not change banksb

9.1

10.0

9.9

9.7

10.2

10.0 8.8

Total Observations Percent of total

82 7

106 9

112 9

120 10

768 64

1,188 100

A. Time Since Last Change $20 billion -0.480 0.098 *** Regional effects Yes Yes Yes Yes Yes Industry effects Yes Yes Yes Yes Yes Form of business effects Yes Yes Yes Yes Yes Constant -0.759 0.524 -0.773 0.438 * -0.797 0.506 -0.716 0.515 -0.798 0.710 Test statistic of null hypothesis: all importance rating coefficients = 0 (Prob > χ2) No. of observations pseudo-Rsq

2,068 0.053

13.25 (.278)

12.31 (.305) 2,068 0.057

13.21 (.280) 2,068 0.050

12.89 (.301) 2,068 0.060

* indicates significance at the 10 percent level; ** indicates significance at the 5 percent level; and *** indicates significance at the 1 percent level.

12.79 (.308) 2,068 0.063

Table 5 Determinants of Which Firms Change Banks Estimates of the determinants of the reasons for changing banks are presented using data from the 2001 Credit, Banks and Small Business survey. Panel A presents the baseline results; Panel B allows for a non-linear years in business and firm size response; and Panel C excludes owners reporting a denial on their most recent loan application. Multinomial logit is used estimate the effect of the independent variables on the different reasons given for changing banks (shown at the top of the columns): Relationship change (column 1), Loan Term Change (column 2), Merger Change (column 3), and Other change (column 4). The omitted category from the estimation is firms that did not change banks. The significance of the coefficients should be interpreted as the effect on the log odds of changing for the reason reported versus not changing banks. Robust standard errors are reported that allow for clustering on firm size. (1) (2) (3) (4) Relationship Loan term change change Merger change Other change A. Baseline model Importance characteristics Coeff. Std.Err. Coeff. Std.Err. Coeff. Std.Err. Coeff. Std.Err. Knows my business 0.262 0.083 *** -0.019 0.085 -0.094 0.110 0.048 0.107 Provides helpful suggestions 0.036 0.080 -0.070 0.064 -0.018 0.070 -0.026 0.066 Wants cheapest money -0.120 0.060 ** 0.072 0.064 -0.060 0.085 0.003 0.060 Location important -0.048 0.097 -0.097 0.091 0.065 0.074 0.093 0.064 Reliable credit source -0.047 0.099 -0.020 0.102 0.120 0.096 -0.097 0.060 Knows my industry 0.000 0.067 0.055 0.060 -0.089 0.082 0.012 0.049 Speed of decisions 0.130 0.094 0.245 0.121 ** 0.152 0.115 0.089 0.081 Easy access to loan officer 0.003 0.085 -0.045 0.158 0.024 0.128 -0.116 0.070 * Wide range of services -0.171 0.078 ** 0.093 0.112 -0.116 0.108 0.048 0.063 Knows local market -0.043 0.071 -0.053 0.075 0.068 0.079 -0.040 0.075 Social contact with loan officer 0.138 0.065 ** -0.073 0.046 -0.012 0.067 -0.132 0.053 ** Firm characteristics Ln (Age-adjusted) -0.044 0.085 -0.218 0.103 ** 0.089 0.129 -0.166 0.088 * Ln (FTE) 0.096 0.076 0.082 0.081 0.078 0.090 0.033 0.086 ** Sales growth 0.091 0.054 * 0.123 0.054 -0.038 0.060 0.065 0.057 *** Denied in last loan attempt 0.572 0.231 ** 0.812 0.254 0.350 0.287 0.383 0.222 * * *** Non-borrower -0.315 0.171 -1.024 0.325 -0.280 0.210 -0.360 0.158 ** Product/service usage Ln (transaction services) 0.331 0.168 ** 0.102 0.142 0.132 0.183 0.224 0.148 *** ** Ln (loan services) 0.292 0.169 * 0.761 0.193 0.358 0.163 -0.195 0.133 Market/location characteristics Ln (number of banks used) 0.145 0.263 0.400 0.285 0.553 0.273 ** 0.164 0.271 *** MSA location 0.444 0.153 0.211 0.154 0.703 0.173 *** 0.044 0.125 HHI -0.204 0.569 0.034 0.499 0.725 0.535 -0.216 0.521 Primary bank is >$20 billion -0.462 0.137 *** -0.272 0.154 * -0.560 0.197 *** -0.570 0.144 *** Regional effects Yes Yes Yes Yes Industry effects Yes Yes Yes Yes Form of business effects Yes Yes Yes Yes ** *** Constant -3.004 0.729 *** -2.400 0.959 -3.797 0.768 -0.760 0.619 Test statistic of equivalence of all importance characteristic coefficients by reason (Prob > χ2) Relationship change=Loan term change Relationship change=Merger change Relationship change = Other change

44.5 29.8 69.7

(.000) (.002) (.000)

Table 5 Determinants of Which Firms change Banks (continued) (1) B. Non-linear Age and Size Effects Importance characteristics Knows my business Provides helpful suggestions Wants cheapest money Location important Reliable credit source Knows my industry Speed of decisions Easy access to loan officer Wide range of services Knows local market Social contact with loan officer Firm characteristics Year in business (1st quartile): 0-5 Year in business (2nd and 3rd quartile): 6 Year in business (4th quartile): 23 and hig FTE (1st quartile): $20 billion Regional effects Industry effects Form of business effects Constant

Relationship change Coeff. Std.Err. 0.261 0.083 0.037 0.081 -0.115 0.059 -0.054 0.099 -0.047 0.100 -0.008 0.068 0.126 0.092 -0.005 0.083 -0.162 0.079 -0.035 0.071 0.144 0.065 0.000 0.011 -0.003 -0.026 -0.020 0.003 0.100 0.593 -0.309

0.038 0.011 0.005 0.067 0.016 0.002 0.052 0.231 0.172

0.336 0.294

0.169 0.171

0.163 0.447 -0.217 -0.460 Yes Yes Yes -2.936

0.262 0.151 0.568 0.138

0.709

(2)

***

*

**

**

* ** *

** *

***

***

***

Loan term change Coeff. Std.Err. -0.022 0.084 -0.062 0.065 0.077 0.064 -0.095 0.092 -0.020 0.103 0.050 0.061 0.235 0.121 -0.052 0.157 0.100 0.114 -0.049 0.074 -0.065 0.046 -0.057 -0.011 -0.015 -0.021 0.018 0.005 0.131 0.856 -1.000

0.058 0.017 0.008 0.066 0.022 0.002 0.052 0.253 0.323

0.102 0.720

0.142 0.191

0.368 0.196 0.046 -0.282 Yes Yes Yes -2.600

0.278 0.151 0.509 0.155

0.874

*

*

** ** *** ***

***

*

***

(3)

(4)

Merger change Coeff. Std.Err. -0.095 0.110 -0.020 0.070 -0.056 0.087 0.066 0.074 0.121 0.097 -0.094 0.084 0.145 0.114 0.023 0.127 -0.111 0.110 0.069 0.079 -0.005 0.066

Other change Coeff. Std.Err. 0.046 0.108 -0.023 0.065 0.004 0.060 0.093 0.064 -0.102 0.060 0.010 0.051 0.077 0.081 -0.122 0.071 0.052 0.063 -0.034 0.075 -0.127 0.053

0.006 0.010 0.001 -0.059 -0.004 0.004 -0.042 0.355 -0.274

0.041 0.014 0.007 0.121 0.025 0.002 0.059 0.286 0.211

0.119 0.341

0.190 0.159

0.554 0.693 0.727 -0.566 Yes Yes Yes -3.076

0.284 0.175 0.532 0.200

*

0.713

***

*

**

***

***

-0.025 -0.008 -0.016 0.081 0.017 0.005 0.062 0.398 -0.358

0.047 0.016 0.008 0.071 0.022 0.002 0.055 0.215 0.154

0.225 -0.227

0.146 0.136

0.140 0.021 -0.225 -0.568 Yes Yes Yes -0.923

0.273 0.123 0.515 0.146

0.612

*

*

**

**

**

**

*

***

Table 5 Determinants of Which Firms change Banks (continued) (1) C. Without Loan Denials Importance characteristics Knows my business Provides helpful suggestions Wants cheapest money Location important Reliable credit source Knows my industry Speed of decisions Easy access to loan officer Wide range of services Knows local market Social contact with loan officer Firm characteristics Ln (Age-adjusted) Ln (FTE) Sales growth Denied in last loan attempt Non-borrower Ln (transaction services) Ln (loan services) Market/location characteristics Ln (number of banks used) Number of banks used MSA location HHI Primary bank is >$20 billion Regional effects Industry effects Form of business effects Constant

Relationship change Coeff. Std.Err. 0.230 0.088 0.022 0.074 -0.132 0.062 -0.044 0.102 -0.045 0.109 0.004 0.073 0.129 0.092 0.064 0.093 -0.200 0.067 -0.037 0.066 0.140 0.067 -0.020 0.104 0.120

0.097 0.075 0.053

-0.344

0.175

0.366 0.187

0.175 0.172

0.163 0.525 -0.206 -0.486 Yes Yes Yes

0.255 0.165 0.570 0.154

-3.260

0.787

(2)

***

**

***

**

Loan term change Coeff. Std.Err. 0.035 0.091 -0.080 0.077 0.097 0.070 -0.118 0.088 0.000 0.099 0.061 0.062 0.266 0.136 -0.082 0.166 0.110 0.120 -0.076 0.074 -0.100 0.054

**

*

-0.252 0.078 0.123

0.110 0.085 0.062

**

**

-1.063

0.332

***

**

0.104 0.690

0.162 0.203

0.473 0.214 0.150 -0.245 Yes Yes Yes

0.300 0.154 0.540 0.154

-2.461

1.112

**

***

***

***

**

***

**

(3)

(4)

Merger change Coeff. Std.Err. -0.094 0.112 -0.031 0.076 -0.030 0.084 0.061 0.082 0.129 0.108 -0.107 0.089 0.166 0.126 -0.005 0.129 -0.145 0.109 0.093 0.085 -0.001 0.068

Other change Coeff. Std.Err. 0.042 0.104 -0.037 0.068 0.025 0.064 0.113 0.068 -0.125 0.060 0.020 0.052 0.079 0.084 -0.119 0.073 0.059 0.069 -0.009 0.077 -0.127 0.059

*

**

0.055 0.104 -0.040

0.133 0.090 0.071

-0.208 0.023 0.050

0.093 0.087 0.066

**

-0.317

0.210

-0.386

0.155

**

0.168 0.311

0.192 0.176

0.246 -0.188

0.149 0.147

*

0.547 0.826 0.539 -0.534 Yes Yes Yes

0.281 0.172 0.554 0.205

*

0.079 0.025 -0.204 -0.567 Yes Yes Yes

0.286 0.135 0.510 0.139

-3.941

0.795

***

-0.677

0.617

*

***

***


***

Table 6 Determinants of Distance to Current Primary Bank OLS estimates of the determinants of the log of the distance (in minutes) are presented using data from the 2001 Credit, Banks and Small Business survey. The baseline results are in column 1. An alternate definition of theChanged banks based on the Year the Relation Started is shown in column 2. The coefficient onChanged banks varies by the outcome of the most recent loan application in column 3. In column 4, theChanged banks coefficient varies by three years in business categories based on quartile distributions: "Young" takes a value of 1 if firm age 24 years). In column 5 the coefficient on Changed banks varies by the reason for the change. Changed banks varies by the size of the current bank (CFI versus LFI) in column 6. F-tests for the equivalence of selected interactive variable coefficients in columns 3 - 6 are presented at the bottom of the table. (1) (2) (3) (4) (5) (6)

Independent variables Changed banks Year relation started

Robust Coeff. Std. Err. *** 0.167 0.037

Robust Coeff. Std. Err. 0.046

0.012

Robust Coeff. Std. Err.




***

Changed banks x Relationship change Changed banks x Loan term change Changed banks x Merger change Changed banks x Other change

0.183 0.208 0.161 -0.001

0.048 0.048 0.055 0.065

*** *** ***

Changed banks x Young Changed banks x Midage Changed banks x Old

0.117 0.168 0.178

0.093 0.053 0.047

*** *** ***

Changed banks x Denied Changed banks x Not denied

0.383 0.140

0.061 0.039

*** ***

Changed banks x CFI Changed banks x LFI Market/location CFI HHI-MSA MSA location Firm characteristics Ln (Age-adjusted) Ln (FTE) Has credit card Has checking account Non-borrower

-0.080 0.341 0.041

0.032 0.165 0.045

-0.003 -0.002 -0.004 -0.101 -0.014

0.022 0.017 0.042 0.056 0.040

** **

*

-0.076 0.360 0.047

0.032 0.162 0.045

-0.002 0.000 0.000 -0.103 -0.026

0.022 0.018 0.041 0.054 0.039

** **

*

-0.075 0.339 0.039

0.031 0.162 0.045

-0.004 -0.006 -0.003 -0.108 -0.006

0.022 0.017 0.040 0.055 0.038

** **

*

-0.081 0.338 0.040

0.032 0.168 0.045

-0.008 -0.002 -0.005 -0.102 -0.013

0.031 0.017 0.042 0.056 0.041

** **

*

-0.073 0.327 0.037

0.033 0.165 0.046

0.003 0.001 -0.001 -0.105 0.004

0.023 0.018 0.041 0.056 0.041

** **

*

0.243 0.132

0.055 0.044

***

-0.151 0.343 0.036

0.046 0.165 0.047

***

-0.005 0.000 -0.002 -0.101 -0.011

0.022 0.017 0.042 0.055 0.040

***

**

*

Table 6: Determinants of Distance to Current Primary Bank (continued) (1) (2)

Independent variables Agriculture Construction Manufacturing Transportation Wholesale FIRE Services Professional services Constant Pseudo r-squared F-tests on Changed banks Relationship change vs. Loan term change Young vs. Old Denied vs. Not Denied CFI vs. LFI

Robust Coeff. Std. Err. *** 0.748 0.075 0.423 0.053 *** *** 0.483 0.071 0.667 0.106 *** 0.335 0.047 *** -0.069 0.060 *** 0.299 0.051 0.082 0.079 *** 1.558 0.104 0.101

Robust Coeff. Std. Err. 0.763 0.077 *** 0.433 0.053 *** *** 0.492 0.071 0.669 0.106 *** 0.343 0.048 *** -0.063 0.060 *** 0.301 0.051 *** 0.089 0.080 1.538 0.100 *** 0.097

(3)

(4)

Robust Coeff. Std. Err. 0.744 0.076 *** 0.418 0.052 *** *** 0.479 0.069 0.647 0.108 *** 0.333 0.046 *** -0.066 0.060 0.297 0.051 *** 0.082 0.080 1.595 0.101 *** 0.103 F 0.24

(5)

Robust Coeff. Std. Err. 0.746 0.075 *** 0.424 0.052 *** *** 0.485 0.071 0.671 0.106 *** 0.335 0.046 *** -0.067 0.062 0.298 0.051 *** 0.082 0.080 1.570 0.117 *** 0.101

Sign 0.624

F 0.31


(6)


Sign 0.581


F

Sign

13.95

0.000

F

Sign

3.02

0.084