Deterioration in borrowing terms of small businesses: an agency perspective.

Hedges, Peggy ; Wu, Zhenyu ; Chua, Jess 等

Introduction

Small business owners frequently refer to banks as "fair weather" lenders, i.e., that bank debt is available only when the business is doing well and banks may recall their loans or apply more stringent borrowing terms when the economic situation deteriorates and the borrower's needs are acute. When this happens, the financial crisis faced by the small business is likely to be much worse than that of not obtaining debt financing in the first place because the small business will have made irreversible financial and strategic commitments plus incurred irrecoverable sunk costs. Therefore, understanding how to decrease this possibility should be of great importance to both owners of small business and small business management researchers. This dynamic aspect of debt financing for small business has not been examined before.

Recent theories of debt financing have focused on the agency problems between lenders and borrowers arising from asymmetric information. While the risk of default retains a role in determining access to and the terms of borrowing, both researchers and practitioners have observed that debt financing is more than just a matter of default risk. For example, default risk cannot explain the practice of capital rationing wherein debt financing from banks is unavailable even when the small business is willing to pay higher interest. Neither can it explain why borrowers tend to become captives of their lenders and find it difficult to get better terms from competing banks. This and many of the complex problems in lending have been explained by agency theory and empirical studies of both large and small firms have shown that dealing with the agency problems between lender and borrower is key to accessing debt financing.

The effective means for dealing with agency problems may be different, however, for large and small businesses. For large firms, the means for controlling the adverse selection and moral hazard problems include signalling, bonding, and monitoring (Harris and Raviv, 1991; Shleifer and Vishny, 1997). Unfortunately, the information asymmetry problems for small businesses are more serious than those in large firms. Consequently, these tools do not appear to be as effective for small firms. Instead, the small business has relied on developing relationships with lenders to deal with agency problems (Petersen and Rajan, 1994; Berger and Udell, 1995); relationship has the ability to decrease the information asymmetry between the lender and borrower with respect to both the quality of the business and the business owner's opportunism.

This study investigates whether dealing with agency problems affects the likelihood that the borrowing terms of a small business will deteriorate. It makes two important contributions to the literature on small business management. First, using Industry Canada data for 2001, a time soon after the economic slowdown in 2000, this study provides evidence that dealing with agency problems does decrease the likelihood that borrowing terms for a small business will deteriorate. This means that small businesses should nurture and continually improve such ability. Second, we confirm the US evidence regarding the importance of relationship in dealing with the agency problems of small business debt financing.

Hypotheses

Agency theory began as a rigorous explanation of the idea that non-owner managers will not watch over the affairs of a firm as diligently as owners would if the latter were managing the firm themselves. Ross (1973) formalized this conflict of interest arising from the separation of ownership and management as a principal-agent problem and Jensen and Meckling (1976) coined the phrase "agency costs" to represent the costs of all activities and operating systems designed to align the interests and/or actions of managers (agents) with the interests of owners (principals). Expanding on the concepts, Myers (1977) and Smith and Warner (1979) showed that agency problems also exist in the owner-lender relationship.

Over the years, theoretical developments pinpointed information asymmetry, due to complexity, observability, information search costs, and bounded rationality, as the sources of agency problems. In the context of the borrower-lender relationship, asymmetric information leads to adverse selection when the lender systematically prices the terms of borrowing lower than the risk justifies (Akerlof, 1970). It also leads to a wide range of moral hazard problems wherein the firm changes the riskiness of the debt after the debt has been extended (Smith and Warner, 1979).

Full disclosure and ensuring observability, if possible and not value dissipating, are the lowest cost means of dealing with information asymmetry. Unfortunately, this is not possible in most cases because intangible risk sources such as the borrower's ability, future commitment, and opportunism plus the business' future potential cannot be accurately and credibly communicated. Thus, costly signals have to be employed (Harris and Raviv, 1991) with the cost making the signal credible. For a small business, signals to the lender include willingness to put personal wealth at risk and certification by a third party. It is quite possible that signals continue to be effective even when economic conditions deteriorate and, thus, lower the probability that credit terms for the small business will be made more stringent. Therefore, we have the following Hypothesis 1:

Hypothesis 1: The stronger the signal provided to the lender by the borrower, the less likely credit terms for the borrower will change adversely.

While signalling deals with the asymmetric information problem before the loan is obtained, it does not deal with the opportunism of the borrower after the lender has turned over the funds. To deal with these moral hazard problems, the literature suggests monitoring and bonding (Smith and Warner, 1979). Bonding is more popular because its costs are mostly borne by the borrower. Monitoring not only has higher costs, exacerbated by the loss of economies of scale in the case of small business, or may even be impossible, as in the case of determining the commitment of the owner's creativity. Bonding in the small business context would include providing collateral, personal guarantees, or both. We hypothesize that when the small business has provided adequate collateral, personal guarantees, or both, the bank is less likely to change the firm's borrowing terms negatively. Hence, we have Hypothesis 2:

Hypothesis 2: The stronger the bonding provided by the small business to the lender, the less likely the credit terms will change adversely.

As explained by Peterson and Rajan (1994) and Berger and Udell (1995), tools recommended for dealing with the debt-related agency problems of large firms are either unavailable or less effective in the case of SME debt financing. Instead, small businesses have to rely on long-term relationship to cope with the asymmetric information problems. The efficacy of long-term relationship as an agency cost control mechanism in SME debt financing has been studied both theoretically (Petersen and Rajan, 1994; Boot and Thakor, 1994) and empirically (Peterson and Rajan, 1994; Berger and Udell, 1995). Essentially, a long-term relationship reduces the information asymmetry between the lender and borrower, thus minimizing both adverse selection and moral hazard problems. Hence, we have Hypothesis 3:

Hypothesis 3: The stronger the relationship between the borrower and the bank, the less likely the credit terms will change adversely.

In summary, the agency theory of debt financing predicts that signalling, bonding, and the relationship between lender and borrower are able to resolve the agency problems between lenders and borrowers by minimizing the effects of information asymmetry. We hypothesize that they not only have a static function, but also a dynamic one in the sense that they affect the probability that the borrowing terms of a small business will deteriorate when economic conditions weaken.

Methodology

This study examines small business debt financing focusing on whether arrangements for reducing agency problems between lenders and borrowers affect the probability that the borrowing terms of the small business will deteriorate. As discussed previously, deterioration in the borrowing terms of a small business could create a financial crisis worse than that of not being able to borrow in the first place because the irreversible financial commitments and irrecoverable sunk costs made after debt financing is secured could make the business unsustainable.

Data

The data used come from a survey commissioned by Industry Canada in 2001. (1) The purpose of the survey was to understand better the financial needs of Canadian small and medium size enterprises (SME). The survey provided an overview of Canadian SME financing over the preceding three years and focussed on the financing that had occurred in the six months prior to the survey along with the financing intentions for the upcoming year.

A total of 10,000 questionnaires were mailed to eligible Canadian firms (2) of which 1,014 surveys were completed and returned. Given the low response rate, the survey was modified in order to facilitate telephone interviews. A total of 1,134 telephone surveys were completed resulting in a total of 2,148 responses. (3) Thirty-two firms were dropped from the sample as they did not meet the minimum eligibility requirements; therefore 2,116 observations were left.

Since telephone surveys were conducted after the mail surveys, we follow Oppenheim (1966) in assuming that these late respondents are more similar to non-respondents. Comparison of these two sub-samples did not reveal statistically significant differences in terms of the major variables; therefore, this suggests that there is no reason to suspect nonresponse bias in the key variables analyzed in this study.

There are, however, other important data limitations. The data are self-reported; no independent verification is possible. Many of the questions asked for subjective opinions. Our review of the dataset reveals that many variables necessary to fully and directly test agency and relationship theories are missing or incomplete in nature. For example, there are no variables related to monitoring activities. That is why we did not propose a hypothesis with respect to monitoring.

The Research Institute for SMEs, housed at the University of Quebec, found a high rate of "anomalous" values for several variables, including the ratio of employees to total sales, indebtedness ratio, and financial risk measures. A number of these errors are attributed to faulty data capture and others due to misunderstanding on the part of respondents. It is not possible for us to ascertain the existence and extent of any bias introduced by these values.

Meaningful inclusion and analysis of financial factors are not possible because many of the respondents did not answer the questions related to details about their firms. The financial data for which there are a larger number of responses, on the other hand, yield imprecise information. For example, liabilities are not separated into financial institution debt and trade payables. Nevertheless, recognizing the flaws in the financial data, the combined dataset is sufficient to examine the effects of relationships, signalling and bonding on changes to credit terms. Given the dearth of Canadian data, this particular set provides the opportunity to take a unique first step in examining the important question about credit term deterioration and the Canadian situation.

Dependent Variable

The survey covered different types of short-term borrowing, long-term borrowing, and equity. We focused on the line of credit (LOC) following Berger and Udell (1995). Berger and Udell (1995) note that the first formal lending relationship between the bank and a business is typically a LOC. They argue that LOCs are ideal for studying bank-borrower relationships as banks specialize in lending to highly information-problematic classes of borrowers. Nitani et al. (2004) also find that the LOC is one of the significant form of financing for new technology firms. Therefore, LOCs provide a reasonably good barometer of the effect of relationships to changes in terms and the intent to continue the banking relationship.

For this study, LOCs are also ideal because their terms can be changed quickly in response to the financial results of the borrower or economic conditions. On the other hand, because LOCs are typically reviewed annually, some of the firms in the sample may not have had their cases reviewed resulting in higher seeming stability than might actually exist. However, LOCs and their reporting requirements are typically stringently monitored and any deterioration in firm performance may result in an interim review, increased monitoring, or both.

The dependent variable is CHANGE, a measure of deterioration in borrowing terms. It is measured by the responses to the survey question, "During the past six months, have the financing conditions on your business line of credit been modified?" and the follow-up questions about how the conditions had been modified.

Terms of LOCs included interest rates, service fees, and credit limit. If a respondent indicated adverse change to all three terms or adverse change to one or more terms with the others unchanged, we deemed the credit terms to have deteriorated and the binary dependent variable (Change) was assigned a value of one. If all three terms had improved or one or more had improved with the remaining ones unchanged, the binary variable (Change) was assigned a value of zero. It was not possible to ascertain whether borrowing terms deteriorated if there was improvement in one but adverse change in another. These few cases were excluded.

For a number of firms, there were no changes to any one of their credit terms. For robustness of our results, we assigned different values to these respondents' CHANGE variable. In one set of tests, we grouped them with the positive change group and assigned to them a value of zero. In the alternative set, we excluded them. We found no qualitative differences in the results and report those for the first set here.

If the deterioration in credit terms was drastic, the small business may have failed or changed its lender. The former would introduce a survivorship bias but the latter would not because the firms would still be in the sample. We did not correct for the possible survivorship bias.

Independent Variable--Signalling

Four variables were used to measure signalling by the borrower. The first variable (Personal) was measured by whether or not the owner used a personal line of credit to help finance the business during the past three years. Although risky, particularly for the small business without a strong track record, the willingness of the owner to put personal wealth at risk alongside human capital is a strong signal of the owner's commitment to and confidence in the business.

The second variable (Reporting) is a dummy variable indicating whether or not a person other than the owner is in charge of finance and/or accounting matters. This individual, although employed by the enterprise, presumably is a reasonable "third party" to ensure that the information about the enterprise is relayed to the lender in a timely and accurate manner. The survey did not distinguish between an independent third party or related individual (i.e., family member). If an independent third party, the lenders would be provided with some assurances the information they are depending upon in their credit analysis processes is likely to be less biased than when the owner prepares the information. Researchers (e.g., Shleifer and Vishny, 1997) have suggested that having a third party in charge of financial and accounting matters signals the importance the firm attaches to such matters. They also note that the existence of a full-time financial manager indicates actual competence and signals the professionalization in such matters to the lenders (Thornton, 1981). (4)

The other two signalling variables are Founder and Manage. Founder indicates whether the borrower participated in creating the business. Such a borrower may be more committed to the survival of the firm although we must recognize that in some cases the founder may be more of an entrepreneur rather than a manager and hence employment of a professional manager may be desirable as the firm grows. The data set did not speak to whether or not the company employed both the founder and a professional manager. Finally, if the owner also manages the business (Manage), behaviour toward risk may also be different. Mishra (1999) believes that owner-managers tend to be more risk averse.

Independent Variable--Bonding

Bonding is typically achieved through both collateral and guarantee (Stulz and Johnson, 1985). The survey grouped these together. The ideal measure of bonding would be a comparison of the value of the collaterals plus guarantees to the size of the LOC; unfortunately, this was not available from the survey. Instead, the independent variable (Bonding) was measured by the level of borrowers' satisfaction with guarantees required by the institution. This measure is imprecise because satisfaction is determined by the difference between the expectation of the borrower and the requirement set by the lender. Therefore, even if the requirement is low, the borrower will be dissatisfied if the expectation is even lower. Nevertheless, there is likely a high correlation between the bonding required and satisfaction; the higher the bonding requirement, the less likely that the borrower will be satisfied.

Independent Variables--Relationship

Three dimensions of lender-borrower relationship were included in the analysis. Reliability measures whether the financial institution supported the firm when times were hard; Flexibility measures the willingness of the bank to negotiate credit terms; and Amturnover measures the turnover in account manager. All three variables were measured on a 5-point scale.

Several dimensions of banking relationships were not captured by the above construct. For example, there was no information about the length of time the LOC had been in place, a measure used by previous researchers to measure relationship. This information was not available. Relationship may be affected by the type of lending institution. Therefore, we included a dummy variable (Bank) to classify the lending institutions into two types: commercial banks (given a value of one) and others (given a value of zero).

Control Variables--Firm

Other firm-specific variables, aside from those related to agency problems, could also affect credit terms. Many of these, such as size, age, industrial sector, and stage of development, may be interpreted as proxies of default risk.

Size has been found to be an important variable in determining the credit rating of firms (Kaplan and Urwitz, 1979). In addition, research consistently shows that size affects the amount of debt capital available to the firm (e.g., Hooks, 2003). We measured size by the natural logarithm of the number of employees (Lnemployee) because there were too many missing observations for total assets. For analyzing potential strata problems, we also categorize the firms into five groups by size, according to the categories defined by Statistics Canada. These are firms with 0, 1-4, 5-19, 20-99, and 100-499 employees.

Diamond (1991) and Petersen and Rajan (1994, 1995) found the age of the firm to be positively related to leverage; older firms are more likely to have higher debt ratios than younger firms. The natural logarithm of the age of the firm (Lnagefirm) was therefore included in the model.

Consistent with Berger and Udell's (1995) approach, categorical variables were adopted to measure industry characteristics. Following Statistics Canada, firms were categorized into seven groups: agriculture, knowledge-based industries, manufacturing, primary, professional services, wholesale and retail, and other sectors. Six dummies were included in the analysis.

Stage of development is considered crucial for small and medium-sized enterprises. It has four categories: seed and start-up, growth, maturity, and winding down, and is represented by three dummy variables (Startup, Growth and Maturity).

The legal form of the business may be important in terms of legal separation between personal and firm assets and also as an indicator of business sophistication. Adummy variable, Sole, which measures if a firm is a sole proprietorship, was included in the analysis.

Control Variables--Borrower

The personal attributes of the borrower, aside from firm characteristics, could also contribute to the risk exposure of the lender and, thus, may affect both the credit terms. Prior research (e.g., Coleman, 2002; Fielden et al., 2003) has shown that female entrepreneurs (Gender, male [1] or female [0]) and visible minorities (Minority, yes [1] or no [0]) generally find it difficult to obtain capital. (5) Language was also included as a characteristic because there might be cultural differences between francophones, anglophones, and others in terms of their dealings with financial institutions. Two dummy variables differentiated mother tongue into English, French, or other.

Education is a categorical measure of borrowers' human capital (e.g., Coleman, 2002), and we measured this with three dummy variables indicating whether the borrower received high-school or post-secondary education. Years of experience in the industry (Indexp) as a predictor of success is a risk measure. Age of the owner (Ageowner) could also be a proxy for risk. (6)

Analysis

We examined deterioration in a borrower's credit term in terms of the probability that it would occur. However, there was a potential sample selection problem since firms that had not used LOC as a type of debt financing had missing data in the variable Change. Thus, we had to use a two-stage probit model to deal with it by calculating Inverse Mills' Ratio (IMR) from the first stage.

The correlation test also shows that relationship variables are highly and significantly related to other agency variables (e.g., Bonding, Personal) and some control variables (e.g., Bank, Lnagefirm). Therefore, we adopted another two models as robustness tests. In Model 2 and Model 3, for analytical simplicity, we constructed a new relationship variable, Relationavg, whose values were the average scores of the three dimensions of the lender-borrower relationship. This should be considered a continuous variable. In Model 2, we re-estimated Model 1 using Relationavg instead of the three relationship variables, while in Model 3, we used the residual of the variable Relationavg after taking the influences of other variables on the relationship variable into consideration. We also conducted a series of robustness tests for further exploring the effects of collinearity. These are discussed in a later section.

Results

Table 1 provides the descriptive statistics for Change, the firm, and the owner. As shown in the table, the average value for Change was positive at 0.53 indicating that slightly more than half of the sample firms suffered deterioration in borrowing terms during the period. This had to be expected following the economic slowdown in 2000. The average firm had 22.4 employees, was 15 years old, and more than half (0.56) were in professional, wholesale, and retail. More than 83% of the respondents were founders and more than 88% were managed by the owner. Industry experience of the owners averaged 20.5 years. To save space, we do not repeat the results from the descriptive analysis here.

Hypothesis 1. Hypothesis 1 proposes a negative relationship between signals provided to the lender by the borrower and the adverse change in credit terms. This hypothesis is mainly rejected since only the variable Founder is weakly significant (at the 10% level) and negatively related to the deterioration in the borrowing terms of small firms in Model 1 and Model 2. After dealing with the multicollinearity in Model 3, however, it is not significant any more. Results are presented in Table 2.

Personal did not have a significant relationship with adverse change in credit terms. There are two possible explanations. One is a strong relationship allows the bank and the Canadian small firm to diminish the information asymmetry that leads to agency problems. Another is firms of this size may have to rely on both the firm's debt capacity and that of the owner to finance their expansion. Therefore, using personal capital may be standard practice for Canadian small firms and cannot explain any difference in credit terms. Around 70% of the firms in our sample did not use personal capital.

Reporting was not significantly related to adverse change in credit terms either. A possible explanation is that the individual in charge of finance and accounting matters may have a special relationship with the firm owner, and therefore the signal may not be credible. Unfortunately, the data do not allow us to tell if the third party was independent of the firm owner.

Hypothesis 2. Hypothesis 2 proposes a negative relationship between the borrowers' satisfaction with bonding requirements and the adverse change in credit terms. This hypothesis was supported after taking the multicollinearity into consideration in Model 3.

In Models 1 and 2, in which no interaction between relationship variables and other variables are considered, the influence of the bonding variable is not significant due to the multicollinearity problem. The results from Model 3 presented in Table 2, however, indicate that the borrowers' satisfaction with guarantees required by lenders is negatively related to the adverse change in credit terms and significant at the 1% level. It may be standard practice for owners of small firms to give personal guarantees. Further, borrowers may perceive that by offering guarantees, the likelihood the credit terms for their business will deteriorate is lessened, despite prevailing economic conditions.

Hypothesis 3. Hypothesis 3 proposes a negative relationship between the borrower-lender relationship and the adverse change in credit terms. The results presented in Table 2, which show the lender-borrower relationship, as measured by two of the three raw variables, the average of the three dimensions, and the residual after taking multicollinearity into account, is always highly significant (at the 1% level) and negatively related to the adverse change in credit terms. Thus, the result is very robust, suggesting that the stronger the relationship, the less likely credit terms are changed adversely.

Institution variable Bank does not have a significant relationship with change in credit terms. This rejects the frequent complaints by small Canadian firms about their relationships with banks. The results do suggest relationships can solve many of the agency problems arising from information asymmetry. This also reconfirms the findings by Peterson and Rajan (1994) and Berger and Udell (1995) in that relationship lending is critical in small business debt financing from the borrower's perspective using Canadian data.

Control Variables--Firm

Of the firm variables, Sole was positively related to adverse change in credit terms and significantly at the 5% level in all three models. This suggests that during an economic recession period, sole proprietorships are more likely to be treated poorly by lenders since other legal forms of organization may have competitive advantages over them.

The variable Professional Services was negatively related to deterioration in credit terms in Model 3 in which the multicollinearity problem was dealt with. The survey was done in late 2000 when prices of commodities in general, but not those of professional services, were declining. This could explain why lenders did not tighten credit terms for firms in this particular sector of the economy. In general, additionally, part of the size variables are negatively related to the adverse change in credit terms.

Control Variables--Borrower

Only two of the borrower characteristics, Gender and Minority, were significantly related to adverse change in credit terms before and after dealing with the multicollinearity problem. Their significant relationships with an adverse change in the credit terms in Models 1 and 2 suggest that male-owned and non-minority-owned small firms are less likely to suffer deterioration in credit terms, indicating possible gender and racial effects. Their insignificant relationships after dealing with the multicollinearity problem indicate, however, that laws and policies to eliminate discrimination in lending have been successful. In addition, firm owners with better education are less likely to get adverse change in credit terms.

Sample Selection Bias. As mentioned earlier, results from all three models presented in Table 2 suggest strong sample selection bias since the Inverse Mill's Ratios are all significant, although at different levels in different models. Due to the missing data from firms not choosing LOCs as a debt financing source, it was necessary to correct for the sample selection bias before making conclusions. Thus, a two-staged probit model was employed.

Stratified Sample. (7) The sample can be stratified through both sector and size and, as a result, a potential bias may have been generated. Table 3 highlights the proportion of firms experiencing adverse change in credit terms in each sub-category determined by sector and size. Results indicate that 15.3% of the firms with more than 99 but less than 500 employees experienced adverse change in the terms of LOC's as did 14.6% of the firms in the primary sectors. To further test if these results were biased by the strata, the above models were rerun with six dummies for sectors and four for size. (8) No qualitative change was found.

Collinearity. By checking the correlation matrix, we identified the independent and control variables that were highly correlated and were potential sources of multicollinearity problems. We ran base case regressions with all variables and checked the variance inflation factors (VIF). The highly correlated variables were removed, one by one, making sure that a sufficient number of variables remained to account for the different groups of economic drivers, and reran the regressions. This was repeated until no VIF exceeded 10.0.

However, results from both the two-stage probit models including many non-significant variables and the correlation coefficients among variables suggest that collinearity problems may extend beyond the interactions between relationship variables, other agency variables and some control variables. To further explore the potential influences of collinearity, we followed the standard solution proposed in Campbell and Reyment (1978) and Campbell (1980) and adopted Principal Component Analysis (PCA). Twenty-eight components with eigenvalue values between 0.061 and 2.608 were extracted from the PCA. As Campbell observed, the principal component with the smallest eigenvalue is the one involving collinear variables, and the standard rules about eigenvalue values should not be used to decide the variables kept as the smallest eigenvalue points to the most significant eigenvector. These are also confirmed by Hadi and Ling (1998).

A logistic model was then run with the principal components that had the highest and the lowest eigenvalue values (9) to get a good fit. The coefficient on principal component 1 is -0.637, and it is significant at the 1% level. The coefficient on principal component 28 is 0.034, but it is insignificant. While introducing factor scores introduces an error and understates the significance levels, a good explanatory power helps make it acceptable (Madansky, 1959).

Conclusions

Our analysis attempts to extend the small business lending literature in three distinct areas. First, we examine the banking relationships of small Canadian firms, specifically with respect to the determinants of adverse changes in credit terms which could create a worse financial crisis for the small business than that faced by the firm from not getting the loan in the first place. Second, we support results obtained by U.S. researchers showing agency theory is a fruitful approach to understanding the borrower-lender relationship, and we also examine the relationship from the unique perspective of the borrower. Third, we examine the drivers of negative changes in lending terms using an agency perspective, and propose potential solutions to agency conflicts generated by the information asymmetry between lenders and small business borrowers.

Our results suggest that building a strong long-term relationship with the bank is one of the best solutions to agency problems for Canadian small firms in debt financing. This is in agreement with the results on U.S. borrowers. The guarantee required by lenders also appears to have a significant effect in determining whether the small Canadian firm's credit terms are changed adversely. Although building a strong relationship can be difficult for both the lender and the borrower due to high account manager turnover, both parties need to examine how the relationship can be strengthened to enhance mutual trust. Government programs dealing with small business issues could include information on how to build relationships with lenders as it is one of the most effective ways to mitigate informational asymmetry.

We find that small business borrowers who have strong relationships with their lenders enjoy a lower likelihood that their credit terms will be changed adversely. Bonding also helps. In conclusion, our results suggest a strong relationship between small business and the bank has the potential to solve the agency problems associated with debt. Theoretically, this study contributes further support to the use of agency theory in explaining the borrower-lender relationship.

References

Akerlof, G. 1970. "The Market for Lemons: Quality Uncertainty and the Market Mechanism," Quarterly Journal of Economics 84: 488-500.

Ang, J.S., J.H. Chua and R. Sellers. 1979. "The Profiles of Late-paying Consumer Loan Borrowers: An Exploratory Study," Journal of Money, Credit and Banking (May).

Berger, A. and G. Udell. 1995. "Relationships Lending and Lines of Credit in Small Firm Finance," Journal of Business 68: 351-382.

--. 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: 613-73.

--. 2002. "Small Business Credit Availability and Relationship Lending: The Importance of Bank Organisational Structure," The Economic Journal 112: F32-F53.

Billett, M., M. Flannery and J. Garfinkel. 1995. "The Effect of Lender Identity on a Borrowing Firm's Equity Return," Journal of Finance 50: 699-718.

Boot, A.W.A. and A.V. Thakor. 1994. "Moral Hazard and Secured Lending in an Infinitely Repeated Credit Market Game," International Economic Review 35: 899-920.

Campbell, N.A. 1980. "Shrunken Estimators in Discriminant and Canonical Variate Analysis," Applied Statistics 29: 5-14.

Campbell, N.A. and R.A. Reyment. 1978. "Discriminant Analysis of a Cretaceous Foraminifer Using Shrunken Estimators," Mathematical Geology 10: 347-59.

Cole, R. 1998. "The Importance of Relationships to the Availability of Credit," Journal of Banking and Finance 22: 959-77.

Coleman, S. 2002. "Constraints Faced by Women Small Business Owners: Evidence from the Data," Journal of Developmental Entrepreneurship 7, no. 2: 151-74.

Davidsson, P. 1989. "Entrepreneurship--and After? A Study of Growth Willingness in Small Firms," Journal of Business Venturing 4: 211-26.

Diamond, D. 1991. "Monitoring and Reputation: The Choice Between Bank Loans and Directly Placed Debt," Journal of Political Economy 99: 689-721.

Dyer, W., Jr. 1986. Cultural Change in Family Firms: Anticipating and Managing Business and Family Traditions. San Francisco: Jossey Bass.

Fielden, S.L., M.J. Davidson, A.J. Dawe and P.J. Makem. 2003. "Factors Inhibiting the Growth of Female Owned Small Businesses in North West England," Journal of Small Business and Enterprise Development 10: 152-66.

Hadi, A.S. and R.F. Ling. 1998. "Some Cautionary Notes on the Use of Principle Components Regression," The American Statistician 52: 15-19.

Harris, M. and A. Raviv. 1991. "The Theory of Capital Structure," The Journal of Finance 46: 305-60.

Hooks, L. 2003. The Impact of Firm Size on Bank Debt Use," Journal of Financial Economics 12: 173-89.

Jensen, M. and W. Meckling. 1976. "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure," Journal of Financial Economics 3: 305-60.

Kaplan, R. and G. Urwitz. 1979. "Statistical Models of Bond Ratings: A Methodologyical Inquiry," Journal of Business 52: 231-61.

Madansky, A. 1959. "The Fitting of Straight Lines When Both Variables are Subject to Error," Journal of the American Statistical Association 54: 173-205.

Mishra, C.S. 1999. "Founding Family Control and Capital Structure: The Risk of Loss of Control and the Aversion to Debt," Entrepreneurship Theory and Practice 23: 53-64.

Myers, S. 1977. "The Determinants of Borrowing," Journal of Financial Economics 5: 147-75.

Natani, M., G.H. Haines, J.J. Madill, Jr., B.J. Orser and A.L. Riding. 2004. "Financing New Technology Firms: Testing for Markets Gaps." Pp. 1-18 in L. Murray Gillin et al. (eds)., Regional Frontiers of Entrepreneurship Research 2004: Proceedings from the First Annual Regional Entrepreneurship Research Exchange. Hawthorne, Victoria, Australia: The Australian Graduate School of Entrepreneurship, Swinburne University of Technology.

Oppenheim, A.N. 1966. Questionnaire Design and Attitude Measurement. London: Heinemann.

Petersen, M.A. and R.G. Rajan. 1994. "The Benefits of Lending Relationships: Evidence from Small Business Data," Journal of Finance 49: 3-37.

--. 1995. "The Effect of Credit Market Competition on Lending Relationships," Quarterly Journal of Economics 110: 407-43.

Ross, S. 1973. "The Economic Theory of Agency: The Principal's Problem," American Economic Review 63: 134-39.

Shleifer, A. and R.W. Vishny. 1997. "A Survey of Corporate Governance," Journal of Finance 52: 737-83.

Slovin, M., M. Sushka and J. Poloncheck. 1993. "The Value of Bank Durability: Borrowers as Bank Stakeholders," Journal of Finance 48: 247-66.

Smith, C.W., Jr. and J.B. Warner. 1979. "On Financial Contracting: An Analysis of Bond Covenants," Journal of Financial Economics 7: 117.

Stiglitz, J. and A. Weiss. 1981. "Credit Rationing in Markets with Imperfect Information," American Economic Review 71: 373-410.

Stulz, R.M. and H. Johnson. 1985. "An Analysis of Secured Debt," Journal of Financial Economics 14: 501-21.

Thomas, D.R. 1993. "Inference Using Complex Data from Surveys and Experiments," Canadian Psychology 34: 413-31.

Thornton, D.B. 1981. Small Business Financing and Non-Bank Financial Institutions, Volume I: Text, Chapter 6. Toronto: FACSYM Research Limited.

Vos, E. and C. Forlong. 1996. "The Agency Advantage of Debt Over the Lifecycle of the Firm," Journal of Entrepreneurial and Small Business Finance 5, no. 3: 193-211.

(1.) Industry Canada directed The Research Institute for SMEs, housed at the University of Quebec, to complete the survey.

(2.) The firm could not be a franchise, holding company or not-for-profit agency. Further, the number of employees of the firm had to be lower than 500 and firm had to have sales in excess of $30,000.

(3.) SMEs completing the mail survey did not participate in the telephone surveys.

(4.) The authors thank one of the anonymous referees for this observation.

(5.) This does not necessarily imply discrimination in lending because gender and race may be highly correlated with other economic drivers.

(6.) Prior research (Ang, Chua and Sellers, 1979) suggests age is a nonlinear predictor of default. Therefore, we also included a square term for this variable (Age-Squared), but found no qualitative change in the results.

(7.) This part is added along the lines suggested by one of the anonymous referees. The authors thank the referee for pointing out this issue.

(8.) This method is suggested by Thomas (1993).

(9.) We also ran a robustness test by including the principle components with the first two highest eigenvalue values and with the lowest eigenvalue value. No qualitative change was found on those with the highest and the lowest eigenvalue values, but the component with the second highest eigenvalue value was not significant.

Peggy Hedges, Haskayne School of Business, University of Calgary

Zhenyu Wu, Haskayne School of Business, University of Calgary

Jess Chua, Haskayne School of Business, University of Calgary

Contact Information

For further information on this article, contact

Zhenyu Wu, 2500 University Drive, NW, Calgary, Alberta, Canada T2N 1N4.

Tel.: (403) 220-3185

Fax: (403) 282-0095

Email address: zhenyu.wu@haskayne.ucalgary.ca.

Table 1. Selected Descriptive Statistics

 Std.
 N Mean Deviation

CHANGE 2093 0.053 0.224
RELIABILITY 1261 3.30 1.28
FLEXIBILITY 1252 3.26 1.19
AMTURNOVER 1262 3.60 1.26
BONDING 1291 2.961 1.378
PERSONAL 2078 0.299 0.458
REPORTING 2097 0.576 0.494
FOUNDER 2095 0.836 0.371
MANAGE 2088 0.886 0.318
BANK 2104 0.796 0.403
LNAGEFIRM 2050 2.204 1.08
SOLE 2097 0.190 0.392
LNEMPLOYEE 2021 2.00 1.49
 1-4 2076 0.384 0.486
 5-19 2076 0.292 0.455
 20-99 2076 0.228 0.420
 100-499 2076 0.050 0.218
STARTUP 2072 0.053 0.224
GROWTH 2072 0.715 0.452
MATURITY 2072 0.167 0.373
GENDER 2093 0.780 0.414
MINORITY 1998 0.121 0.326
HIGHSCHO 2083 0.346 0.476
COLLEGE 2083 0.232 0.422
UNIVERSITY 2083 0.375 0.484
FRENCH 2092 0.268 0.443
ENGLISH 2092 0.581 0.493
INDEXP 2072 20.454 11.612
AGRICULTURE 2069 0.049 0.216
KNOWLEDGE-BASED 2069 0.042 0.200
MANUFACTURING 2069 0.124 0.329
PRIMARY 2069 0.020 0.139
PROFESSIONAL SERVICES 2069 0.260 0.438
WHOLESALE and RETAIL 2069 0.305 0.461

Table 2. Logistic Models

 Model 1

 B S.E.

RELIABILITY -0.457 *** 0.152
FLEXIBILITY -0.604 *** 0.167
AMTURNOVER -0.119 0.117
RELATIONAVG
RELATIONAVG Residual
BONDING 0.083 0.122
PERSONAL 0.079 0.288
REPORTING 0.043 0.310
FOUNDER -0.656 * 0.384
MANAGE -0.254 0.408
BANK 0.066 0.415
SOLE 6.327 ** 3.120
STARTUP 0.467 1.038
GROWTH -1.833 1.219
MATURITY -1.044 1.004
GENDER -1.140 * 0.654
MINORITY 1.070 * 0.602
HIGHSCHO -0.992 1.058
COLLEGE -1.909 1.154
UNIVERSI -1.091 1.179
FRENCH 0.841 0.606
ENGLISH -0.495 0.898
Agriculture -0.605 0.971
Knowledge-based Industries 0.145 0.685
Manufacturing -0.568 0.604
Primary -0.240 1.072
Professional Services -0.661 0.538
Wholesale and Retail 0.374 0.419
Size (1-4) -1.372 1.275
Size (5-19) -2.476 * 1.480
Size (20-99) -3.524 ** 1.754
Size (100-499) -3.216 2.097
IMR -8.695 ** 3.640
Constant 15.414 ** 6.714
N 1022 1105
Pseudo R Square 0.220 0.205

 Model 2

 B S.E.

RELIABILITY
FLEXIBILITY
AMTURNOVER
RELATIONAVG -1.075 *** 0.158
RELATIONAVG Residual
BONDING -0.010 0.116
PERSONAL 0.039 0.278
REPORTING -0.051 0.299
FOUNDER -0.662 * 0.370
MANAGE -0.246 0.387
BANK -0.147 0.379
SOLE 6.917 ** 3.036
STARTUP 0.418 1.002
GROWTH -1.900 1.196
MATURITY -1.146 0.987
GENDER -1.259 ** 0.640
MINORITY 1.167 ** 0.576
HIGHSCHO -1.510 0.937
COLLEGE -2.238 ** 1.044
UNIVERSI -1.561 1.063
FRENCH 0.533 0.564
ENGLISH -0.755 0.854
Agriculture -0.771 0.953
Knowledge-based Industries 0.225 0.677
Manufacturing -0.597 0.576
Primary -0.435 1.045
Professional Services -0.793 0.516
Wholesale and Retail 0.326 0.403
Size (1-4) -1.270 1.214
Size (5-19) -2.335 * 1.412
Size (20-99) -3.427 ** 1.683
Size (100-499) -3.288 2.012
IMR -9.148 *** 3.528
Constant 16.604 6.500
N 1071
Pseudo R Square 0.197

 Model 3

 B S.E.

RELIABILITY
FLEXIBILITY
AMTURNOVER
RELATIONAVG
RELATIONAVG Residual -1.003 *** 0.162
BONDING -0.486 *** 0.110
PERSONAL 0.286 0.281
REPORTING -0.134 0.306
FOUNDER -0.581 0.378
MANAGE -0.133 0.398
BANK 0.014 0.380
SOLE 7.120 ** 3.096
STARTUP 0.474 1.010
GROWTH -1.963 1.213
MATURITY -1.265 1.004
GENDER -1.081 0.663
MINORITY 0.883 0.634
HIGHSCHO -1.197 1.041
COLLEGE -2.010 * 1.145
UNIVERSI -1.295 1.161
FRENCH 0.472 0.568
ENGLISH -1.003 0.867
Agriculture -0.735 0.963
Knowledge-based Industries 0.229 0.679
Manufacturing -0.663 0.591
Primary -0.490 1.057
Professional Services -0.906 * 0.532
Wholesale and Retail 0.276 0.417
Size (1-4) -1.242 1.219
Size (5-19) -2.487 * 1.428
Size (20-99) -3.473 ** 1.706
Size (100-499) -3.512 * 2.050
IMR -9.417 *** 3.610
Constant 14.276 ** 6.603
N
Pseudo R Square

Note: * indicates significance at the 10% level, ** indicates
significance at the 5% level, and *** indicates significance
at the 1% level.

Table 3. Strata: Proportion of Adverse Change in Credit Terms

 Sector 1 2 3 4

Size 1 0.000 0.000 0.000
 2 0.081 0.100 0.022 0.091
 3 0.000 0.000 0.015 0.188
 4 0.000 0.043 0.103 0.083
 5 0.000 0.000 0.179 0.500
Total 0.052 0.048 0.072 0.146

 Sector 5 6 7 Total

Size 1 0.043 0.000 0.000 0.011
 2 0.021 0.023 0.046 0.035
 3 0.038 0.074 0.050 0.052
 4 0.049 0.058 0.029 0.058
 5 0.100 0.154 0.235 0.153
Total 0.036 0.047 0.049 0.050

Note: Sector: 1. agriculture, 2. knowledge-based industries,
3. manufacturing, 4. primary, 5. Professional services, 6. wholesale
and retail, and 7.other sectors. Size: 1. 0 employees, 2.
1-4 employees, 3. 5-19 employees, 4. 20-99 employees, and
5. 100-499 employees.