Do small and medium-sized companies benefit from maintaining a closer relationship with their main bank?

(1)

MSc Finance

Track: Corporate Finance

Master Thesis

Do small and medium-sized companies

benefit from maintaining a closer

relationship with their main bank?

Written by: Anastasia Romm

Thesis supervisor: Dr. Razvan Vlahu

July 1

st

_{, 2017}

University of Amsterdam,

Amsterdam Business School

(2)

2

Statement of Originality

This document is written by Student Anastasia Romm who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

3

Abstract

I investigate the benefits of establishing and maintaining a closer relationship with its main bank for small businesses in the U.S.A. Using survey data collected by the Federal Reserve Board in 2003, I show that increase in relationship scope has positive effect for the borrower when it comes to costs of borrowing: firms with fewer bank relationships and located closer to their financing bank obtain loans at lower interest rates. The higher the proportion of that loan in the firm’s credit portfolio, the lower the interest rate. Collateral requirements become less stringent both in relationship scope, measured by number of bank relationships and proportion of the loan in firm’s credit portfolio, and in relationship length. Guarantee requirements are more stringent for firms with multiple bank relationships. Overall, I conclude that it is beneficial, in terms of price and non-price dimensions of loan contracts, for a borrower to maintain a stronger relationship with its main bank. This bank-borrower relationship is a mechanism to reduce information asymmetry problem between “informationally opaque” SMEs and banks that provide them with loans.

(4)

4

1. Introduction

In my study I investigated the effects of borrower-lender relationship on loan contract terms to small and medium-sized enterprises (SMEs). In broad terms, my study falls into the area of research dedicated to relationship banking1. It studies various effects of stronger connection between the

bank and the client such as: availability of credit, loan contract characteristics and usage of covenants, competition in debt market, relationship versus transaction lending, evolution of firm-bank interactions over time.

I am zooming in on the benefits for the SME borrower of establishing and maintaining a closer relationship with its bank, which manifests in terms of loan agreement. I consider three types of benefits a firm can get: possibility to obtain loans at a lower interest rates and loosening of collateral and guarantee requirements. Since SMEs tend to be more “opaque” in a sense that their financial statements are usually less comprehensive than those of public companies and they are not obliged to disclose as much information, greater role in assessing credit worthiness of such companies is attributed to “soft” information. This information a loan officer can gather through direct communication with debtor’s management and owners. Thus, building a closer relationship with a client should decrease information asymmetry, help to evaluate credit risks more precisely and lead to better terms of loans to SMEs. Indeed these predictions were supported by empirical studies such as Berger and Udell (1995) and Bharath et al. (2011).

On the other hand, closer relationship with a bank leads to a so called “hold-up” problem (see Boot (2000)) meaning that it is costly for a client to switch the lender because the new one will not have the same amount of information about the potential borrower. In other words, this “soft” information is not easily transferable (see Rajan (1992)). Consequently, current lender can use this leverage to its own advantage by increasing interest rates or requiring additional collateral for loan renewal or increase in the credit facility. This means that in order to evaluate borrower’s benefits,

1_{For a comprehensive review of existing literature see, for example, Boot (2000) and Dedu and Nițescu}

(7)

7

we need to know the degree of influence the incumbent bank has. This leads me to the main research question: do SMEs benefit from maintaining a closer relationship with their main bank?

It is an interesting topic to research because, despite the abundance of literature dedicated to investigating these implications, there is still no fundamental agreement on the subject. Further investigation will help to answer questions like: is the tight connection between the firm and the bank more beneficial for one or the other, and whether or not such a connection can clash with interests of society at large. For example, the bank can become less objective in determining borrower’s creditworthiness over time due to personal connections between firm’s management and loan officer, which can lead to higher default rates, hurting the real sector of economy and negatively impacting employment.

I was able to add to existing research in several ways. First of all, in order to answer the research question I used survey data on U.S. small businesses collected by the Federal Reserve Board. Some previous studies such as Berger and Black (2011), Cole et al. (2004) and others also used this survey but for previous interviewing rounds (1988-89, 1998, 1993). Moreover, I used a new measure of relationship strength – a share of the most recent loan in the whole credit portfolio of the client. This was used in robustness check for the influence of relationship scope on terms of financing. Though, it is not an ideal proxy for relationship scope, it is still more precise than measures used by other researchers: for example, Degryse and Van Cayseele (2000) used a dummy variable, which equals to one if the lender is the firm’s main bank. On top of that I consider a new loan characteristic, which can be influenced by bank-borrower relationship strength and has not been investigated in previous studies – bank’s requirement to provide a guarantee on loan. Surprisingly, although some research addresses the effects of relationship on collateral requirements, guarantees were not previously considered. Nonetheless they are another significant means of securing the loan and it was beneficial to unveil the effects (if any) of stronger relationship on probability of SME’s loan to be guaranteed. Finally, I distinguished collateral into sub-categories that, to my knowledge, have not been used before. Particularly, I distinguish between accounts receivable and

(8)

8

inventories (usually provided by the riskiest borrowers), “business” and “personal” collateral (while the former can be both “outside” and “inside”, the letter is always “outside” meaning that it is not only rearranges the order of claims between the creditors, but also provides additional assets to secure the loan). This allowed me to be more precise when controlling for collateral provided to secure the loan, since different types of collateral are, firstly, provided on loans of different quality and, secondly, decrease bank’s exposure to credit risk to various degrees.

In order to answer my research question I test three hypotheses: do stronger firm-bank relationships lead to lower interest rates on loans and lower probability of the loan to be secured by collateral or guarantee. I use four main relationship measures: length of relationship, physical proximity of the firm’s headquarters to the bank, share of the most recent approved loan in client’s total credit portfolio, and number of alternative sources of finance. For estimation of multiple linear-logarithmic regression I used ordinary least squares method, for models of binary choice – logit and probit regressions. I also control for multiple loan, firm and macroeconomic characteristics.

I managed to provide empirical support for the hypothesis that relationship strength is beneficial for the borrower in terms of loan interest rates: companies with established relationships with their banks are obtaining loans at lower rates than those without. The dimension of relationship that has a significant impact on loan interest rates is relationship scope measured by physical proximity to the bank, fewer sources of finance and higher proportion of bank’s loans in client’s loan portfolio. Longer relationship, on the other hand, does not help small businesses benefit from lower cost of financing. Since there is no overall unity of opinions on this subject matter, I consider my contribution to provide support for proponents of the theory that stronger relationship helps to decrease information asymmetry between the borrower and the lender, helping to estimate default risks more precisely and, thus, offer loans at lower interest rates such as Petersen and Rajan (1994), Berger and Udell (1995), Bharath et al. (2011), Bolton et al. (2016).

(9)

9

I also showed that collateral requirements become less stringent with the increase in relationship strength. In this case, I provided empirical evidence that both relationship scope and relationship length matter for determination of collateral requirements. This result is in line with theoretical model of Boot (2000), but provides counterargument to the empirical study of Degryse and Van Cayseele (2000).

Finally, on the new topic of influence of relationship strength on guarantee requirements from the bank, I showed that only relationship scope, measured by the number of alternative sources of financial services, plays a role. The bigger the scope of relationship, the lower the probability of a loan to be secured by a guarantee.

The thesis proceeds as follows: in Section 2 I provide a comprehensive overview of existing literature on relationship banking. Section 3 specifies hypotheses to be checked and corresponding regression models, and also gives description of methodology applied. Section 4 provides information on the sources used to collect the data and descriptive statistics for the sample of firms used in the study. In Section 5 I present the main results and in Section 6 – robustness checks.

Section 7 concludes.

2. Literature review

My study is most closely related to a strand of literature dedicated to relationship banking which focuses on the effect of strength of relationship between a borrower and a bank on availability of credit and terms of financing. My study is directly related to these works and builds upon them. Second, I provide an overview of other topics in relationship banking such as influence of credit market competition, effects of macroeconomic conditions and business cycle, changes occurring as firm-bank relationship matures, role of relationship lending for financially distressed firms.

2.1 Relationship and availability of credit

Existing literature studying relationship banking draws similar conclusions with regards to its effects on the availability of credit. According to Petersen and Rajan (1994), the availability of credit

(10)

10

increases in relationship length as well as in concentration of borrowings and other financial services from the same lender. Cole (1998) also showed that the smaller the number of sources of financial services, the higher the chances of credit being extended. He also found that having pre-existing relationship with the bank improves the availability of credit, although the length of that relationship is not important. Deviation of this result from Petersen and Rajan (1994) can be explained by the difference in measures of credit availability: Cole (1998) used the probability of existing credit being extended, while Petersen and Rajan (1994) used an inverse measure of the amount of borrowing from alternative sources. The latter implies that firms with longer relationships are less likely to be financed by external sources, since they obtain new credits from their relationship bank but not necessarily due to extension of existing loans. This does not contradict the positive effect of relationship strength on the availability of credit if the new credit is obtained by the firm on better terms, for example, bigger commitment or longer maturity. More recent research of Iyer et al. (2013) investigated the effect of credit crunch during the 2007–2009 crisis on availability credit and found that credit supply reduction is strong and binding for firms with weaker banking relationships.

Elsas and Krahnen (1998), using data from credit-files of five major German banks during the period from 1992-1996, showed that so called “housebank” – primary lender for a firm – is able to guarantee provision of liquidity in case of firm’s financial strain because it uses superior information sources in order to access credit risks.

In a more recent study, Cenni et al. (2015) found that the probability of being rationed – bank’s decision to deny credit to a firm that needed financing and applied for credit – for small and medium-sized companies increases in the number of banks and decreases in debt concentration and longevity of relationship. Results of the studies mentioned above are overall quite consistent, and, additionally, they used different geographical regions and timeframes, as well as different measures for variables of interest, which makes the conclusions more robust. Cenni et al. (2015) used Italian manufacturing firms in 2000-2003, while studies of Cole (1998) and Petersen and Rajan (1994)

(11)

11

used the data on the U.S.A. small businesses in 1988-1989. Moreover, Cole (1998) and Petersen and Rajan (1994) used indirect measures, while Cenni et al. (2015) use a direct measure of being rationed: firm desired more credit at the market interest rate and applied for it but was not granted the credit. In sum, there is extensive support in favor of forming a tight connection with the bank in order to improve company’s access to financing, and more so for small and informationally opaque companies.

2.2 Relationship and cost of borrowing

Researchers, who studied the influence of firm-bank relationship on cost of credit, provide inconclusive results. For example, Petersen and Rajan (1994) found that firms with multiple bank relationships are charged significantly higher rates and that rates are unaffected by the relationship length or usage of other financial services from the bank. At the same time, Berger and Udell (1995) found that borrowers with longer relationships pay lower interest rates. Both studies used NSSBF data for 1988-1989 but divergence of their results can be explained by the fact that, while Petersen and Rajan (1994) used all types of bank loans, Berger and Udell (1995) used only lines of credit. Thus, it is not surprising that the former were able to find a stronger connection, since for lines of credit reputation and relationship effects are significantly more important. Other types of loans, such as mortgages, equipment and motor vehicle loans and other “spot products”, do not imply bank’s commitment to allow the borrower to use a credit facility on a continuous basis in the future, so they are “transaction-driven” in contrast with lines of credit, which are “relationship-driven”. Consequently, inclusion of those loans could have weakened the effect of relationship length on loan pricing. Elsas and Krahnen (1998) also did not find significant link between length of relationship and the loan price. The authors themselves question the conclusion that relationship banking is unprofitable, since lender does not seem to require compensation for providing insurance against liquidity constraint in case of worsening of firm’s financial situation. As a potential resolution to this riddle, one can argue that bank can extract those compensating rents from other products sold to

(12)

12

the same customer such as cash management and consulting services, foreign exchange products, and many more.

Similarly to Berger and Udell (1995), Bharath et al. (2011) identified that repeated borrowing from the same bank leads to a decrease in loan spreads, and this pattern is more pronounced for informationally opaque borrowers. These finding are in line with theoretical model of Boot and Thakor (1994), which shows that prior to the first project realization, a firm is financed by a bank at above-market borrowing cost but upon first success the firms is financed at below-market rates, suggesting decrease in the borrowing costs with maturity of relationship.

Contrary to the studies mentioned above, Degryse and Van Cayseele (2000) showed that loan rates increase in the duration of relationship between a borrower and a bank and decrease in the scope of this relationship. Ioannidou and Ongena (2010) also showed that as relationship matures, the bank is able to extract rents due to the “hold-up problem”: since information asymmetry between a new bank and a firm is much stronger than between the same firm and its existing lender, the latter can charge higher interest rates due to high switching costs. Those empirical findings are in line with the theory of Greenbaum et al. (1989) and Sharpe (1990), who showed that the current lender charges interest rate that exceeds its cost of funds and average offer of outside lenders. The latter, in turn, are willing to provide credit at below-cost of funds rate in order to win the client in the hope to recoup its losses in the future, when client becomes locked in the relationship.

Bolton et al. (2016) studied influence of relationship strength on loan pricing during good and bad macroeconomic conditions. They found that relationship banks provide credit at lower rates during crisis but compensate with higher rates during better times, banks are able to make an intertemporal shift of information rents.

Since the studies mentioned above provide conflicting evidence, one of the hypotheses to check in the study will be regarding dependence of borrowing costs on relationship strength.

(13)

13

2.3 Relationship and collateral requirements

Among studies which considered collateral requirements2_{there is a general agreement that}

likelihood of pledging collateral decreases in longevity of relationship. This result is consistent with theoretical arguments that relationship lending generates valuable information about borrower quality as summarized by Boot (2000). Lower information asymmetry allows for more precise evaluation of credit risks, thus collateral requirements can be loosened. Degryse and Van Cayseele (2000) found that the scope of relationship has a negative impact on collateral requirements (they become more stringent). The reason for this result can be the fact that they use as a measure for the scope of relationship binary variable that equals one if the lender is the company’s main bank.3_First

of all, it is not a very precise measure of relationship scope. Second, it could be expected that a bank that is a main source of credit for the company, having more influence on this borrower, is more likely to request collateral. Since results are again inconclusive, I would also like to study the impact of relationship on collateral requirements further.

2.4 Physical and cultural proximity between a bank and a borrower

An alternative measure of firm-bank relationship, quite extensively used in existing literature, is a physical distance from the firm’s headquarters to its bank’s branch, this firm uses. For example, Agarwal and Hauswald (2010) studied how it affects bank’s ability to collect private information about its borrower. They showed that physical proximity eases acquisition of “soft” information and improves its quality, thus improving availability of credit, although it is offered at a higher interest rate. Building on that, Hauswald, and Marquez (2006) found that the probability of a business to be won over by a competing bank increases in the distance between this firm and the bank (and in proximity to this bank’s competitors), in the loan price and in the quality of the borrower. This happens because competition drives investment in information acquisition down, facilitating the switch.

2_{See, for example, Berger and Udell (1995), Bharath et al. (2011), Degryse and van Cayseele (2000).}

3_{“Main bank” is defined by the following characteristics: the firm has at least 100,000 BEF turnover per}

(14)

14

More recent research finds that cultural proximity of a bank is of greater importance than physical one due to the fact that it not only facilitates information acquisition but also interpretation of that information, since a loan officer and a borrower share cultural background, such as language, ethical codes, beliefs and so on. Cornell and Welch (1996), using data from an Indian bank showed that cultural proximity reduces information asymmetry in lending relationships through improvement of the precision of the signal that the loan officer obtains about borrower’s creditworthiness.

2.5 Transaction lending versus relationship lending

Another strand of literature considers motivation for firms to choose a particular bank and type of financing (transaction or relationship lending) and decisions about leaving the bank. Beck (2014) found evidence that both techniques are equally beneficial during normal macroeconomic state, but relationship lending provides substantial advantages during business cycle downturns. The effects are the most pronounced for younger and smaller firms, which lack tangible assets and cannot compensate with other sources of debt, as well as in case more dramatic cyclical downturns. Research of Iyer et al. (2013) is focused on effects of the credit crunch of 2007-2009 on availability of credit. According to this study, decline in credit supply was the most harmful for small and young firms, firms that do not have access to other sources of external finance and firms with weaker banking relationships.

Ongena and Smith (2001) showed that as a relationship with the bank matures, likelihood of a firm ending it increases, suggesting that benefits of decreasing information asymmetry deteriorate with time and, more importantly, they provide evidence against “hold-up” theories since borrowers are able to end relationships before they get locked in with their main bank. Firms that are relatively smaller and more levered, as well as firms that use multiple banks as financing sources, maintain shorter bank relationships. The theory suggests that those companies should be the most opaque and thus most interested in benefits of a bank acquiring “soft” information about them.

(15)

15

Consequently, they should suffer from lock-in the most. However, empirical evidence from Ongena and Smith (2001) suggests that switching costs are still low enough for the firm to be able to transfer its business to a different bank. According to Farinha and Santos (2002), “firms with more growth opportunities and firms with poor performance” are more likely to switch banks. The fact that after those companies obtain additional financing from other banks, they tend to invest more and perform even worse, suggests that increased costs of credit due to hold-up problems and unwillingness of the incumbent bank to extend or increase the amount of credit facility is the key driver for multiple bank relationships.

A number of studies investigated comparative advantages of large and small banks in relationship lending. For example, Berger and Black (2011), using data from the U.S.A. small business, showed that small banks benefit more from relationship lending, especially when financing relatively bigger firms. This can explain lack of incentive for large banks to be involved in collecting “soft” information about its borrowers, pointed out by Uchida et al. (2012). Hower (2016) also found evidence that small banks are better at processing “soft” information. Cole et al. (2004) showed that highly formalized decision-making based mostly on data from firm’s financial statements is usually employed by large banks.4 Small banks pay more attention to borrower’s character, information

about which they acquire through pre-existing relationship with him. Moreover, they show that influence of pre-existing relationship on probability of credit being extended is more pronounced for small banks, although the effect is ambiguous: pre-existing deposits improve availability of credit, while pre-existing loans hurt it. In agreement with aforementioned studies Degryse and Ongena (2007) provided evidence that “larger bank branches lend substantially more on a transactional basis but are less likely to be specialized in particular industries”.

(16)

16

2.6 Competition in debt market

Several studies addressed influence of interbank competition on benefits of relationship banking. Boot and Thakor (2000) propose theoretical model that shows that the higher the interbank competition, the more banks engage in relationship lending, although they also showed that the higher the competition with capital markets, the less are the gains from relationship loans. Empirical paper of Petersen and Rajan (1995) shows that young firms in concentrated credit markets get more access to institutional finance, while for old firms interbank competition influence on availability of credit is less pronounced. They also find that in highly competitive markets creditors tend to smooth interest rates over the business cycle: younger firms obtain credit at below-competitive rate, while old firms get charged above-competitive rate, which is another example of intertemporal shift in banks gains from relationship lending. Sharpe’s (1990) model also shows that competition incentivizes lenders to give loans to new firms at below-cost interest rates, thus initially producing losses on their loans. Degryse and Ongena (2007) provided evidence that stronger competition between banks is associated with higher proportion of relationship banking in the market.

At the same time, there are several studies that show the opposite effect of credit market competition on relationship lending. Theoretical model of Ogura (2010) proposes that increased competition drives rents from acquiring private borrower-specific information down. He also provides empirical evidence supporting this model. Boot and Thakor (2000) in their theoretical model showed mechanisms by which an increase in competition in credit market leads to a decrease in bank’s sector specialization, thus reducing benefits of relationship lending.

Some studies do not find statistically significant difference between concentrated and competitive markets in terms of benefits of investing in acquisition of proprietary information. For example, Neuberger et al. (2008) did not find significant influence of local interbank competition on number of banking relationships. But probably, closest to the true representation of the influence of competition in credit markets is the study of Elsas (2005), who showed that for lower levels of

(17)

17

concentration in credit markets, relationship lending becomes more beneficial with increase in competition. This provides evidence contradicting the proposition that relationship “lending requires monopolistic market structures”. Although for high levels of credit market concentration, the effect is opposite. Presbitero and Zazzaro (2011) argue that in markets, where large banks dominate the market, increase in competition hurts relationship lending more severely than in markets, where small banks are more powerful and “soft” information technologies are already widely used. In the latter type of markets increased competition can actually increase banks’ incentives to invest in developing tighter connections with their clients.

2.7 Relationship and use of covenants

An important instrument that banks can use to ease the process of relationship lending is the usage of covenants – special conditions of loan contracts that facilitate monitoring and make credit facility contingent on borrower’s financial position. Banks will be more willing to provide credit on special terms, if they are able to make contracts more flexible, thus, ability to introduce covenants should facilitate relationship lending. Denis and Wang (2014) showed that when a loan contract is signed, covenants are usually relatively strict (most likely due to the severity of information asymmetry at the start of relationship) but they usually get renegotiated before they are violated and even before they are close to be violated. This implies that as banks collect inside information about their borrowers, they are willing to relax existing constraints because of reduction in perceived credit risk.

2.8 Costs and benefits for a lender

A branch of studies discusses costs and benefits of engaging in relationship lending for banks themselves. One of them by Schenone (2010) investigated whether banks exploit their information advantage by charging their relationship borrowers higher interest rates. She used IPO as a natural experiment since it is a significant information-revealing event for a company. Thus, if relationship lending is beneficial for banks, there should be a difference in terms of financing prior- and post-IPO. Specifically, she found that prior to IPO banks extract rents from acquiring proprietary

(18)

18

information about their borrowers since firms face high switching costs and are “locked into a relationship”. After IPO, when those costs significantly decrease, relationship banks decrease those borrowers’ interest rates on loans.

According to Bharath et al. (2007), bank is more likely to win a loan contract from its relationship borrower. This means that as long as the increase in the amount of credit is profitable for a bank, it is beneficial for this bank to invest in relationships. Same applies to other fee-generating banking products.

The major “cost” of relationship lending that needs to be taken into account is probability of default. If relationship lending helps to mitigate information asymmetry problem and evaluate credit risk more accurately, there should be relatively less defaults on relationship loans. On the other hand, bank-borrower proximity can introduce agency costs because of personal relationship between a loan officer and a borrower, in which the former may be incentivized to conceal negative information regarding firm’s financial perspectives. In support of the first proposition, Norden and Weber (2010) showed that, twelve months prior to default, borrower’s cash flows and credit line usage “exhibit abnormal patterns”. This means that relationship bank, able to observe client’s cash flows, as well as provide right incentives and closely monitor the borrower with the help of credit covenants, will be more capable of preventing default. In support of the second theory, Berg et al. (2013) showed that loan officers use multiple trials, when evaluating client’s creditworthiness using internal banking systems, in order to get the desired outcome. In other words, when using scoring systems, loan officers tend to manipulate information in order to get the credit approved for their “relationship clients”. They showed that such loans are more likely to default. Thus, developing personal relationship with a borrower might actually be detrimental for a bank.

2.9 Mechanisms of relationship lending

Existing literature provides limited information regarding exact mechanisms through which banks acquire “soft” information from its borrowers and what does this information actually consist of.

(19)

19

Uchida et al. (2012) studied the role of commercial loan officers in producing information about borrowers in Japanese the loan market for small and medium-sized companies. They found that more information production is associated with higher frequency of client-officer contact and lower frequency of loan officer turnover. This is evidence of critical role of loan officers in collecting “soft” information about borrowers. Uchida et al. (2012) also found that small banks are better at producing “soft” information than large ones, although they show that large banks are not less qualified to produce such information but rather decide not to do so.

Hertzberg et al. (2010) studied the treat of rotation of client from one loan officer to another to be incentivizing the former one to be more accurate in evaluation of its borrower’s financial prospects. The idea is that “soft” information about the borrower cannot be formalized and easily used transferred to the other employees of the same bank. This can lead to loan officers’ moral hazard problem. From this perspective collection of proprietary information might not be that beneficial for a bank. Hertzberg et al. (2010) showed that rotation is a credible treat for a loan officer to reveal any negative information about the borrower’s solvency and to improve communication with other loan officers because the fact of concealing potential problems with the borrower is more hurtful for loan officer’s career than the reveal of the problem by his successor.

2.10 Other topics in relationship banking

Hower (2016) studied the role of banks in financing financially destressed firms and found that those firms are better protected from financial shocks if they maintain multiple bank relationships. At the same time, in case of firm obtaining resources from several banks, coordination and free-riding problems become more pronounced. This makes debt restructuring more problematic. He also showed that firms that do not have long-term relationship with their financing bank have higher chances of experiencing financial distress and defaulting on their debt.

Ongena et al. (2011) studied factors influencing firm’s decisions with regards to how many and which banks to choose for being sources of financial services. Companies whose priority is

(20)

20

reputation of the bank usually use fewer banks and are unlikely to reduce the number and quantity of services used from their servicing banks. Firms for which pricing of banking products is of utmost importance are more likely to leave the bank or to reduce the amount and number of services used. This also implies that they do not experience restrictively high switching costs.

Berlin and Mester (1999) studied effects of banks’ access to funding on relationship banking. They used availability of core deposits with inelastic interest rates, assuming that easier access to them will facilitate relationship banking by making borrower-specific contractual agreements feasible, as opposed to the situation, when banks have to pay market price for funds. Indeed, Berlin and Mester (1999) found that access to core deposits helps banks to insure its relationship borrowers from credit shocks.

Degryse et al. (2011) investigated the effects of bank mergers on relationships with merging banks’ current clients and found that borrowers with a single bank relationship are more likely to be dropped as a result of merger. Firms maintaining multiple bank relationships, in turn, are less harmed by mergers, since they can more effectively hedge themselves against termination of relationship.

3. Methodology

3.1 Outcome variables specification

Firstly, I investigate how relationship strength influences the original interest rate on the most recent loan the company has obtained. Original interest rate is used because, firstly, it allows me to use the whole sample of firms which were approved credit irrespectively of the type of interest rate used: at the loan initiation fixed and floating rates that a bank is willing to offer on a specific loan application should be equal because they both reflect credit risk on this loan, and the bank can set them so that it will be indifferent between the two. Secondly, the initial interest rate reflects bank’s evaluation of client’s creditworthiness and is not dependent on changes in borrowing costs: floating rates change over time with the changes in the underlying rate over which the premium is defined.

(21)

21

This methodological aspect distinguishes my study from similar ones such as Berger and Udell (1995), who used the same survey but for earlier period of time. Their approach was to use only floating interest rate loans and investigate effect of relationship on premium over prime rate.

Secondly, I consider how relationship strength influences collateral and guarantee requirements. For this I use two dummy variables, one of which equals one if collateral was used to secure the loan and zero otherwise, and second equals one if guarantee was provided and zero otherwise.5

3.2 Relationship variables

As measures of relationship strength I will, first, use natural logarithm of one plus distance from the firm’s headquarters to the bank,6 assuming that the shorter this distance is the easier it is to monitor

the borrower and, thus, the stronger the relationship. Secondly, I will use number of financial institutions that can be used by the firm as sources of finance. This is an inverse measure of relationship scope, meaning that the higher the number of financial intermediaries the firm interacts with, the weaker its relationship with the bank which provided the most recent loan.

For robustness checks I will use a share of the most recent loan in the total credit portfolio of the firm and natural logarithm of one plus the number or years the firm was serviced by the bank, which approved the most recent loan. The higher the proportion of financing acquired from the same bank, the stronger the relationship between the bank and the borrower. A cleaner measure for the share of the bank in client’s credit portfolio would have been a proportion of all the loans, provided by the bank that approved the most recent loan to the firm, in borrower whole credit portfolio. Unfortunately, 2003 NSSBF survey structure does not allow construction of such a measure. Relationship length is another dimension of measuring relationship strength (in contrast with relationship scope), which is commonly used in the existing literature.

5_{For detailed information on variables specification (both dependent and independent) as well as expected}

signs for regressors and control variables, please, refer to the Appendix 1.

(22)

22

3.3 Firm characteristics

Among firm characteristics I control for the firm size measured by natural logarithm of one plus number of employees,7_{firm age measured by natural logarithm of one plus number of years the}

firm has existed, firm profitability measured by profit in the last financial year over sales for the same period, firm growth measured by the indicator variable which equals one if total sales in the last financial year were higher than in the previous year, and industry where the firm operates. Additionally as a proxy for company’s quality I use dummy variable equal to one if this company has at least one occasion during the past three years of being delinquent on one of its obligations for 60 or more days. The fact that the firm was delinquent on its obligations signals lower credit quality of this borrower and, thus, it should obtain loans at higher interest rates. Dummy variables for industries help to proxy for differenced in the degree of their riskiness, which influences both interest rates and requirements regarding collateral and guarantees.

Overall, the purpose of firm variables is to control for the observable risk of potential borrower that determines the loan agreement conditions. All else equal, it is expected that riskier clients: smaller, younger, less profitable, faster growing firms, as well as firms from certain industries, – would acquire loans at higher interest rates and are more likely to pledge collateral or provide guarantee on a loan. As proxies for quality of corporate governance I have added four dummy variables: whether the firm is owned exclusively by the members of the same family, whether the firm is owner managed, whether the firm is a corporation, and whether the firm is a partnership. These proxies are included because different ownership structures reflect different degree of transparency of the firm as well as different levels of risk that might be taken by the firm and propensity to shift that risk. The firms in the sample are quite homogeneous in their ownership concentration: in 98% of the firms the owner of the biggest share of the firm controls at least 20% of it, so we can say that

7_{As alternative measures of firm size in unreported results I also used natural logarithm of one plus total}

assets and natural logarithm of one plus total sales but they are stronger correlated with other control variables, thus, in main results I use an alternative measure.

(23)

23

almost all the firms have a blockholder. Since there is no significant variation in that characteristic, there is no need to control for ownership concentration.

3.4 Loan characteristics

With regards to loan characteristics I control for the loan size measured by the natural logarithm of one plus the amount of loan commitment, maturity measured by one plus natural logarithm of the commitment length, interest rate type (dummy variable equals one if the loan is obtained at a floating rate), type of collateral provided to secure the loan: accounts receivable and inventories, personal and business collateral, other; and whether or not a guarantee was provided. I expect interest rate to be increasing both in length and size of loan commitment since both are associated with higher risk. Dummy variable for floating interest rate is included in order to control for potential difference between those loans and loans with fixed interest rate. I expect floating interest rates to be relatively lower because they fluctuate with the underlying cost of capital because part of such rate is tied to a benchmark rate, which reduces the risk the bank takes.

There are two contradictory theories with regards to the use of collateral in debt contracts: borrower’s moral hazard and adverse selection. First argues that borrowers of lower quality (more informationally opaque and, presumably, riskier firms) turn to financial intermediaries who can reduce the demand for collateral by monitoring more closely (see Stulz and Johnson (1985)). All else equal, such borrowers should obtain loans at higher interest rates. Proponents of adverse selection point of view on collateral requirements (see Besanko and Thakor (1987)) argue that willingness to provide collateral serves as a credible signal of borrower’s quality thus should be associated with lower interest rates.

Important distinction that needs to be drawn to control for collateral pledged is between accounts receivable and inventories and other types of collateral, because loans, secured by accounts receivable and inventories are usually viewed as the riskiest type of working capital financing, which, in turn, should significantly affect loan interest rates. These types of loans usually involve a

(24)

24

form of monitoring not used for other types of loans: observation of cash flow movements on borrower’s current accounts, payment discipline of firm’s customers and inventory management, which are perceived as valuable sources of information about borrower’s financial situation. This information is even more important in evaluating young firm’s ability to repay the loan. As highlighted by Berger and Udell (1995), providers of loans secured by accounts receivable and inventories collect more information about their borrowers trough relationship than is typical for other types of loans. This information is then used to design loan contracts in the future.

Moreover, I further separate other types if collateral into three groups: “business” collateral, “personal” collateral and other collateral (that does not fall in any other categories). I expect loan’s most recent interest rate to be decreasing in all of those factors because they all provide additional assets to secure the loan or at least rearrange the priority of claimants to the bank’s advantage.8

In a similar fashion I control for a loan to be guaranteed. Guarantee provides a signal of borrower’s quality: if firm’s owner is willing to provide personal guarantee, it means that this potential borrower is of high quality and corresponding loan has lower probability of default (otherwise owner would not be willing to risk his own property).

3.5 Macroeconomic conditions

Since we use original interest rate on the most recent loan, it is necessary to control for underlying cost of capital. Firstly, to control for the cost of capital I obtain historical values of the U.S. prime rates from www.fedprimerate.com for the period 1986-2005, which covers the most recent loan application dates for the firms in my sample. This rate includes the risk-free rate and a default premium for the bank’s best customers. I collected observations for the prime rate on a monthly basis since each observation in the main sample provides information on a month and year when the loan was approved, thus I have matched it with the prime rate at the respective point in time.

(25)

25

I also control for the competition in the debt market since the degree of interbank competition should incentivize banks to offer more favorable conditions to its borrowers in order to attract them (see Petersen and Rajan (1995)). I measure the degree of interbank competition by commercial bank deposit Herfindahl-Hirschman Index for the year 2003. The index for each observation is matched by the metropolitan statistical area or county where firm's headquarters office is located. The loan market is considered to be concentrated, if this index has the value of at least 1800. The justification of usage of bank deposit index as a proxy for credit market concentration is provided by Petersen and Rajan (1995): the concentration of the market for deposits is a good approximation for the concentration of the market for credit, if the firms in the sample largely borrow from local markets because of the prohibitive informational and transactional costs of going outside. This is a realistic assumption since the sample includes small businesses that are unlikely to seek financing outside of the local markets: 75% of firms in my sample acquire loans from the bank that are located within 15 miles from the firm’s headquarters and only 5% of firms borrow from the banks that are located further than 200 miles away (this radius can be safely assumed to include “local” market).

3.6 Hypotheses

Now let’s consider the first hypothesis:

Hypothesis 1: stronger firm-bank relationship allows the firm to acquire loans with lower interest rates.

In comparison with existing studies, such as Berger and Udell (1995), who used the similar survey to the one I use but for 1988-1989 and measured the cost of borrowing by premium over prime rate, I am able to use a bigger data set, because they had to restrict the sample to floating interest rate loans whose premium is defined over the prime rate. In case of applying the same methodology,

(26)

26

I would have lost 80% of the sample, which would make the model less powerful.9_{Also the firms}

that choose floating interest rate might be different from those that choose a fixed one. Consequently, it is more valuable to study the whole sample of borrowers and control for interest rate type to, first, be able to study effects of relationship on both types of borrowers, and, second, to see if there is an inherent difference between them.

The regression equation to be estimated is the following:

𝑂𝑟𝑖𝑔𝑖𝑛𝑎𝑙 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑟𝑎𝑡𝑒_𝑖 = 𝛼₀+ 𝛼₁∙ 𝑅𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠ℎ𝑖𝑝 𝑚𝑒𝑎𝑠𝑢𝑟𝑒_𝑖+ 𝛼̅̅̅ ∙ 𝐿𝑜𝑎𝑛 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠₂ ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ + 𝛼_𝑖 ̅̅̅ ×₃ × 𝐹𝑖𝑟𝑚 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ + 𝛼𝑖 ̅̅̅ ∙ 𝑀𝑎𝑐𝑟𝑜𝑒𝑐𝑜𝑛𝑜𝑚𝑖𝑐 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛𝑠4 ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ + 𝜀𝑖 𝑖 (I)

where:

𝑅𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠ℎ𝑖𝑝 𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑖 depends on specification:

 Natural logarithm of one plus length of relationship;

 Natural logarithm of one plus distance from the bank;

 Number of financial sources the firm has

 Share of the most recent loan in firm’s loan portfolio.

Variable definitions for vectors of loan and firm characteristics and macroeconomic conditions as well as for relationship measures are provided in Appendix 1.

This multivariate linear regression can be estimated by simple ordinary least squares method. Under the assumption that relationship lending facilitates reduction of information asymmetry and is overall beneficial for the borrower, we would expect loan interest rate to decrease in the strength of firm-lender relationship. Thus, we would expect the coefficient of interest 𝛼₁ to be:

 Negative for the natural logarithm of one plus relationship length;

 Positive for the natural logarithm of one plus distance from the bank;

9_{In fact, models of Berger and Udell (1995) lack statistical significance and predictive power as measured by}

(27)

27

 Positive for the number of financial sources;

 Negative for the share of most recent loan.

Regarding control variables, it is expected that loans with longer maturities and bigger size of commitment, should be approved at higher interest rates due to bigger credit exposure. Loans obtained at a floating interest rate are expected to be lower since, in case of unpredicted change in the cost of capital for the bank, such interest rates will be automatically recalculated, in contrast with fixed rates, which would have to be renegotiated. Loans secured by any type of collateral, except for accounts receivable and inventories, are likely to have lower original interest rates. Exception for accounts receivable and inventories can be explained by the higher riskiness of loans, secured with accounts receivable and inventories, since such collateral is usually used to finance firms that cannot provide other type of collateral of higher quality. Such collateral typically secures loans to firms with low proportion of fixed assets on their balance sheet and operating in riskier industries (for example, retail trade). Older and bigger firms are expected to be able to obtain loans at lower interest rates due to the fact that they have stronger reputation. Moreover, relationship length with the bank that provided the most recent loan and firm age are significantly correlated (at 1% level), thus, older firms tend to have longer relationships with their financing bank, which means that they should be able to negotiate better terms when it comes to interest rate. Nonetheless, it is still needed to include firm age into the model to avoid bias in the coefficient for relationship length, because the age of the firm reflects information, available to the whole market (in other words, it is company’s public reputation), while relationship length reflects proprietary information which the bank collects over time (see Berger and Udell (1995)).

I expect more profitable firms to get loans at higher interest rates, which is in line with the “hold-up” theory, meaning that the bank will exercise its power over the client that it gains as relationship matures in order to make an intertemporal shift in its gains from lending to the firm: in the beginning, when the firm was young unprofitable the bank might have financed it at below cost rate,

(28)

28

but now, when the firm is more established, the bank is able to recoup those losses. Growing firms (for which indicator variable for sales growth equals one) should still obtain loans at lower cost.

Regarding governance characteristics, I expect firms with poorer corporate governance (family owned companies, proprietorships and firms that are not owner managed) to be subject to higher interest rates. Firms that were delinquent on its obligations should also incur higher borrowing costs. In concentrated markets lenders should be able to extract higher rents from borrowers. All the expected directions of influence for explanatory and control variables are summarized in the

Appendix 1.

Secondly, I would like to test how relationship with the bank affects the requirement to pledge collateral. Outcome variable in this case is binary and equals one, if collateral was pledged to secure the loan, and zero, if it was not.

Hypothesis 2: stronger firm-bank relationship decreases probability of the firm to pledge collateral on its loans.

Since dependent variable in not continuous, we need to use the model of binary choice:

Pr(𝐶𝑜𝑙𝑙𝑎𝑡𝑒𝑟𝑎𝑙𝑖= 1) = 𝐹(𝑥̅ ′𝛽) =𝑖

𝑒𝑥̅ ′𝛽𝑖

1 + 𝑒𝑥̅ ′𝛽𝑖

where:

𝐶𝑜𝑙𝑙𝑎𝑡𝑒𝑟𝑎𝑙_𝑖 – dummy variable, equals one, if collateral was pledged for the most recent loan of company i,

β – vector of parameter estimates for the independent variables,

𝑥̅ – vector of relationship and control variables. _𝑖

(29)

29 Ln ( 𝑃𝑟{𝐶𝑜𝑙𝑙𝑎𝑡𝑒𝑟𝑎𝑙𝑖=1} 1−𝑃𝑟{𝐶𝑜𝑙𝑙𝑎𝑡𝑒𝑟𝑎𝑙𝑖=1}) =𝛽0+ 𝛽1∙ 𝑅𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠ℎ𝑖𝑝 𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑖+ 𝛽2 ̅̅̅_{∙ 𝐿𝑜𝑎𝑛 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠}̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅_𝑖_{+ 𝛽} 3 ̅̅̅_× × 𝐹𝑖𝑟𝑚 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅𝑖+ 𝛽̅̅̅₄∙ 𝑀𝑎𝑐𝑟𝑜𝑒𝑐𝑜𝑛𝑜𝑚𝑖𝑐 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛𝑠̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅𝑖+ 𝜀𝑖 (II)

Under the assumption that collateral requirements become less stringent with the strengthening of relationship, expected sign of 𝛽1 is:

Firm characteristics and Macroeconomic conditions vectors of independent variables are the same as

for the first regression specification. I did not exclude Prime rate as a control from regression because I expect cost of capital to have an effect on loan contract terms regarding collateral requirements, because cost of capital reflects bank’s ability to obtain financial sources at reasonable price in order to give out loans. It is likely that collateral requirements become more stringent in case of negative tendencies in the cost of capital since, first, bank’s profit margins on loans are likely to tighten and it will need to compensate for credit risk with the help of other means and, second, bank’s requirement to provide collateral may play a role of a restrictive factor for the firm to obtain a loan in case of increasing of systematic risk in the banking industry. The vector of loan characteristics now does not include any collateral or guarantee dummies because the decisions to require collateral of different types and guarantee are likely to be co-determined. Appendix 2 provides variables description and expected signs for independent variables for testing the

Hypothesis 2.

Finally, I would like to investigate the influence of firm-bank relationship on the requirements to provide a guarantee from a third party: an owner, another individual, a firm, or a government agency.

(30)

30

Hypothesis 3: stronger firm-bank relationship decreases probability of the firm to provide guarantee for its loans.

A guarantee is most often provided by the firm’s owner and gives a lender the right to demand loan repayment by selling owner’s assets in case of default on the loan. It is similar to the owner pledging her own property as collateral but, unlike the latter, it does not specify the lien. Thus, guarantee provides the lender with more flexibility in ensuring loan repayment in case of default than personal guarantee. If the guarantee is provided my another firm, it means that the lender is able to demand loan repayment in case of default by means of selling this firm’s assets or with cash holdings this firm has accumulated.

For this hypothesis I will estimate a model, similar to the second one but with dependent variable reflecting the probability of guarantee being used to secure the loan:

Pr(𝐺𝑢𝑎𝑟𝑎𝑛𝑡𝑒𝑒_𝑖= 1) = 𝐹(𝑥̅ ′𝛽) =_𝑖 𝑒𝑥̅ ′𝛽𝑖 1 + 𝑒𝑥̅ ′𝛽𝑖 Ln ( 𝑃𝑟{𝐺𝑢𝑎𝑟𝑎𝑛𝑡𝑒𝑒𝑖=1} 1−𝑃𝑟{𝐺𝑢𝑎𝑟𝑎𝑛𝑡𝑒𝑒𝑖=1}) =𝛼0+ 𝛼1∙ 𝑅𝑒𝑙𝑎𝑡𝑖𝑜𝑛𝑠ℎ𝑖𝑝 𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑖+ 𝛼̅̅̅2∙ 𝐿𝑜𝑎𝑛 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠𝑖 ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅_{+ 𝛼}_̅̅̅₃_× × 𝐹𝑖𝑟𝑚 𝑐ℎ𝑎𝑟𝑎𝑐𝑡𝑒𝑟𝑖𝑠𝑡𝑖𝑐𝑠̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅_𝑖+ 𝛼̅̅̅₄∙ 𝑀𝑎𝑐𝑟𝑜𝑒𝑐𝑜𝑛𝑜𝑚𝑖𝑐 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛𝑠̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅_𝑖+ 𝜀_𝑖 (III) where:

𝐺𝑢𝑎𝑟𝑎𝑛𝑡𝑒𝑒_𝑖 – a binary variable, which equals one, if a guarantee was provided to secure the most recent loan of company i.

Firm characteristics and Macroeconomic conditions vectors of independent variables are the same as

for the second regression specification. Similarly to specification (II), I did not exclude control for the cost of capital. Appendix 3 provides variables description and expected signs for independent variables for testing the Hypothesis 3.

(31)

31

Under the assumption that relationship strength mitigates information asymmetry problems between a borrower and a lender we would expect the probability of bank requiring a guarantee on a loan to decrease in relationship strength, thus the sign of 𝛽1 is expected to be:

Loans with longer maturity and bigger size of commitment should have stricter collateral and guarantee requirements due to higher risks associated with such loans. I also expect loans with floating rate to be less likely to be guaranteed or secured with collateral because the cost of capital is factored in the rate, which makes required compensation for credit risk lower than in case of fixed rate, which does not automatically change due to unforeseen changes in the cost of capital. With regards to the firm characteristics, I expect bigger and older firms to be more likely to secure the loan with collateral or guarantee because, as you can see from Table 6 Appendix 4, these parameters are highly correlated with loan size, in which probability of collateral or guarantee being used to secure the loan should increase. Moreover, such firms are more likely to have established relationship with the bank, which in turn should have a stronger influence over the client and can make collateral and guarantee requirements stricter.10_{Also, these firms are more likely to have}

assets that can be used as collateral in comparison with younger and smaller firms. I expect more profitable firms and firms with growing sales, as well as companies that have not been delinquent on any of their obligations in the last three years to have less strict collateral and guarantee requirements on their loans due to higher creditworthiness of such firms. Similar implications are expected for firms with better corporate governance: corporations and partnerships and firms that

10_{Alternatively, it is expected that the firm itself will be more willing to provide additional securitization for}

the loan, such as guarantee or collateral, as a sign of good faith, when it has a well-established relationship with its bank.

(32)

32

are owner managed. Finally, in more concentrated markets banks should be able to impose stricter collateral and guarantee requirements on their clients. All the expected signs for the coefficients for explanatory and control variables are summarized in Appendices 2 and 3.

4. Data and descriptive statistics

For the purpose of this study I used the latest survey of U.S. small businesses conducted by the National Opinion Research Center at the University of Chicago for the Federal Reserve Board (NSSBF) in 2003.11 This survey collected comprehensive information on firm and owner

characteristics and their usage of financial services for companies with fewer than 500 employees. The main section of interest in this survey is information on most recent loan obtained by the company, which is the one to be used in this research because it contains detailed information on the loan such as costs of obtaining this loan, type, size and maturity, means of securitization, as well as relationship characteristics with the bank which approved this loan. This source is quite extensively used in the area of research but, to my knowledge, the most recent survey has not been addressed in existing literature. Moreover, my methodological approach is different, thus, the usage of similar dataset should be justified. The limitation of this source is that it represents survey data, thus it can introduce some noise because it is based on the information provided by firms’ executives and not directly from the firms official documents.

The data was collected via telephone interviews with executives of almost 4,000 companies. The study uses information on 1,650 of them since, according to common practice in relationship lending literature, firms with SIC codes 6000-6999 – finance, insurance and real estate companies – are excluded from the sample,12_{as well as firms that did not apply for credit or were not approved}

11_{The program was terminated in 2006 so the data for 2003 is the most recent one.}

12_{Relationships between financial firms, such as banks, are fundamentally different from relationships}

between firms from other industries and financial firms. Thus, including them in the sample would complicate inferences to be made regarding the results of the study.

(33)

33

any credit during the last three years.13_{The sample includes 1,278 corporations (77% of all}

companies in the sample), 257 sole proprietorships (16%) and 120 partnerships (7%). Majority (36%) of firms from the sub-sample of interest operate in the service industry, 19% – in retail trade, 17% – in manufacturing, 12% – in construction. Most of those firms are owner managed (1,429 or 86% of the sample).

Companies in the sample are relatively young (median age is 16 years) and small (median number of employees is only 25, median size of total assets is 880,000 USD). Table 2 Appendix 4 gives a breakdown of firm size and age by industry.14_{The biggest companies are concentrated in the}

manufacturing industry with median size of total assets 2.776 million USD and median number of employees 52. The industry which represents the majority of companies in the sample – services industry – is the one with the smallest companies having median size of total assets 357 thousand USD and median number of employees 14.

Table 1 above provides descriptive statistics for the main sample of interest. Original interest rate

on a loan from the sample ranges from 0.5% to 30.0% with both average and median interest rate being around 6%. Most of the loans (75%) are acquired at a rate 7% or lower. Loan maturity varies from 0 (no fixed maturity) to 48 years in extreme cases. Median loan maturity is 1.5 years and rarely exceeds 5 years (only in 25% of cases). Loans of relatively short maturity are typical for small and medium-sized enterprises. Median loan size is 0.150 million USD and majority of loans are under 0.650 million USD.

13_{The reason is that we cannot compare loan characteristics for them because we can only observe terms of}

credit for the approved loans.