Key Influences on Loan Pricing at Credit Unions and Banks

Transcription

1 Key Influences on Loan Pricing at Credit Unions and Banks Robert M. Feinberg Professor of Economics American University With the assistance of: Ataur Rahman Ph.D. Student in Economics American University Prepared for the Center for Credit Union Research and the

2 Copyright 2004 by Filene Research Institute. ISBN All rights reserved. Printed in U.S.A.

3 Filene Research Institute and Center for Credit Union Research The Filene Research Institute is a non-profit organization dedicated to scientific and thoughtful analysis about issues affecting the future of consumer finance and credit unions. It supports research efforts that will ultimately enhance the wellbeing of consumers and will assist credit unions in adapting to rapidly changing economic, legal, and social environments. Deeply imbedded in the credit union tradition is an ongoing search for better ways to understand and serve credit union members and the general public. Credit unions, like other democratic institutions, make great progress when they welcome and carefully consider high-quality research, new perspectives, and innovative, sometimes controversial, proposals. Open inquiry, the free flow of ideas, and debate are essential parts of the true democratic process. In this spirit, the Filene Research Institute grants researchers considerable latitude in their studies of highpriority consumer finance issues and encourages them to candidly communicate their findings and recommendations. The name of the institute honors Edward A. Filene, the father of the U.S. credit union movement. He was an innovative leader who relied on insightful research and analysis when encouraging credit union development. The Center for Credit Union Research is an independent academic research center located in the School of Business at the University of Wisconsin Madison. The Center conducts research and evaluates academic research proposals on subjects determined to be priority issues by the Research Council of the Filene Research Institute. The Center also supervises Filene Research Institute projects at other universities and institutions. The purpose of the Center s research is to provide independent analysis of key issues faced by the credit union movement, thus assisting credit unions and public policymakers in their long-term planning. Progress is the constant replacing of the best there is with something still better! Edward A. Filene i

4 ii

5 Table of Contents Executive Summary CHAPTER 1: Introduction CHAPTER 2: Bank Pricing Behavior CHAPTER 3: CHAPTER 4: Credit Union and Bank Pricing Behavior A Comparison Internal Influences on Credit Union Pricing APPENDIX: Data and Methodology REFERENCES About the Author Filene Research Institute Administrative Board/Research Council Filene Research Institute Publications iii

6 iv

7 Executive Summary PURPOSE How does the competitiveness of local markets for consumer loans affect bank and credit union pricing of these loans? How do bank and credit union behaviors differ in this respect? How do internal variables affect credit union loan pricing? How can regulatory policy support or inhibit credit unions in carrying out their service and social missions with regard to loan pricing? These are the questions we address in this study. Their answers have important implications for credit unions and their regulators, in particular, with regard to capital requirements. METHOD We begin with a model that explains average prices of banks in local markets, for which the literature provides considerable precedent. This includes some prior research on how the presence of credit unions influences the average price in local markets. From this model of average prices, we move to a model that explains prices at individual banks, for which a few precedents exist. We then extend the analysis to a model that explains prices at individual credit unions, for which no precedents exist. This approach allows us to confirm elements of these models which existed in the literature as well as to extend these models in new ways to answer our research questions. The source of data for developing these models includes a Federal Reserve survey of bank loan rates on two types of loans, new vehicle loans and unsecured (non-credit card) loans. We also rely on call report data for both banks and credit unions and on market share data from Sheshunoff Information Services. This study is one of the first to model how internal variables, specific to an institution, affect its pricing behavior. Because it employs the latest available technical research tools, we include in the text and appendix of this report detailed information about these methods and their support of the research findings. KEY FINDINGS As predicted by their different ownership, governance structure, and objectives, credit unions pricing behavior differs notably from that of banks. 1

8 For unsecured (non-credit card) loans, measures of market competitiveness affect bank pricing but not credit union pricing, and credit union rates adjust to changing conditions even slower than the already slow adjustment speed of bank rates. However, compared to new vehicle loan rates, unsecured loan rates are relatively insensitive to changes in market conditions and are slow to change for both types of institutions. Banks exhibit pricing behavior consistent with profit maximizing behavior. New vehicle loan prices are higher at larger banks, in less competitive markets, and among banks that are members of top ten bank holding companies. Proxies for ease of expansion of credit unions in a state indicate that relative ease reduces new vehicle loan rates at individual banks and credit unions. New vehicle loan prices are lower at larger credit unions. Credit unions do not raise their rates in less competitive markets, a behavior that is consistent with their overall objectives. Internal variables at credit unions can influence their loan rates for new vehicles. In particular, we expect capitalconstrained credit unions to be less able to offer attractive rates to members. Of the four internal variables that reflect possible effects of capital constraints, three operate in the expected direction and are statistically significant. IMPLICATIONS The evidence indicates that internal as well as external variables influence loan rates. Capital-constrained credit unions appear to charge higher loan rates on both new vehicle loans and unsecured loans. This suggests that the regulation of capital for credit unions has two important objectives. The traditional objective is to assure enough capital to maintain safety and soundness. Our results suggest a second objective: Regulatory policy should also avoid requiring capital standards so high relative to risk that they distort pricing decisions, thereby undermining the purpose of credit unions which is to meet social and service objectives rather than to maximize profits. Regulatory policy should avoid capital standards that distort pricing decisions. 2

9 CHAPTER 1: Introduction PURPOSE The purpose of this research is to: 1) evaluate what determines loan prices for individual banks and credit unions of similar size and market share; 2) evaluate how credit union loan rates are influenced by size, net income, and capital constraints that may limit growth; and 3) develop the policy implications of the answers to these questions. Prior research examined the influence of credit unions on local market rates for loans and found that as a group their presence moderates market prices. This research extends the analysis to the level of individual institutions, in order to examine what effects the presence of credit unions has on bank loan prices. In addition, it allows us to assess whether capital constraints on credit unions could adversely affect credit union members and bank customers, whose rates also may be influenced by rates offered at credit unions. APPROACH We obtained data on bank rates for unsecured (non-credit card) loans and on new vehicle loans from a survey conducted by the Federal Reserve that provides quarterly data for the period , over a total of 72 markets. We obtained data from the same period for credit unions for the same markets from call reports. In addition we obtained information on market shares, based on deposits, from Sheshunoff Information Services. The appendix provides more detail on the data and statistical approaches used in the study. RESULTS ON PRICING BEHAVIOR: BANKS VERSUS CREDIT UNIONS Expected Differences in General: Banks versus Credit Unions Banks are investor-owned firms. We expect them to price so as to maximize profits. Credit unions are governed by an uncompensated board, elected by the membership on a oneperson, one-vote basis. They have service and social objectives beyond profit maximization, and they generally wish to keep loan 3

10 rates low to benefit their members. Therefore, we expect banks and credit unions to exhibit different behavior in pricing. To examine this issue we construct a pricing model for banks and credit unions to explain pricing for two types of loans, new vehicle loans and unsecured (non-credit card) loans. These models include variables that are likely to influence pricing. Although in general we anticipate these variables will affect new vehicle and unsecured rates in the same direction, we recognize that unsecured rates move considerably less than new vehicle rates in response to market conditions, for a number of reasons. Therefore, we expect the influence of the variables in the model to be stronger on new vehicle rates and weaker on unsecured rates. Put another way, we expect fewer variables to have a statistically significant influence on unsecured loans than on new vehicle loans. Variables Table 1.1 Findings (Empty cells represent relationships that were not statistically significant) Bank Effect on Loan Rates Credit Union Bank Credit Union Larger size institution plus minus plus More concentrated market* Institution has larger market share plus minus Higher cost of funds plus plus plus plus Ease of CU expansion in state** minus minus plus Speed of rate adjustment faster slower very slow slowest Belongs to a top ten bank holding company plus n/a n/a 3rd Quarter New Vehicle minus 4th Quarter minus minus plus Unsecured Larger amounts of random variation*** higher lower higher lower * A more concentrated market is one dominated by a few, large non-credit union institutions. ** Measured by potential members as a percentage of the adult population in the state where the institution is located. *** Random variation is variation that is not explained by the variables in the market. Non-credit card, unsecured loans. As measured by deposits. 4

11 FINDINGS In the far left column of Table 1.1 we list changes in variables that might influence loan rates. On the right we show the statistical findings on whether the influence raises or lowers loan rates (indicated by plus or minus ). For each of the two types of loans, we show the results for banks and for credit unions. We first discuss influences on new vehicle loans. New Vehicle Loans In principal, larger institutions 1 can either raise or lower rates. To the extent larger size creates economies of scale and reduces cost, this would tend to lower rates. However, to the extent larger size creates pricing power in a market, this could raise rates. Table 1.1 indicates that the latter effect is stronger for bank pricing, with larger size institutions having higher prices, other things equal. By contrast, credit unions appear not to use size to raise prices. The results suggest that economies of scale at larger credit unions tend to lower rates. More Concentrated Market A more concentrated market refers to one that is dominated by a few large firms. This could reduce prices if the presence of large institutions lowered costs through economies of scale. At the same time, a few large firms are more likely to have more market power, which would raise prices. Table 1.1 indicates that where the market is dominated by a few large non-credit union institutions, banks charge higher rates. However, a more concentrated market has no effect on credit union pricing, which is consistent with objectives that differ from those of banks. Larger Institutional Market Share This variable is the market share of the particular bank or credit union. 2 This is highly correlated with the size of the individual institution. After controlling for size and overall market concentration (the market share of the largest three non-credit union institutions), we see no additional effects on price from the market share of either an individual bank or credit union. 1 As measured by deposits. 2 Based on deposits. 5

12 Higher Cost of Funds A higher cost of funds raises rates at both banks and credit unions, as expected. Ease of Credit Union Expansion in State Potential entry into a market can restrain the exercise of market power (the power to influence prices). As a measure of potential entry or ease of expansion by credit unions, we use potential members as a percentage of the adult population in the state where the institution is located. We find that ease of credit union expansion lowers rates at both banks and credit unions. Speed of Rate Adjustment In a model of pricing, we need to control for the fact that some types of institutions may adjust their rates at different speeds as market conditions change. We expect banks and credit unions to differ somewhat in the speed at which they adjust their rates because credit union boards are often consulted on rate changes. Our results suggest that credit unions adjust their rates more slowly than banks. Large Bank Holding Companies The presence in a market of banks that belong to large bank holding companies could have two effects. To the extent that access to greater resources could lower costs, this also could lower rates. However, since bank holding companies have much contact with one another in many markets, they can learn that keeping prices up avoids the possibility of price wars. As Table 1.1 indicates, we find that banks that belong to a top ten bank holding company have higher rates. Seasonal Effects A model of pricing also needs to control for seasonal effects. We find that at banks, loan rates for new vehicles dip in the fourth quarter. Perhaps this is related to the tendency for auto sales to diminish in the fourth quarter, although the sale of less durable goods rises in this quarter. We find that credit unions tend to lower rates on new vehicles in both the third and fourth quarters. 6

13 Degree of Random Variation The statistical results of pricing models also indicate how much of the price variation is explained by the variables in the model and how much of the variation remains unexplained. Less variation explained means that more unidentifiable random effects influence price changes. As Table 1.1 indicates, we find a larger random component in pricing at banks than at credit unions. 3 This suggests that in addition to the variables in our model, banks are more sensitive to other market conditions. This is consistent with their expected objective of maximizing profits. Unsecured Loans Table 1.1 also shows the results for rates on unsecured (non-credit card) loans. As expected, we find these rates less sensitive than rates on new vehicle loans in response to changes in the explanatory variables. Recall that empty cells in Table 1.1 represent relationships that were not statistically significant. Credit union rates on unsecured loans show a statistically significant relationship only to the cost of funds, which also influences bank pricing. Banks show a statistically significant response to more variables, but the patterns are mixed and seem somewhat inconsistent. However, we do see an interesting seasonal pattern. Banks charge more on unsecured loans in the fourth quarter, and less for new vehicle loans in this quarter. This suggests that banks respond to market demand, which for new car loans diminishes in the fourth quarter, while demand for unsecured loans generally rises in this quarter. THE EFFECT OF INTERNAL VARIABLES ON LOAN RATES Internal Variables We also evaluate the extent to which internal variables influence loan rates at credit unions. We investigate whether credit unions that were capital constrained charge higher loan rates. This constraint this can be viewed as a lower capital/asset ratio, or as a 3 In subsequent chapters the adjusted R 2 statistic is given for each regression, which indicates the variation explained in the model. 7

14 higher savings/asset ratio. 4 Concern about potential capital constraints could result from a higher growth rate in assets or a lower growth rate in net income. Adding internal variables to the model also alters the measured effect of other variables. New Vehicle Loans Table 1.2 shows the results of adding internal variables to credit union rates. With this model, for new vehicle loans all the variables previously used are statistically significant. However, now a more concentrated market and a larger market share for the credit union both appear to reflect economies of scale, which lowers rates. Of the four internal variables added that reflect possible effects of capital constraints, three operate in the expected direction and are statistically significant. With this model we still see a dip in rates in the third quarter, but also a dip in the second rather than the fourth quarter. This may occur because potential capital constraints induce credit unions to reduce new vehicle rates to gain loans in the second quarter when savings inflows are typically high. Unsecured Loans Internal variables dominate the explanation of rates on unsecured loans in this model. Only one of the original variables is statistically significant, whereas three out of four of the internal variables are statistically significant. Credit unions with less capital charge higher rates on both unsecured and new vehicle loans. However, while a higher growth rate in assets did not significantly influence new vehicle loan rates, we do see a significant relationship of lower rates on unsecured loans, perhaps to increase loan volume as savings grow. The growth rate of net income significantly influences both new vehicle and unsecured loan rates, but in opposite directions. Lower growth in net income means higher new vehicle rates, but lower rates on unsecured loans. This may suggest that the motives that drive rates on new vehicle loans and unsecured loans differ. This would be consistent with their notably different speeds of adjustment. Additional research focused on lagged effects of explanatory variables would shed more light on this issue. 4 Since savings plus capital equals virtually all of liabilities and equity, savings is approximately equal to assets less capital. 8

15 Table 1.2 Statistical Results (Empty cells represent relationships that were not statistically significant) Variable Original Variables from Table 1.1 Larger size institution More concentrated market* Institution has larger market share Higher cost of funds Ease of CU expansion in state** Speed of rate adjustment Internal Variables Lower capital/asset ratio Higher savings/asset ratio Higher growth rate of assets Lower growth rate in net income Seasonal Variables 2nd Quarter 3rd Quarter 4th Quarter Larger amounts of random variation*** Effect on Loan Rates at Credit Unions New Vehicle minus minus minus plus minus slow plus plus plus minus minus smaller Unsecured slower plus minus minus minus larger * The market is dominated by a few, large non-credit union institutions. ** Measured by potential members as a percentage of the adult population in the state where the institution is located. *** Random variation is variation that is not explained by the variables in the market. 9

16 10

17 CHAPTER 2: Bank Pricing Behavior TERMINOLOGY In what follows, the term loans includes both new vehicle and unsecured (non-credit card) loans, except where we specify one or the other. The term banks excludes thrifts unless we specify otherwise. DETERMINING THE AVERAGE LEVEL OF BANK RATES IN EACH MARKET Our first model examines the determination of the average level of bank loan rates in each market. THE MODEL The explanatory variables in determination of loan rates in the model are: 1. The deposit share of the top two non-credit union institutions. This is based on extensive literature that indicates when a small number (two or three) firms dominate a market, prices tend to be higher. The symbol we use for this variable is CR2. The letters CR refer to concentration ratio. A concentrated market is one in which a few large firms dominate the market. 2. The share of deposits in the market held by all credit unions. Since credit unions are not-for-profit institutions and want to keep loan rates to their members low, we expect that a greater credit union presence in the market would lower average loan rates at banks. The symbol is CU. 3. The size of deposits at the largest depository institution in the market. We expect the presence of the largest institution to reduce loan rates because economies of scale would reduce its cost, thereby putting downward pressure on prices of all banks in a market. The symbol is TOPDEPOSIT. 11

18 4. The federal funds rate in the prior quarter. We expect a higher federal funds rate (which will vary with time) to lead to higher rates about one-quarter later. 5 The symbol is FEDFUNDS t A dummy variable with a value of 0 or 1. The value is 1 for markets in states in which more than 25% of adults are credit union members, and otherwise is 0. We expect states with a credit union presence above a threshold produce an environment in which the supply of services of credit unions is more elastic, in response to situations in which banks raise loan rates significantly above competitive levels. This could happen in highly concentrated markets. The symbol for this dummy variable is HISTATE. 6. We also include dummy variables to represent the quarters of the year to represent seasonal effects on loan rates. 6 Although in all cases above we anticipate these variables would affect new vehicle and unsecured rates in the same direction, we recognize that unsecured rates move considerably less than new vehicle rates in response to market conditions. Therefore, we expect the influence of the variables in the model to be stronger on new vehicle rates and weaker on unsecured rates. The model also controls for possible unknown market-specific influences that do not vary systematically over time. 7 5 Empirically, loan rates adjust to market rates with some lag. 6 We use dummy variables for the second through fourth quarters to indicate influence in comparison with the first quarter. 7 We use a fixed effects model with a constant term in the regression equation to pick up market-specific influences. 12

19 Table 2.1 Regression Results Explaining Average Rates in Markets for Bank Loans FEDFUNDS t-1 CR2 CU TOPDEPOSITS HISTATE QTR2 QTR3 QTR4 observations R 2 (t-statistics are in parentheses) All variables in natural logs Explanatory Variable Dependent Variable Unsecured New Vehicle 0.103*** (8.09) (0.62) ** (2.13) *** (3.71) ** (2.49) (0.66) (1.50) (0.57) *** (28.13) 0.084** (2.16) *** (5.74) *** (8.61) *** (2.91) (0.36) ** (2.45) *** (3.44) *, **, and *** mean significant at the 10%, 5%, and 1% levels, respectively. Market fixed effects (market-specific influences) are not presented here. RESULTS Table 2.1 shows the results of the regression using the model and data described above. For the model of new vehicle loan rates, all of the first five variables listed influence rates in the expected direction and all are statistically significant. The quarterly dummy variables suggest some limited seasonal variation in vehicle loan rates; they dip slightly in the third and fourth quarters of the year. This may be related to the introduction of new model year autos. For the model of unsecured (non-credit card) loans, we find that all five variables operate in the expected direction and that all but CR2 are statistically significant. However, unlike new vehicle loan rates we find no seasonal influences on unsecured loan rates. 13

20 DISCUSSION The type of model used above to determine the average price in the market across a number of providers has been used many times in studies of how market structure may influence average prices. 8 Ideally, however, analysis of price determination explains not the average price across providers in the market, but price determination at each individual provider. Such models of price determination at individual institutions are ideal but much less common, because they require data at a level of detail that often is unavailable. In our analysis we did obtain such data. Next we present the model that explains prices at individual banks. We will then compare the results with the results of the more limited model of average rates presented in the previous section. DETERMINATION OF RATES AT INDIVIDUAL BANKS The first two variables we use in the model to explain loan rates at individual banks are the same as those used for the previous model (of average rates in local markets), and are used for the same reason. The third variable used in the previous model does not appear in the new model. This variable is the size of deposits at the largest depository institution in the market, which we used as a proxy for the effects of economies of scale on average prices in the market. Now that we wish to model how prices are determined at each individual bank, we use the size of the deposits at each bank, which is an indicator of economies of scale for that particular bank. This is our third variable in the new model, and we represent it with the symbol BANKSIZE. The fourth and fifth variables in the model are the same as in the previous model and are present for the same reasons. The new model has two new variables. One is the share of the market of each individual bank. We include this because an individual bank may have some market power based on a sizeable market share of its own. We expect that with a larger market share a bank may have enough market power to charge higher loan rates. The symbol is BANKSHR. 8 Market structure refers to the number of firms in a market and the market share of each. 14

21 The other new variable is a dummy variable which takes on a value of 0 or 1, where a 1 indicates that it belongs to one of the largest 10 bank holding companies in the country. The inclusion of this variable requires some additional explanation. Large bank holding companies compete with each other in many markets. This frequent contact could lead to a policy of mutual forbearance, which means they have learned to avoid price wars. We would expect a bank belonging to one of these holding companies to have higher loan rates than other banks. The symbol for this variable is TOP10BHC. As with the previous model, we also include quarterly dummy variables to pick up any seasonal effects on loan rates at individual banks, and we control for possible market specific influences that do not vary systematically over time. 15

22 Table 2.2 Regression Results Explaining Rates at Individual Banks FEDFUNDS t-1 CR2 CU BANKSIZE HISTATE BANKSHR TOP10BHC QTR2 QTR3 QTR4 observations R 2 (t-statistics are in parentheses) All variables in natural logs Explanatory Variable Dependent Variable Unsecured New Vehicle 0.083*** (6.09) * (1.60) (1.28) *** (2.68) *** (2.82) 0.114*** (3.08) 0.063** (2.37) (0.36) (1.41) (0.33) *** (35.33) *** (5.83) *** (4.97) *** (5.11) *** (4.25) 0.044* (1.89) (0.12) (0.49) (3.01) *** (4.20) *, **, and *** mean significant at the 10%, 5%, and 1% levels, respectively. 16

23 RESULTS The results using this model are shown in Table 2.2. New Vehicle Loan Rates First we consider the results for new vehicle loans. Recall that four of the first five variables in the model are the same as in the previous model. They are the market share of the top two banks in the market, the share of all credit union deposits in the market, the federal funds rate in the prior quarter, and whether the market is located in a state in which more than 25% of adults are credit union members. All of these variables continue to operate in the same direction as in the previous model, and all are statistically significant. In this new model we take into account effects on loan rates from economies of scale by including the size of each bank s deposits. As expected, the results show that larger banks tend to have lower loan rates, and this effect is statistically significant. The two new effects we include in this model of rates at individual banks are the share of the market of each individual bank and whether it belongs to one of the top 10 bank holding companies. We find that, as expected, a bank with a larger market share has enough market power to charge higher new vehicle loan rates, but this effect is only weakly significant (at the 10% level). We find that belonging to a top 10 bank holding company is not associated with significantly higher new vehicle loan rates. As in the previous model, new vehicle rates exhibit some seasonality and dip slightly in the third and fourth quarters. Unsecured Loan Rates For rates on unsecured loans, we first consider effects that are the same as in the previous model: The effect of the federal funds rate in the previous quarter is as expected and is statistically significant. The CR2 variable is not statistically significant in either model. 17

24 The proxies for economies of scale in both models have the expected influence and are statistically significant. Being in a state where more than 25% of adults are credit union members has the expected influence and is statistically significant in both models. In neither model do we see statistically significant quarterly variations in unsecured loan rates. Next we consider results that differ between the two models. The most notable change is that in the first model the market share of all credit unions in a local market reduces average unsecured loan rates among banks in that market. In the second model, the market share of all credit unions in a local market operates in the same direction on individual bank rates, but the effect is no longer statistically significant. What is statistically significant is a new variable, which is the market share of the individual bank. We find that as the market share of an individual bank is larger, the unsecured loan rate rises. Finally, we find that belonging to one of the top 10 bank holding companies raises the unsecured loan rate. 18

25 CHAPTER 3 Credit Union and Bank Pricing Behavior: A Comparison INTRODUCTION In this chapter we first develop a model of pricing behavior for individual institutions which we can apply to both banks and credit unions. Then we use the model to evaluate differences in pricing behavior between the two types of institutions. We would expect them to differ, because the cooperative ownership and governance structure of credit unions generates objectives that differ from those of banks. THE MODEL The model developed for this chapter contains eight variables, plus quarterly dummy variables, to explain rates on unsecured (non-credit card) loans and new vehicle loans. We estimate the model from a combined sample of banks and credit unions. The variables in the model are as follows: 1. Deposit size Total deposits of the institution within a local market. This is a proxy for economies of scale. 2. Market concentration We use the percentage of the market held by the top three non-credit union institutions as a measure of potential market power by large banks in the market. 3. Market share Market share of the institution. 4. Cost of funds An average of the federal funds rate and a measure of credit union cost of funds, lagged one quarter. 5. State credit union share Potential credit union members as a percentage of the adult population in the state. 6. Lag rate A measure of how fast loan rates adjust to market conditions. 7. CU dummy variable The value is 0 for an institution that is a bank and 1 if it is a credit union. 8. Bank holding company dummy variable The value is 0 or 1, with 1 meaning the institution belongs to a top ten bank holding company. 19

26 9-11. Dummy variables for the second through fourth quarters, to pick up seasonal variations. RESULTS The results are shown in Table 3.1. The two versions of the model differ in just one respect. In both versions, we use the cost of funds lagged one quarter. However, in version two we use a model which also includes the previous quarter s loan rate (a two-quarter lag), which we call the lag rate. 9 We find that institution deposit size has no statistically significant effect on unsecured loan rates. However, it does have a significant effect on new vehicle rates, with larger size leading to lower rates, which suggests economies of scale. Market concentration increases new vehicle rates significantly, but this effect is only weakly significant in the rate adjustment version of the model. Institutional market shares are significant for neither type of loan. This could be the result of the large number of credit unions in the sample, since they would have no incentive to translate greater market power into higher rates. 10 As expected, a higher cost of funds raises rates for both types of loans, whether the lag rate variable is included or not. A higher credit union share in the state in which the market is located has no significant effect on unsecured rates but does reduce new vehicle rates. This result is consistent with the threat of potential competition from credit unions helping to discipline pricing for new vehicle loans. The variable measuring the speed of adjustment for rates indicates that prices for both unsecured loans and new vehicle loans exhibit a good deal of inertia in adjusting to changes in the market. Of the two types of loans, the rate of adjustment is much slower for unsecured loans. 9 This allows us to measure (indirectly) the speed at which the institution adjusts its rate to changes in market conditions. The appendix provides the details of the specification. 10 In general, credit unions prefer to keep their loan rates as low as possible for the benefit of their members. 20

27 Banks that belong to one of the top ten bank holding companies charge higher rates on both unsecured loans and new vehicle loans. We find that new vehicle loans tend to be lowest in the third quarter. The dummy variable that denotes whether an institution is a credit union or a bank is of special importance. This allows us to answer the following question: If two depository institutions of the same deposit size in a particular market have the same cost of funds, the same share of the market, and the same market structure conditions, 11 does the pricing behavior of the two institutions differ if one is a credit union and the other is a bank? For new vehicle loans, the answer is clearly yes, with credit unions offering lower loan rates. The picture is somewhat mixed with unsecured loan rates, which in general tend to exhibit low sensitivity to market conditions. For unsecured rates we do not see a difference at credit unions in Version 1 of the model, which does not measure the speed of adjustment in rates. We do see a tendency toward lower rates at credit unions in Version 2, which measures speed of adjustment, although the coefficient is small and is only weakly significant Recall that market structure refers to the number of competing firms in the market and the market share of each. 12 The t-statistic is

28 Table 3.1: Regression Results (combined sample) 13 (t-statistics in parentheses) All variables in natural logs Version 1 Version 2 Variables Unsecured New Vehicle Unsecured New Vehicle Deposit size *** * (1.51) (-2.68) (1.52) (-1.80) Market Concentration *** * (Top 3) (-1.27) (3.34) (-1.69) (1.58) Market Share * (-1.20) (-0.34) (-1.65) (-0.92) Cost of Funds *** *** *** *** (5.54) (42.44) (5.01) (23.03) State CU Share *** ** (0.44) (-6.75) (0.69) (-2.01) Lag Rate n/a n/a *** *** (241.88) (169.56) CU Dummy *** *** (0.61) (-8.40) (-1.55) (-10.89) BHC Dummy *** *** ** *** (4.58) (3.92) (2.20) (4.42) QD *** *** (-1.26) (-3.26) (-0.42) (-2.70) QD ** *** *** (-2.40) (-6.01) (-0.48) (-3.36) QD *** * *** (0.57) (-2.66) (1.84) (-4.51) Adjusted R DW D Statistics D h Statistics n/a n/a No. of Cross Sections Observations *, **, and *** mean significant at the 10%, 5%, and 1% levels, respectively. 13 We use a random effects model instead of a fixed effects model for this estimation, since testing indicated that this was more appropriate. 22

29 MODELING THE DIFFERENCES IN BANK AND CREDIT UNION PRICING BEHAVIOR While the CU dummy variable identifies the effect of credit unions on loan pricing after controlling for other influences, this approach implicitly restricts this credit union effect to changing the intercept term of the loan pricing equations. Tables 3.2 and 3.3 provide results for a more general alternative in which we estimate the pricing equation separately for banks and credit unions, rather than with the combined sample which produced Table Considering the relative size of the coefficients in the regression reported in Tables 3.2 and 3.3, we can summarize the results as follows: Larger banks charge higher interest rates than other banks on both types of loans. In contrast, larger credit unions, compared to other credit unions, do not charge a significantly different rate on unsecured loans, and they charge a lower rate on new vehicle loans. This suggests that for banks, the effect of market power of a larger size institution is stronger than any cost reduction from scale economies. 15 For credit unions, scale economies appear to lower their loan rates. Banks in concentrated markets charge higher rates for new vehicle loans than banks in other markets, although banks with higher market shares charge less on unsecured loans than do other banks. However, neither market share nor market concentration affects credit union pricing. Both banks and credit unions in states with a higher credit union share charge less for loans on new vehicles, and the reduction in rates from this effect is twice as large for credit unions as for banks. However, banks in states with a larger credit union market share charge more for unsecured loans. This is puzzling unless banks find unsecured loans relatively unattractive assets relative to, for example, credit card loans. 14 In estimating the model to produce tables 4 and 5, we eliminate the dummy variable that indicates whether an institution is a credit union or not, since table 4 is for all credit unions and table 5 is for all banks. Also in table 4 we eliminate the dummy variable indicating whether an institution belongs to one of the top ten bank holding companies, since this does not apply to credit unions. 15 An alternative explanation is that bank size is correlated with the size of the market. A market with a large population may have higher operating expenses which contribute to higher loan rates. 23

30 A higher cost of funds raises both bank and credit union rates on both new vehicle and unsecured loans. However, banks appear to respond somewhat more to cost of funds in pricing new vehicle loans than do credit unions. Compared to other banks, those which belong to one of the top ten bank holding companies charge notably higher rates on vehicle loans, but membership in one of these holding companies has no significant effect on rates for unsecured loans. This suggests that unsecured loans may be less attractive assets to banks than other types of loans. 16 Credit unions show substantially greater inertia than banks in adjusting rates for both types of loans when market conditions change. CONCLUSIONS When we pool banks and credit unions into a common data base to estimate a model we find that, other things equal, the fact that an institution is a credit union generates pricing behavior different from that of banks. When we separate the data into two parts, one for credit unions only and one for banks only, we see in more detail how pricing behavior differs. We use a similar model for credit unions and banks to avoid the possibility that differences in models might cause spurious differences in results between the two types of institutions. Having established clear evidence that with the same model we find significant differences in pricing behavior between banks and credit unions, further research could examine in more detail the reasons behind these differences. We address this issue in the following chapter by investigating how internal variables at credit unions affect their pricing behavior. In addition to finding that credit unions and banks exhibit different pricing behavior, we also find that pricing behavior for both types of institutions differs between new vehicle and unsecured loans. While these differences probably stem from differences in the profitability of these two types of loans, future research could illuminate this issue by developing different models for rates on different types of loans. 16 This is possible because making an unsecured loan via a credit card is much more efficient. 24

31 Table 3.2: Regression Results (bank sample) 17 (t-statistics in parentheses) All variables in natural logs Variables USL NVL Deposit size ** *** (2.30) (4.53) Market Concentration (Top 3) *** -(1.50) (3.23) Market Share ** (-2.13) (1.53) Cost of Funds * *** (1.85) (12.92) State CU Share ** ** (3.74) (-2.02) Lag Rate *** *** (127.69) (76.35) BHC Dummy *** (1.36) (4.58) QD (0.71) (-1.56) QD (-0.36) (-1.09) QD ** ** (2.49) (-2.50) Adjusted R DW D Statistics D h Statistics Nos. Of Cross Section Nos. of Observations *, **, and *** mean significant at the 10%, 5%, and 1% levels, respectively. 17 We used a random effects rather than a fixed effects model to estimate this regression. 25

32 Table 3.3: Regression Results (credit union sample) 18 (t-statistics in parentheses) All variables in natural logs Variables USL NVL Deposit size *** (1.34) (-2.58) Market Concentration (Top 3) (-0.26) (-1.05) Market Share (-1.16) (-0.30) Cost of Funds *** *** (8.87) (21.58) State CU Share *** * (-1.06) (-1.75) Lag Rate *** (296.72) (193.85) QD (-0.85) (-1.17) QD *** (-0.69) (-2.64) QD ** (-0.20) (-2.33) Adjusted R DW D Statistics D h Statistics Nos. Of Cross Section Nos. of Observations *, **, and *** mean significant at the 10%, 5%, and 1% levels, respectively. 18 We used a random effects model rather than a fixed effects model to estimate this regression. 26

33 CHAPTER 4: Internal Influences on Credit Union Pricing INTRODUCTION We have seen in the previous chapter that credit unions and banks differ in their pricing behavior. In this chapter we explore in more detail the pricing behavior of credit unions by investigating the influence of internal variables. INTERNAL VARIABLES The following internal variables influence credit union pricing: capital/asset ratio; savings/asset ratio; growth rate of assets; and growth rate of net income. To see why we expect these variables to influence pricing, consider a credit union with low capital, low profitability, but high asset growth rate. This situation could constrain growth through inability to meet minimum regulatory requirements for capital relative to assets. In such a case, the credit union may wish to increase loan rates in order to increase retained earnings, since by regulation it has no other source of capital. A for-profit firm generally might raise prices whenever this would increase earnings. However, as not-for-profit institutions, credit unions generally wish to keep loan rates low to benefit members, except when capital requirements generate a need for higher earnings. A second reason that internal variables could affect credit union pricing is that they pay no explicit dividends on capital. Therefore, the cost of funds is zero for capital, and less capital raises the overall cost of funds. This higher cost of funds could in turn lead to higher loan rates. MODELING We begin with the seven variables used in the last model of chapter 3, plus the quarterly dummy variables. To these we add the four internal variables specified in the previous section. We have discussed how the capital to asset ratio could influence pricing. We expect the influence of the other three internal variables to operate as follows: The savings to asset ratio is an alternative way to measure capital to assets. Savings plus capital is almost equal to total liabilities and equity and therefore almost equal to assets. 27

34 Therefore, savings to assets is inversely related to capital to assets. As a result, if higher capital to asset ratios lowered loan rates, we would expect higher savings to asset ratios to raise loan rates. A higher growth rate of assets would reduce the capital/asset ratio, which we would expect to raise loan rates. A higher growth rate of net income would raise the capital to asset ratio, which we would expect to lower loan rates. RESULTS Results are shown in Table 4.1. Asterisks indicate statistically significant variables. Variables are grouped by: 1) original variables from the previous chapter, 2) internal variables, and 3) the quarterly dummy variables used before. The new internal variables often significantly influence rates on both types of loans. In addition, the presence of these new variables in the pricing model influences the effects of the original variables from the previous chapter that are in the model. Therefore, we will discuss the results for each of the variables in model represented in Table

35 Table 4.1: Explaining Rates at Individual Credit Unions Including Internal Variables 19 (t-statistics in parentheses) All variables in natural logs Variable Unsecured loans New Vehicle loans Original Variables (from Table 3.3) Deposit size 3.73E-10 (0.79) -1.21E-09*** (-2.66) Market concentration (top 2 banks) (-1.47) ** (-2.38) Market share (-0.72) ** (-2.04) Cost of funds (1.36) *** (15.12) State CU Share (-1.02) *** (-3.17) Lag rate *** (16.70) ** (2.02) Internal Variables Capital/asset ratio *** (-2.91) *** (-2.71) Savings/asset ratio (-0.03) *** (3.52) Growth rate of assets *** (-2.90) (0.03) Growth rate in net income *** (10.15) -9.45E-05** (-2.17) Quarterly Dummy Variables Quarter (0.56) *** (-2.96) Quarter ** (-2.16) *** (-7.47) Quarter (-0.13) (-0.76) Rho (autocorrelation adjustment term) *** (2.96) *** (20.83) Adj. R Durbin Watson D Durbin H Number of cross sections Total panel observations 4,725 4,782 *, **, and *** mean significant at the 10%, 5%, and 1% levels, respectively. E means exponential. +Rho is not an explanatory variable; it is included to make a necessary statistical adjustment. 19 We used a fixed effects model to estimate this regression. 29

36 PREVIOUS VARIABLES VERSUS INTERNAL VARIABLES The new variables dominate as the statistically significant explanatory variables for unsecured loans to such an extent that three of the four new variables are statistically significant but none of the original variables are statistically significant, except for the third quarter dummy variable. Three of the four new variables are also statistically significant in explaining new vehicle rates. However, for new vehicle rates all of the previous seven variables are statistically significant. We discuss the likely reasons for these patterns below. ORIGINAL VARIABLES Since none of the original variables (other than a quarterly dummy variable) are statistically significant in explaining unsecured rates in this model, we focus on the effect of the original variables on rates for new vehicle loans. INFLUENCES ON NEW VEHICLE LOANS Deposit Size From other sources we know that in general, as credit union asset size increases, the ratio of operating expense to assets declines. 20 Therefore, in our model we would expect a larger deposit size to reduce costs and thereby lower loan rates. The results in Table 4.1 show this effect and it is statistically significant. Bank Concentration Greater bank concentration in the market could produce two effects. It could allow banks to increase prices, which would in turn allow credit unions to raise prices. Greater bank concentration could also represent greater economies of scale in the market, which would lower prices. Since credit unions have objectives other than profit, they do not necessarily raise prices when banks with market power raise theirs. However, if the higher concentration ratio for banks generates economies of scale in the market, this could lower bank prices, which credit unions likely 20 For the relationship between size and the ratio of operating expenses to assets in credit unions, see Operating Ratios and Spreads, Department of Economics and Statistics, CUNA and Affiliates, Madison, WI, 2001, p