The determinants of bank rates in local consumer lending markets: comparing market and institution-level results.

Feinberg, Robert M.

1. Introduction

With the exception of a few very recent studies, the sizeable literature on the impact of market structure in banking markets (1) has ignored the potential competitive role played by credit unions. This is surprising, since the continuing consolidation of the financial services industry in recent years has naturally raised concerns about competitive effects and credit unions would seem to be a likely source of market discipline. (2) This paper explores these issues using both market and firm data.

Using a variety of approaches, economists have started paying some attention to the interaction between banks and credit unions. Emmons and Schmid (2000), using county-level data, examine two-way intertemporal linkages between credit union participation rates and market concentration of the commercial banking sector to support the view that the two types of institutions compete in the market for consumer deposits; similarly, Feinberg and Rabman (2001) find that credit union and bank rates for two consumer loan products can each be shown to be influenced by the other. Tokle and Tokle (2000) found a competitive influence of credit unions on bank certificate of deposit (CD) rates offered in Idaho and Montana, while Feinberg (2001) explored their impact on consumer loan rates in a broader sample of relatively small local markets (3) over the 1992-1997 period.

The latter article found that both unsecured and new vehicle loan rates offered by banks in these markets were affected in a significant manner by the market share held by the two leading banks (implying a competitive role for smaller financial institutions generally) and--for new vehicle loan rates alone--by the share of credit unions in those markets. Whether these results would generalize to larger metropolitan areas is unclear and is the topic of what follows. I focus on the same two loan products, 24-month unsecured (non-credit card) loans and 48-month new vehicle loans, both of which seem likely to be provided in a local market, and empirically explain the determinants of bank loan rates in a sample of 56 markets--both large and small--for the period 1992-1998. An innovative aspect of this paper is that I analyze both market-level data and institution-level data for 81 banks in those 56 markets and consider as well the role of multimarket linkages affecting these banks.

2. Theoretical Framework

Virtually all models of imperfect competition imply that increasing entry and supply from fringe suppliers will lower prices. This clearly suggests that an increasing credit union presence should discipline prices in local financial services markets. To formalize, I employ a modified version of the dominant firm-price leadership model. The modification involves the notion that while credit unions may generally be thought of as fringe suppliers, not all banks and savings and loans (S&Ls) would realistically constitute a dominant group. Nevertheless as a group we can think of banks and S&Ls as being relatively dominant, with the degree to which a monopoly position over their residual demand (i.e., netting out credit unions) is exploited depending on how concentrated bank and S&L deposits are in the leading two institutions. (4)

The number of major firms is often quite small in local consumer lending markets, and there is a genuine concern that in the absence of pressure from a "competitive fringe" these leading firms may be able to act in a collusive manner. In the spirit of Saving (1970), I assume a homogeneous product, with market demand for loans Q = D(P). Credit unions are treated as a price-taking fringe, with their supply [S.sup.CU](P) and the demand faced by banks and S&Ls, [D.sup.B](P), a residual:

[D.sup.B](P) = D(P) - [S.sup.CU](P). (1)

Taking first derivatives with respect to price, multiplying all terms by (P/Q), and multiplying the lefthand side expression by ([D.sup.B](P)/[D.sup.B](P)) and the last term on the right-hand side by ([S.sup.CU](P)/[S.sup.CU](P)), I obtain

[PD.sup.B]'(P)[D.sup.B](P)/[D.sup.B](P)D(P) = D'(P)P/D(P) - [S.sup.CU]'(P)[PS.sup.CU](P)/[S.sup.CU](P)D(P) (2)

or, simplifying, and defining CU to be equal to [S.sup.CU](P)/D(P), the credit union market share, and then dividing through by 1 - CU, I obtain an expression in terms of price elasticities of demand and supply,

\[[eta].sub.B]\ = \[eta]\/(1-CU) + [[epsilon].sub.CU] (CU/(1-CU)). (3)

If I assumed that all non-credit union institutions acted to jointly maximize profits, I would of course find that their Lemer Index, LI = (P - MC)/P = 1/\[[eta].sub.B]\ . More realistically, I parameterize the extent to which the Lemer Index approaches this value as a function of dominance within the bank and S&L group of the two leading firms (to simplify, a linear function) so that

LI = [theta]/\[[eta].sub.B]\, (4)

where [theta] = kCR2/(1 - CU), CR2 = the share of the two largest financial institutions in total market deposits (including credit unions), k is a constant, and 0 < [theta] < 1.

Substituting Equation 3 into Equation 4, I obtain the bank/thrift Lemer Index to be

LI = kCR2/\[eta]\ + [[epsilon].sub.CU]CU. (5)

Without any further mathematical analysis, the following implications clearly emerge: (i) As CR2 increases the LI increases as well; (ii) as CU increases the LI declines; (iii) as [[epsilon].sub.CU] increases the LI declines; and, not quite as obvious (but easy to derive), (iv) the declines in (ii) and (iii) are larger in absolute value as CR2 is larger. All of the qualitative results above follow when I replace LI with the market price, which is the variable I seek to explain in the empirical work to follow.

While obviously this model is quite ad hoc in nature, the clear implication is that bank concentration, the credit union share, and the supply elasticity of credit unions all matter in determining the exercise of market power in these markets.

3. Data

Data on bank loan rates (5) were obtained (via a Freedom of Information Act request) from the Federal Reserve Board's Quarterly Report of Interest Rates on Selected Direct Consumer installment Loans, and on local market structure from Sheshunoff Information Services, for the 1992-1998 period. The two bank loan rates (reported for the second week of Februaiy, May, August, and November) are for 48-month new auto loans and for 24-month non-credit card unsecured consumer loans, for which banks are requested to report their most common rate. These types of loans are the only ones reported on in the Federal Reserve Board survey.

Markets were included for which data were available on bank rates for both types of loans for at least two-thirds of the 28 quarterly observations from 1992 through 1998. I also deleted six markets in which a single credit union was among the top two depository institutions during this period (as these markets seem inconsistent with the model above in which credit unions are viewed as a competitive fringe)6 and three markets in which local market structure data were distorted by the presence of bank credit card operations. (7) This left 56 markets (listed in Appendix A), defined as metropolitan statistical areas (MSAs) or nonmetropolitan counties. (8) Of these, 52 are MSAs ranging in size (based on 1990 population figures) from Victoria (Texas)--with a population of 74,361--to the New York City consolidated MSA--with almost 20 million people. The other four are rural counties ranging in size from Atchison County (Kansas)--16,932--to Sussex County (Delaware)--1 13,229. The median 1990 population for the 56 mar kets was 674,267.

My initial analysis was performed at the level of the market, with a simple average of loan rates of surveyed banks within that market (or if only one bank was surveyed that bank's loan rate) taken to proxy the market rate. Based on deposit data from banks, thrifts, and credit unions (collected by Sheshunoff Information Services), the share of total market deposits held by credit unions and by the top two depository institutions and (as a proxy for scale economies being exploited in the market) the absolute total of the largest institution deposits in the market were recorded for each market and time period. (9) To proxy the elasticity of supply by credit unions, data were obtained from the Credit Union National Association (CUNA) on state-level credit union membership as a percentage of adult population. The federal funds rate (obtained from Federal Reserve statistics) was employed as a proxy for the cost of funds to banks.

Descriptive statistics for the market-level analysis are given in Table 1, which contains a breakdown by small and big markets, those below or above one million in 1990 population. Clearly, the average big market has a somewhat smaller credit union presence (7.9% vs. 9.1%) and slightly smaller top two bank share (39.6% vs. 41.4%). Not surprisingly, the largest institution is considerably bigger (averaging more than $10 billion vs. just over one billion dollars for the small markets). The average rates are roughly one-quarter (for new vehicle loans) to three-quarters (for unsecured loans) of a percentage point higher in the big markets. In the individual bank analysis to follow, bank deposit market shares and absolute deposit size were included and the loan rate of the particular bank was used; in addition, I examine the impact of participation by banks in leading bank holding companies for performance in particular markets (descriptive statistics for the individual bank sample are presented below).

4. Regression Results

I employ a fixed effects regression model, (10) with coefficient estimates reported in Table 2, explaining (with all variables in natural logs (11)), across both time and markets, bank loan rates (UNSECURED and NEWVEHICLE) by the deposit share of the top two non--credit union institutions (CR2), the total share of deposits held by credit unions (CU), deposit size of largest institution in the market (TOPDEPOSITS), the federal funds rate one quarter lagged (12) (FEDFUNDS), quarterly dummies (to control for seasonal factors), and market-fixed effects. Since the model above indicates a role for the elasticity of supply of credit unions (essentially a measure of the ease of expansion), I attempted to proxy this by the state-level credit union membership penetration ratio (actually a dummy variable, HISTATE, equal to one where this was greater than 25%). (13) While no interaction term between CR2 and CU is included (which would seem to be necessary to test hypothesis (iv) above), a log specification implies that g reater bank concentration will always increase the price-reducing impact of credit unions, as long as a greater credit union share does reduce prices and increased concentration raises prices. (14) No time fixed effects (other than the quarter dummies) were included since I already am including a variable--the federal funds rate--which varies intertemporally but not across markets. (15)

It should be noted that while the dependent variables reflect loan rates on products representing relatively small shares of a typical bank's portfolio, the market structure variables are developed from the bank's deposit position. On the one hand this may be seen as introducing measurement error into the estimation; however, on the other hand, deposit shares may be seen as instruments that deal with the potential simultaneity, often noted in critiques of structure-performance studies, between price or profit rates and market shares. In the current context it is hard to imagine a bank's choice of loan rate on unsecured or auto loans having any significant effect on its deposits.

While initial expectations were for less of an impact of credit unions when larger population centers were included than for the small-market results reported in Feinberg (2001), the results presented here suggest the opposite, at least for unsecured loans--the coefficient of the CU share is twice what it was in the small-market sample and is now statistically significant (which it had not been in the earlier results). In fact, for both equations all estimated coefficients have the expected signs, and all but CR2 in the UNSECURED equation are statistically significant at conventional levels in explaining bank loan rates. Market-fixed effects are important (F-tests reject the hypothesis of all zero coefficients for the market dummy variables), and the quarter dummies suggest some limited seasonality in auto loan rates (slightly lower in the third and fourth quarters of the year, perhaps in connection with the introduction of the new season of cars).

The elasticities of the loan rate with respect to the lagged federal funds rate may seem at first glance to be quite small. However, full passthrough (around mean values of the loan rates and federal funds rate) would imply elasticities of +0.35 for unsecured loans and of +0.53 for new vehicle loans; in this context the estimated elasticities of 0.1 for unsecured loans and 0.3 for new vehicle loans are more reasonable. The impact of the top two bank share (CR2) is relatively small in both equations (0.03 vs. 0.08). The CU effect is only half as large for unsecured loans as for the new vehicle rate (-0.06 vs. -0.12) but is clearly not trivial in magnitude; an increase from the mean CU share (8.68%) to 10% would imply a reduction in the unsecured bank loan rate from its mean of 13.42% to 13.30% and a reduction in the new vehicle bank loan rate from 8.80% to 8.65%. In addition, it is important to note the significant negative impact on rates for both types of bank loans of HISTATE, indicating a spillover effect on local markets of a large CU presence in the state (and, in the context of the model described earlier, possibly capturing the role of greater ease of CU expansion). An argument for scale economies in banking is supported by the statistically significant estimated effect of leading firm absolute size.

While these results suggest a significant role of credit union influence (and market structure more generally) on loan pricing in local markets, the use of market-level observations prevents me from attempting to separate out firm-level from market-level impacts. In what follows, I replicate the above results using data on 81 banks (in the 56 markets examined above) over the quarterly observations from 1992 to 1998. Another feature of this analysis is that I can see whether banks that are part of major bank holding companies determine loan prices differently from other firms; while a negative effect could be explained by efficiencies of multibank operation, there is both theory and previous empirical support for a positive price effect of multimarket operation (both generally and in the banking sector). (16)

Descriptive statistics for the bank-level variables--loan rates as defined earlier, market share (BANKSHR), total deposits (BANKSIZE), and a dummy variable distinguishing banks belonging to one of the top 10 bank holding companies, as of 1996 (TOPlOBHC) (17)--are shown in Table 3. The banks in our sample (listed in Appendix B) have a mean share of 17.1% of market deposits, but this varies from as little as 0.04% to a high of 47.38%. They range in deposit size from $12.2 million to $83.8 billion. About 18% of the banks in our sample were subsidiaries of one of the 10 leading bank holding companies (BHCs). (18)

Table 4 describes changes over time in several of our key variables. Consistent with the wave of bank consolidation occurring during the decade of the 1990s (and continuing to the present), the mean bank market share in our sample rises steadily from 16.1% in 1992 to 18.7% in 1998, while the share of the two leading banks and thrifts rises from 38.9% to 43.0%, and the proportion of our banks associated with the leading bank holding companies rises from 14% to more than 19% over the sample period. (19) Perhaps providing something of a competitive counterweight to this trend, the share of deposits held by credit unions slowly rises from 7.9% in 1992 to 8.3% in 1996 before falling in 1997 and (somewhat dramatically, to 7.5%) in 1998. (20)

Regression estimates are reported in Table 5, explaining (as before, with all variables in natural logs)--in a fixed effects framework--bank loan rates by the bank's deposit share of the market, the bank's total volume of deposits within the market, a dummy variable representing a subsidiary of one of the 10 leading bank holding companies, (21) and variables previously used in the market-level regression--CR2, CU, HISTATE, FEDFUNDS, and quarterly dummy variables.

Not surprisingly, the effects of the federal funds rate, credit union influence variables, and quarterly dummies on bank loan pricing are very similar to their influence in the market-level regression. Of particular interest, however, are the impacts of BANKSHR, BANKSIZE, TOP10BHC, and CR2. For both types of consumer loans, increased market share increases loan rates (although the effect on auto loans is only weakly significant, perhaps reflecting the emergence of competition from national automobile finance companies), while bank size has a strongly significant negative impact on loan pricing--consistent with scale economies playing a role.

After controlling for firm-level effects, the concentration variable now has a negative effect, statistically significant for new vehicle loans; this result is consistent with much of the prior literature. (22) Looking to influences beyond the immediate market, there is a clear indication of higher unsecured loan rates (but not new vehicle rates) in banks that belong to leading bank holding companies and that thus meet each other in multiple markets (73) This result is more consistent with the mutual forebearance view of multimarket contact than with the notion of efficiencies of branching across multiple markets.

5. Summary and Conclusions

Findings like those in Feinberg (2001) showing a positive impact of market concentration on prices (and more often in other studies on profit rates) are often taken to indicate the presence of collusion among few sellers leading to higher market prices. When a comparable analysis is performed at a lower level of aggregation, however, that effect is often seen to be a spurious result of aggregating individual firm market share effects. In fact, as was found in this study, after controlling for individual firm effects the impact of market concentration is often negative; the latter may be the result of scale economies exploited by leading firms and (through competitive pressures) transmitted into lower prices for all firms in a market.

The lack of a positive market concentration effect does not deny the presence of market power in local consumer lending markets. However, this power is more likely the result of unilateral bank behavior and must be considered alongside apparent scale economy benefits of large size. Price-increasing effects can be limited by a significant credit union presence (both actual and potential) and enhanced by multimarket contacts among bank holding companies.

The results presented here point to a strong procompetitive role for credit unions on bank loan pricing. Generally the greater the local market share of credit unions, and the greater the state-level membership penetration, the lower bank loan rates are--both for unsecured and new vehicle loans. In addition, the results support the importance of local markets in the consumer banking sector; differences in market structure defined at the local level clearly produce differences in market and bank loan rates.

Appendix A

List of Markets

1. Birmingham, AL

2. Little Rock, AR

3. Sussex County, DE

4. Fort Pierce/Port St. Luice, FL

5. Macon, GA

6. Des Moines, IA

7. Sioux City, IA

8. Evansville, IN

9. Fort Wayne, IN

10. Atchison County, KS

11. Wichita, KS

12. Louisville, KY

13. Baton Rouge, LA

14. Lafayette, LA

15. Portland, ME

16. Grand Rapids, MI

17. Marquette County, MI

18. Leflore County, MS

19. Billings, MT

20. Fargo, ND

21. Omaha, NE

22. Rena, NV

23. Dayton, OH

24. Mansfield, OH

25. Youngstown, OH

26. Oklahoma City, OK

27. Tulsa, OK

28. Johnstown, PA

29. Amarillo, TX

30. El Paso, TX

31. Victoria, TX

32. Richmond, VA

33. Roanoke, VA

34. Burlington, VT

35. Huntington, WV

36. Hartford, CT

37. Atlanta, GA

38. Chicago, IL

39. Boston, MA

40. Detroit, MI

41. Minneapolis, MN

42. St. Louis, MO

43. Charlotte, NC

44. Buffalo, NY

45. New York, NY

46. Rochester, NY

47. Cincinnati, OH

48. Cleveland, OH

49. Columbus, OH

50. Portland, OR

51. Pittsburgh, PA

52. Providence, RI

53. Memphis, TN

54. Nashville, TN

55. Dallas, TX

56. Milwaukee, WI

Appendix B

Sample of Banks (Names as of 1996) and Their Markets

1. Amsouth Bank ot Alabama, Birmingham

2. Boatmens National Bank, Little Rock

3. Glastonbury Bank and Trust, Hartford

4. First Omni Bank, Sussex County (DE)

5. First NB&TC Treasure Coast, Ft. Pierce (FL)

6. First Union National Bank, Atlanta

7. Suntrust Bank, Atlanta

8. Suntrust Bank Middle Georgia, Macon

9. Norwest Bank Iowa, Des Moines

10. Security National Bank, Sioux City

11. First National Bank, Chicago

12. Northern Trust Company, Chicago

13. LaSalle National Bank, Chicago

14. Aurora National Bank, Chicago

15. Citizens National Bank, Evansville

16. Norwest Bank Indiana, Fort Wayne

17. Exchange NB&TC, Atchison County (KS)

18. Intrust Bank, Wichita

19. National City Bank, Louisville

20. PNC Bank Kentucky, Louisville

21. Hancock Bank of Louisiana, Baton Rouge

22. First National Bank, Lafayette (LA)

23. First National Bank, Boston

24. Fleet Bank New Hampshire, Boston

25. Fleet Bank Maine, Portland

26. Comerica Bank, Detroit

27. NBD Bank, Detroit

28. Michigan National Bank, Detroit

29. Old Kent Bank, Grand Rapids

30. MFC First NB, Marquette County (MI)

31. First Bank, Minneapolis

32. Mercantile Bank, St. Louis

33. Bank of Commerce, Le Flore County (MS)

34. First Interstate Bank, Billings

35. First Union National Bank, Charlotte

36. Norwest Bank ND, Fargo

37. Norwest Bank Nebraska, Omaha

38. Pioneer Citizens Bank, Reno

39. Marine Midland Bank, Buffalo

40. Manufacturers and Traders Bank, Buffalo

41. Bank of New York, New York City

42. Chemical Bank, New York City

43. National Bank, Rochester

44. Star Bank, Cincinnati

45. Fifth Third Bank, Cincinnati

46. PNC Bank, Cincinnati

47. First National Bank of Ohio, Cleveland

48. Bank One Akron, Cleveland

49. National City Bank, Cleveland

50. Premierbank and Trust, Cleveland

51. Huntington National Bank, Columbus

52. National City Bank, Columbus

53. Bank One, Columbus

54. Bank One, Dayton

55. Bank One, Mansfield

56. Mahoning National Bank, Youngstown

57. Liberty Bank and Trust, Oklahoma City

58. Boatmens First National Bank, Oklahoma City

59. Bank of Oklahoma, Tulsa

60. United States National Bank of Oregon, Portland

61. Johnstown Bank and Trust, Johnstown

62. Integra Bank, Pittsburgh

63. PNC Bank, Pittsburgh

64. Mellon Bank, Pittsburgh

65. Rhode Island Hospital Trust NB, Providence

66. Union Planters National Bank, Memphis

67. First American National Bank, Nashville

68. Suntrust Bank, Nashville

69. Amarillo National Bank, Amarillo

70. Bank One Texas, Dallas

71. Sunwest Bank, El Paso

72. First Victoria National Bank, Victoria

73. Crestar Bank, Richmond

74. Signet Bank, Richmond

75. Central Fidelity National Bank, Richmond

76. First Union National Bank, Roanoke

77. First Virginia Bank--Southwest, Roanoke

78. Key Bank of Vermont, Burlington

79. Firstar Bank, Milwaukee

80. Marshall & Jisley Bank, Milwaukee

81. Bank One West Virginia, Huntington

Table 1

Descriptive Statistics (Number of Observations)

 Mean SD Minimum Maximum

56 markets, 28 quarters (1992:1 to
 1998:IV)
 CU (1568) 8.68% 4.77 0 23.34
 CR2 (1568) 40.72% 10.41 16.26 70.08
 TOPDEPOSITS (1568) $4.58B 8.75B 66.07M 83.8.2B
 STATE-CU PENETRATION(1568) 26.64% 8.51 8.60 58.40
 FEDFUNDS (28 quarters) 4.68% 1.04 2.99 6.02
 UNSECURED (1519) 13.42% 2.08 7.00 18.99
 NEWVEHICLE (1515) 8.80% 1.08 4.82 15.00

Small markets (under one million
 1990 population)
 CU (980) 9.12% 5.11 0 23.34
 CR2 (980) 41.42% 10.83 16.26 62.77
 TOPDEPOSITS (980) $1.20B 1.02B 66.07M 5.66B
 STATE-CU PENETRATION (980) 27.39% 9.71 8.60 58.40
 UNSECURED (924) 13.11% 2.18 7.00 18.88
 NEWVEHICLE (921) 8.70% 1.08 4.82 15.00

Big markets (over one million 1990
 population)
 CU (588) 7.93% 4.04 1.44 20.41
 CR2 (588) 39.55% 9.56 18.82 70.08
 TOPDEPOSITS (588) $10.23B 12.32B 1.65B 83.82B
 STATE-CU PENETRATION (588) 25.40% 5.79 16.60 42.40
 UNSECURED (595) 13.91% 1.82 8.50 18.99
 NEWVEHICLE (594) 8.97% 1.06 6.00 14.83

Table 2

Regression Results Explaining Bank Loan Rats--Market-Level Observations
(t Statistics in Parentheses)

 Dependent Variable
 UNSECURED NEWVEHICLE

[FEDFUNDS.sub.t-1] 0.103 0.277
 (8.09) (28.13)
CR2 0.031 0.084
 (0.62) (2.16)
CU -0.056 -0.118
 (2.13) (5.74)
TOPDEPOSITS -0.095 -0.174
 (3.71) (8.61)
HISTATE -0.046 -0.042
 (2.49) (2.91)
QTR II -0.005 -0.002
 (0.66) (0.36)
QTR III -0.012 -0.015
 (1.50) (2.45)
QTR IV -0.005 -0.021
 (0.57) (3.44)
Observations 1465 1460
[R.sup.2] 0.597 0.523

All variables in natural logs.

Market fixed effects are not presented here.

Table 3
Descriptive Statistics, Bank-Level Data (Number of Observations)

 Mean SD Minimum Maximum

81 banks, 28 quarters
(1992:I to 1998:IV)--some
missing observations
UNSECURED (2123) 13.65% 2.45 5.61 24.38
NEWVEHICLE (2110) 8.77% 1.06 5.99 15.00
BANKSHR (2152) 17.14% 8.62 0.04 47.38
BANKSIZE (2152) $4.4B $7.9B $12.2M $83.8B
TOP10BHC (2152) 0.18 0.38 0 1

Table 4

Mean Values by Year, Selected Variables

 BANKSHR CR2 CU TOP10BHC

1992 16.1 38.9 7.9 0.14
1993 16.4 38.3 8.1 0.18
1994 16.6 38.4 8.3 0.19
1995 16.6 38.8 8.2 0.19
1996 17.3 40.4 8.3 0.19
1997 18.5 42.3 8.1 0.18
1998 18.7 43.0 7.5 0.19

Table 5

Regression Results Explaining Bank Loan Rates--Bank-Level Observations
(t Statistics in Parentheses)

 Dependent Variable
 UNSECURED NEWVEHICLE

[FEDFUNDS.sub.t-1] 0.083 0.295
 (6.09) (35.33)
CR2 -0.006 -0.146
 (1.60) (5.83)
CU -0.037 -0.090
 (1.28) (4.97)
BANKSHR 0.114 0.044
 (3.08) (1.89)
BANKSIZE -0.086 -0.103
 (2.68) (5.11)
TOPIOBHC 0.063 0.002
 (2.37) (0.12)
HISTATE -0.049 -0.047
 (2.82) (4.25)
QTR II -0.003 -0.002
 (0.36) (0.49)
QTR III -0.012 -0.0155
 (1.41) (3.01)
QTR IV -0.003 -0.021
 (0.33) (4.20)
Observations 2064 2051
[R.sup.2] 0.528 0.563

All variables in natural logs.

Bank fixed effects are not presented here

Received February 2002; accepted October 2002.

(1.) These include Amel and Liang (1997), Rhoades (1997), Humphrey and Pulley (1997), Simons and Stavins (1998), Berger and Hannan (1989, 1998), Rhoades (1987), and Hannan (1991).

(2.) Related to this is a recent policy debate on the role of credit unions and whether the government should encourage or limit access to these organizations.

(3.) While innovations in technology have led some to the view that all lending markets are national, Simons sad Stavins (1998) present evidence suggesting that banking markets are stilt predominately local in nature. They point to the Federal Reserve Board's 1992 Survey of Consumer Finances, which shows that 94.1% of households using a financial institution identified a local institution as their primary provider of financial services; both deposit accounts and sourees of credit were primarily local in nature. More recently, Amel and Starr-McCluer (2002) have noted (examining more recent versions of that survey) that consumer loans--especially new vehicle loans--became less locally limited during the 1990s.

(4.) Obviously the choice of which concentration ratio to use is somewhat arbitrary (in particular the top two or top three or even the top four share); nevertheless I focus in this paper on the top two firm share. The top four firm share, often used in national-level manufacturing sector studies, was frequently equal or close to 100% (especially for the smallest markets) and did not show as much variation as the top two measure; furthermore, the top two share has been used in other banking studies of local markets. Another rationale for choosing the top two share is found in Kwoka (1979), based on its greater ability (in his empirical study) to explain profit margins in manufacturing.

(5.) These are for commercial banks, both federally and state chartered, and the loan rates are specific to the particular market in question (e.g., NationsBank is not asked for a single loan rate for all markets in which they operate).

(6.) In addition, some of these credit unions were multibranch credit unions with operations far beyond the boundaries of the particular market, and one served essentially as a central bank for credit unions with little impact on the local lending market. Markets excluded for this reason were Albany (NY), Anchorage (AK), Kansas City (MO), Lynchburg (VA), Peoria (IL), and Washington (DC).

(7.) These were Dover (DE), JCPenney; Sioux Falls (SD), Citicorp; and Casper (WY), Norwest.

(8.) Market-fixed effects should pick up the impact of variables (including perhaps per capita income), which may be correlated with market size.

(9.) Since bank loan rates were available quarterly, but market structure variables were only available on an annual basis (reflecting data as of June 30th), some smoothing of the market structure variables was employed; initially the annual observation was repeated for each quarter, and then a three-quarter centered moving average was constructed. As a practical matter this implied that the second and third quarter observations were the actual reported as of June 30th, while the first and fourth quarter observations reflected weighted averages of current and previous or future year figures.

(10.) A Hausman test suggested this was a more appropriate specification than a random effects approach.

(11.) The few zero observations on credit union shares were arbitrarily adjusted to 0.05% in order to take logs.

(12.) Using current quarter FEDFUNDS made very little difference in the results, although the impact of the lagged federal funds rate on all bank and credit union loan rates was consistently larger than the contemporaneous effect.

(13.) The 25% cutoff corresponds roughly to the mean value of the continuous measure. HISTATE performs better than the continuous measure of state-level CU penetration and conforms to our expectation that the relationship should be a discontinuous one reflecting cultural and regulatory factors differing by state that would make future credit union expansion--potential competition in the market--more or less likely (although another interpretation would be that membership penetration could proxy potential future demand for loans). As noted by a referee, the state-level membership penetration is positively correlated with the market-level CU deposit share, making it difficult to separate their impacts on loan rates; HISTATE may be thought of as an instrument allowing the two effects to be better estimated.

(14.) In earlier work, a regression on levels containing an interaction term between CU and CR2 did find the estimated coefficient of this term to be negative as predicted and statistically significant.

(15.) Preliminary attempts to include a demand growth variable (differing by market) failed to find significant effects. To the extent these followed national trends, the federal funds rate would likely capture demand growth; to the extent certain markets had secularly high or low growth over the period, market-fixed effects would capture these effects. Similarly, cross-market income differences would be picked up by fixed effects.

(16.) See Feinberg (1985) and Bernheim and Whinston (1990) and, for banking specifically, Alexander (1985) and Hannan and Prager (2001).

(17.) These were Chase Manhattan, Citicorp, BankAmerica, JP Morgan, Nationsbank, First Union, First Chicago NBD, Bankers Trust, Bank One, and Norwest. Of course, the restriction to 10 BHCs is arbitrary (although encompassing most of the household names in consumer banking); however, virtually all the banks surveyed were part of some bank holding company (some quite small or limited in geographical scope), so some such cutoff was required for this variable to have any meaning.

(18.) I continue to maintain the inclusion criteria of at least 19 observations during the 28 quarters in the sample period.

(19.) While this implies a bank surveyed in one market often shares a common parent with a bank surveyed in another, loan rates reported appear to be determined independently (perhaps more accurately, they are far from perfectly correlated).

(20.) Anecdotally, the drop in the credit union deposit share after 1996 corresponds to a major legal effort by banks to block the growth in credit union membership, which culminated in a U.S. Supreme Court decision in February 1998, ruling that federal credit unions had illegally expanded to include unrelated occupational groups (legislation signed by President Clinton in the summer of 1998, and which took effect in January 1999, largely overturned that decision).

(21.) The list of 10 leading bank holding companies is defined as of 1996. As banks were acquired during the period by these bank holding companies, there is some variation in this dummy variable over time.

(22.) An early finding along these lines was Ravenscraft (1983).

(23.) In preliminary results not reported here, there appear to be some patterns of common BHC effects across markets. Unsecured rates were significantly higher at subsidiaries of Bank of Boston, First Chicago NBD Boatmens, Wells Fargo, Bank of America, and First Union; they were significantly lower at US Bank, PNC, Bank One, and SunTrust.

References

Alexander, Donald L. 1985. An empirical test of the mutual forebearance hypothesis: The case of bank holding companies. Southern Economic Journal 52:122-40.

Amel, Dean F., and J. Nellie Liang. 1997. Determinants of entry and profits in local banking markets. Review of Industrial Organization 12:59-78.

Amel, Dean F., and Martha Starr-McCluer. 2002. Market definition in banking: Recent evidence. Antitrust Bulletin 47:63-89.

Berger, Allen N., and Timothy H. Hannan. 1989. The price-concentration relationship in banking. Review of Economics and Statistics 71:291-9.

Berger, Allen N., and Timothy H. Hannan. 1998. The efficiency cost of market power in the banking industry: A test of the "quiet life" and related hypotheses. Review of Economics and Statistics 80:454-65.

Bernheim, B. Douglas, and Michael D. Whinston. t990. Multimarket contact and collusive behavior. The Rand Journal of Economics 21:1-26.

Emmons. William R., and Frank A. Sebmid. 2000. Bank competition and concentration: Do credit unions matter? Federal Reserve Bank of St. Louis Review 82:29-42.

Feinberg, Robert M. 1985. "Sales-at-risk": A test of the mutual forbearance theory of conglomerate behavior. Journal of Business 58:225-41.

Feinberg, Robert M. 2001. The competitive role of credit unions in small local financial services markets. Review of Economics and Statistics 83:560-3.

Feinberg, Robert M., and A. F. M. Alaur Rabman. 2001. A causality test of the relationship between bank and credit union lending rates in local markets. Economics Letters 71:271-5.

Hannan, Timothy H. 1991. Bank commercial loan markets and the role of market structure: Evidence from surveys of commercial lending. Journal of Banking and Finance 15:133-49.

Hannan, Timothy H., and Robin A. Prager. 2001. The competitive implications of multimarket bank branching, Federal Reserve Board, Finance and Economics Discussion Paper No. 2001-43.

Humphrey, David B., and Lawrence B. Pulley. 1997. Banks' responses to deregulation: Profits, technology, and efficiency. Journal of Money, Credit, and Banking 29:73-93.

Kwoka, John E., Jr. 1979. The effect of market share distribution on industry performance. Review of Economics and Statistics 61:101-9.

Ravenscraft, David J. 1983. Structure-profit relationships at the line of business and industry level. Review of Economics and Statistics 65:22-3 1.

Rhoades, Stephen A. 1987. The effect of nonbank thrift institutions on commercial bank profit performance in local markets. Quarterly Review of Economics and Business 27:16-28.

Rhoades, Stephen A., ed. 1997. Special issue on industrial organization topics in banking. Review of Industrial Organization 12:1-139.

Saving, Thomas. 1970. Concentration ratios and the degree of monopoly. International Economic Review 11:1396.

Simons, Katerina, and Joanna Stavins. 1998. Has antitrust policy in banking become obsolete? New England Economic Review, March/April:13-26.

Tokle, Robert J., and Joanne G. Tokle. 2000. The influence of credit union and savings and loan competition on bank deposit rates in Idaho and Montana. Review of Industrial Organization 17:427-39.

Robert M. Feinberg *

* Department of Economics, American University, Washington, DC 20016-8029, USA; Email [email protected].

The author thanks Ataur Rahman for research assistance; Bill Kelly, Doug Davis, Tim Hannan, Larry White, and two referees for helpful suggestions; and acknowledges research funding support from the Filene Research Institute and the Center for Credit Union Research, School of Business, University of Wisconsin--Madison. All views expressed are those of the author alone.