Banking markets and the use of financial services by households.

Gregory E. Elliehausen and John D. Wolken, of the Board's Division of Research and Statistics, prepared this article. Ronnie McWilliams provided research assistance.

When a bank proposes to absorb another bank through merger or acquisition, analysts must determine whether the proposed transaction is likely to reduce the competitiveness of banking services. And whether competition would be diminished depends crucially on the definition of the financial services and geographic area that constitute the "banking market." The current definition assumes that competition occurs only in relatively small geographic areas among financial institutions offering the full range of banking products. Therefore, only local commercial banks (and, when their offerings warrant, local thrift institutions), with their broad range of services, are included in the current definition of a banking market.

The vast majority of banking customers households and small businesses-historically have relied heavily on local commercial banks for their financial services; hence, the current definition of a banking market has worked well for assessing most dimensions of banking competition, such as deposit taking and the provision of credit to small businesses. Yet, although past evidence supports the current approach to defining banking markets, little recent data has been available regarding the banking practices of small businesses and households. The lack of current data has been troublesome because changes in the financial markets in the 1980s may have altered the banking practices of these customers. Among the key market changes are the authorization of interest-bearing checking accounts at all depository institutions; the introduction of money market deposit accounts; the spread of automated teller machines; legislation in most states permitting the interstate acquisition of banks by bank holding companies; and the growth of large, nationwide issuers of credit cards.

To assess the importance of these changes for the analysis of banking markets, the Board of Governors of the Federal Reserve System surveyed small businesses and consumers to learn more about their use of financial services and financial institutions. The survey results regarding small businesses have already been published.1 This article examines evidence on banking markets for households based on the 1989 Survey of Consumer Finances. These data permit an investigation of the full range of financial services and institutions used by households and the distances over which these households conduct their financial affairs.

DEFINING BANKING MARKETS

Analyzing proposed bank mergers for their effect on competition and hence for their potential violation of antitrust statutes requires a case-by-case examination of the relevant economic market. To perform the required review, one must identify all firms that significantly affect the price, quantity, and quality of the services produced by the merging parties. Typically this involves specifying both the variety of products (product market) and the geographic extent (geographic market) over which the firms compete. This section briefly examines the current approach to defining banking markets, reviews arguments concerning changes in the product and geographic dimensions of banking markets, and discusses the information needed to help resolve the issues.

The Current Definition

Until recently, markets for financial services have generally been thought to be local and segmented along institutional lines. This view as applied to banking is based on the Supreme Court's 1963 decision in the Philadelphia National Bank case and has been supported by numerous subsequent empirical studies and several judicial decisions.2 In the Philadelphia decision, the Court concluded that the product market for antitrust purposes was the entire bundle or "cluster" of financial services offered by commercial banks. The Court said that bank customers cluster their purchases because of a cost advantage or a "settled consumer preference" for joint consumption, and therefore only institutions offering the full cluster of bank services-including demand deposits and commercial loans-belonged in banking markets. In addition, the Court concluded that banking markets were local because the vast majority of commercial bank customers obtained financial services from local banks. This product definition-the bundle of commercial bank services-and geographic market definition-local-is still used today in antitrust analysis in banking, although thrift institutions are now included in banking markets when they provide the same financial services as commercial banks.

More recently, some analysts have questioned whether this thirty-year-old view of banking markets is outdated because of subsequent deregulation, market innovation, and advances in electronic technology. We now examine some of the factors that may justify broadening the product and geographic dimensions of banking markets.

Expanding the Product Market

Among the reasons for expanding the product market is that the distinctions among different types of financial institutions appear to have blurred during the 1980s. For example, commercial banks were the sole source of checking accounts when the Supreme Court made its determination. Today, savings institutions and credit unions also offer checking, and many nondepository institutions offer money market accounts with a limited checking feature.

The erosion of traditional distinctions does not end with checking. During the 1980s, legislation allowed savings institutions to enter the consumer credit market and allowed depository institutions to compete with money market mutual funds by offering money market deposit accounts. In addition, some depository institutions began offering discount brokerage services, while many brokerage companies sought to broker customer funds into depository institutions.

On the other hand, commercial banks and other depository institutions still offer products for which there may be no close substitutes-namely insured checking, savings, and time deposits.3 If households cluster their financial services at insured depository institutions, or if insured checking, savings, and time deposits are distinct products and have no close substitutes, then the current practice of limiting banking markets to only commercial banks and comparable other depository institutions may be appropriate.

Expanding the Geographic Market

The theoretical basis for defining banking markets over small geographic areas is a consideration of transaction costs. The theory holds that economic markets are likely to be local whenever the transaction costs associated with purchasing or using services produced by distant producers are high in relation to the value of the service. These high transaction costs render the nonlocally produced services imperfect substitutes for locally produced services. Two groups of transaction costs important in banking are those for transportation and those for information. Transportation costs vary directly with the number of transactions a buyer has with a financial institution, the need to conduct transactions with the institution in person rather than by telephone or mail, and the distance between the buyer and the financial institution. Information costs include the costs for the buyer to search for information about alternative suppliers and the costs for the supplier to evaluate and monitor the credit-worthiness of customers. These costs tend to vary directly with the frequency of search, the distance between seller and consumer, and the degree to which the services supplied are heterogeneous.

Recent developments in financial markets and institutions have almost surely reduced the transaction costs associated with doing business with distant financial institutions. For example, the expansion of ATMs and ATM networks generally increased the number of locations and the hours at which consumers can gain access to their accounts, thereby allowing consumers to conduct some of their banking business away from a branch office and outside regular business hours. Advances in information technology have reduced creditors' costs of credit evaluation, which may allow creditors to serve larger geographic areas. This development has probably facilitated the growth of nationwide issuers of credit cards. The increased availability of credit cards and home equity lines of credit has also reduced consumers' transaction costs for some forms of credit by eliminating the need to apply each time an extension of credit is desired.

The question is whether the level of transaction costs has fallen sufficiently to make locally and nonlocally produced financial services close substitutes. Despite the reduction in transaction costs through electronic technologies, distance-sensitive transaction costs such as those for transportation, information, and search may remain a consideration in choosing financial institutions. If this is still the case, then the geographic extent of banking markets may still be limited for either the cluster or some specific products.

Resolving the Issue

Whether banking markets have changed is ultimately an empirical question. The 1989 Survey of Consumer Finances is particularly well suited to analyzing the geographic and product dimensions of banking markets for households because it provides comprehensive coverage of the sources, locations, and types of services used by households.4 This article uses the survey to examine several questions on household use of financial services and financial institutions:

What is the distance between the offices of the firms from which households obtain financial services and the household?

To what extent do financial institutions other than commercial banks provide financial services to households and is their geographic distribution similar to that of commercial banks?

What is the geographic area for each of the different types of financial services used by households? For example, do services involving frequent transactions tend to be more geographically concentrated than others?

Do households tend to purchase their financial services from one institution? Do some households purchase these services from separate institutions? And are the bundled services obtained from the same types of institutions as services purchased separately?

THE SURVEY OF CONSUMER FINANCES

The 1989 Survey of Consumer Finances (SCF), which was sponsored by the Federal Reserve Board and other government agencies, is the most recent in a series of consumer financial surveys conducted since 1947 by the Survey Research Center of the University of Michigan. The 1989 SCF collected a detailed inventory of assets and liabilities from a representative sample of the population of U.S. households.1 As part of the inventory, the survey three-fourths of all households (table 1). However, other types of depository institutions are also important. About two-fifths of households use savings institutions, and about one-fourth use credit unions. The most frequently used type of nondepository institution is finance companies, used by one-fifth of households.

Overall, nearly every household that uses any financial institution uses a local financial institution, while only one in five uses a nonlocal institution. For depository institutions, households are eight times more likely to use a local institution than a nonlocal institution, but for nondepository institutions, the preference for local offices is only 50 percent greater than it is for nonlocal offices. Number of Institutions and Accounts Commercial banks account for nearly half of the 2.72 financial institutions used on average by households (table 2). In contrast, only one in five financial institutions used is a savings institution, and about one in ten financial institutions used is a credit union. Among the nondepository institutions, finance companies are the most commonly used, accounting for about one in ten of all financial institutions used.

Local institutions are the dominant providers of household financial services, accounting for 84 percent (2.29 of 2.72) of all institutions used. Again, the preference for local over nonlocal institutions is far more pronounced for depository institutions than it is for nondepository institutions.

The pattern is similar for the number of accounts by type of institution (table 3). On average, depository institutions provide 83 percent of the accounts used by households (3.92 of 4.73), and the overwhelming majority of these accounts are obtained locally. Commercial banks account for a little more than half of household accounts, and savings institutions and credit unions account for another third. Only 17 percent of household accounts are at nondepository institutions, and these accounts are distributed more nearly equally between local and nonlocal institutions.

In sum, the data on the number of institutions used, the number of accounts, and the frequency of use lead to the conclusion that the financial relationships of households are heavily dominated by local commercial banks. The finding that the importance of local institutions is less for nondepository institutions raises the question of whether nondepository institutions are used differently, perhaps for fewer or different services, than are depository institutions.

Prima Institution and Main Checking Institution Households were asked to designate a financial institution as their "main" or primary financial institution. Ninety-four percent of all institutions identified by households as primary were local depository institutions, and 63 percent of primary institutions were local commercial banks. About 4 percent of institutions identified as primary are nonlocal, and about 4 percent are nondepository institutions (table 4).

Checking accounts are the financial service most frequently used by households. Checking accounts are particularly important for defining banking markets because they are one of the unique products provided by commercial banks and other depository institutions. A household's main checking account is defined as the account on which most of the household's checks are written. If transaction costs play a role in the selection of any financial institution, it is most likely to be the one used for the main checking account. About 80 percent of designated primary institutions are also the main checking institution, a fact underscoring the importance of the checking account in household financial relations.

Almost all main checking accounts are at local depository institutions (table 5), with 68 percent at local commercial banks, 21 percent at local savings institutions, and 9 percent at local credit unions. Only 2 percent of main checking institutions are nonlocal, and only 0.5 percent are at nondepository institutions.8

The data on the primary institution and the main checking institution suggest that local depository institutions are especially important suppliers of financial services for households. The high percentage of local institutions for the main checking account suggests that transaction costs may indeed make nonlocal institutions imperfect substitutes for local institutions for at least some financial services.

Multiple Product Usage

The average number of accounts used per type of financial institution provides further evidence on the relative importance of the various institutions to households and indicates where households may be bundling or clustering their purchases of financial services. Households on average have about 2.4 accounts at their primary institution and about 2.5 at their main checking institutions, regardless of whether they are commercial banks, savings institutions, or credit unions (table 6). As shown earlier, primary and main checking institutions are usually local depository institutions.

Multiple accounts are less frequent at nondepository institutions than at local depository institutions. Both finance companies and other financial institutions appear to be single-product institutions, each having an average of 1.1 accounts. The only type of nondepository institution that is associated with multiple-account usage is the brokerage company, where the average number of accounts for households using these firms is about 1.7.

In sum, local depository institutions are the principal suppliers of financial services to households, and a local commercial bank is the single most important financial institution. Local savings institutions and credit unions are also important to many households, and nonlocal and nondepository institutions are also used somewhat. But unlike depository institutions, which are almost always local, nondepository institutions are more equally divided between local and nonlocal. Also, nonlocal and nondepository institutions are almost never the household's primary institution nor its main checking institution.

The data suggest the possibility of clustering-purchasing multiple services-at primary financial institutions; at checking institutions, which are generally local depository institutions; and at brokerage companies. In contrast, nonprimary institutions, finance companies, and other financial institutions are more apt to be single-product institutions.

GEOGRAPHIC DISTRIBUTION OF SPECIFIC FINANCIAL SERVICES

In this section we investigate whether nondepository institutions are used by households for the same financial services they obtain from depository institutions and whether the geographic distributions of the financial institutions supplying households varies by the type of service supplied.

Local and Nonlocal Service Use

We divide household uses of financial institutions into asset services-such as checking, savings, and brokerage accounts-and credit services-such as mortgages, credit lines, and installment loans.

Asset Services. For each of the asset services, whether measured by frequency of use (table 7) or average number of accounts (table 8), the use of local offices of institutions is much greater than the use of nonlocal offices. Ninety-three percent (2.65 of 2.84) of asset accounts, for example, are at local offices. Checking accounts are almost always obtained from local institutions. Nonlocal offices are used slightly more frequently for liquid asset accounts (savings, certificates of deposit, and money market accounts), but even so, local institutions are used about nine times more often than nonlocal institutions. About six times more households use local offices for IRAs and Keogh accounts than use nonlocal offices, and about four times more households use local offices for brokerage accounts than use nonlocal offices. Trust accounts are obtained relatively most often from nonlocal institutions, but only 3.2 percent of households use trust services.

These product differences in the distribution of local and nonlocal financial institutions are consistent with hypotheses about the incidence of transaction costs associated with particular products-that is, products with more frequent transactions are more likely to be obtained from local institutions than are products with less frequent transactions. These data also suggest that nonlocal suppliers are not particularly good substitutes for most of the asset services covered. This conclusion seems especially true for institutions supplying checking and savings products.

Credit Services. Overall, nearly three quarters of respondent households have some credit relationship with a financial institution, but households do not depend quite as much on local institutions for credit as they do for asset services. The average number of credit accounts at financial institutions per household is about 1.9. Bank credit cards, used by 56 percent of households, are the most widely used credit product. About two-fifths of households have a mortgage, a little more than one-third have, a vehicle loan, and one in ten have a home equity or other credit line.

Measured by number of accounts, credit lines are the most local credit product, and mortgages are the least local. Still, a little more than three-fourths of mortgages are at local institutions.9

These results show a surprisingly large percentage of local suppliers for credit considering the existence of national suppliers and secondary markets for many of these credit products. Apparently, transaction costs are a significant factor for credit products as well as asset products.10

Geographic Dispersion of Service Use

Data on the geographic dispersion of the financial institutions supplying households with various services can provide further insights into how large geographic markets might be. Indirectly, these data also suggest the relative importance of transaction costs for different financial services.

The survey evidence indicates that geographic areas for financial services used by households may indeed be small. For all but one service, trust accounts, the median distance to offices of financial institutions is ten miles or less; and, again except for trust accounts, at least 75 percent of households' financial institutions are thirty miles or less from home or work (table 9). For nine of the twelve financial services, the median distance from the financial institution is five miles or less. These findings suggest that transaction costs may be quite important to the selection of financial institutions.

As expected, the institutions at which households have checking accounts have the smallest geographic distribution: 50 percent of the institutions are two miles or less from home or work, 75 percent are five miles or less, and 90 percent are fifteen miles or less. Institutions used for other liquid asset accounts are only slightly more widely distributed, with 90 percent of institutions used for these accounts being thirty miles or less from home or work. Institutions used for credit products are more widely dispersed than institutions used for checking or other liquid asset accounts, but even most of these institutions are still not very far from home or work-the median distance for most credit products is five or six miles (chart 1). Again, these findings underscore how tightly circumscribed is the geographic market for household financial products.

TYPES OF FINANCIAL INSTITUTION USED FOR SPECIFIC PRODUCTS

As shown above, the types and numbers of financial services purchased by households differ by location of financial service supplier and type of product. The analysis in this section examines which products are obtained from specific financial institutions and explores how these products may differ between multiple financial service suppliers and single financial service suppliers. The analysis permits an assessment of which financial services belong in the same market and which ones belong in distinct markets.

Use by Type of Supplier

Tables 10 and 11 show the percentage of households obtaining each financial service and the number of accounts for each service obtained from the various types of financial institutions. The tables also include a column showing the use of nonfinancial sources for each financial service, an aspect not considered above.

A little more than one-fourth of all households obtain one or more financial services from a nonfinancial source (table 10). This statistic, however, probably overstates the importance of nonfinancial sources because the financial service obtained from them is almost always credit in the miscellaneous "other loans " category, and generally, the outstanding balance on such loans is small.12 Besides "other loans," few accounts of any kind are obtained from nonfinancial sources.

Asset Services. Checking and other liquid asset accounts may differ from the other financial services considered in that they are almost always obtained from a depository institution; commercial banks are the most frequently used depository source, but savings institutions and credit unions are also important suppliers. The only liquid asset account for which nondepositor institutions are important is money market accounts; nearly one-fourth (0.07 of 0.31) of the accounts are obtained from nondepository sources, which are almost always brokerage companies (table 11).

IRAs and Keogh accounts, brokerage accounts, and trust accounts have relatively large shares of nondepository institution suppliers. Indeed, a nondepository source, brokerage companies, is the second most important source of IRAs and Keogh accounts for households. For brokerage and trust accounts, nondepository sources are more important sources of supply than depository institutions.

Credit Services. Nondepository institutions are significant suppliers of credit services to households: About two-fifths of households have credit relationships at nondepository institutions. Among all financial institutions, commercial banks are the most frequently used institution for every credit product considered, although the relative importance of commercial banks varies by type of credit product. Commercial banks are a source of supply for mortgages about as frequently as savings institutions or nondepository institutions. For vehicle loans, commercial banks and finance companies are used with about the same frequency, and credit unions and savings institutions supply a smaller but significant percentage of households with vehicle loans. Overall, depository institutions are a source of more mortgages and vehicle loans than are nondepository institutions.

Multiple Product Usage Revisited

Earlier, we described data indicating that clustering or multiple product usage, if it occurs, does not occur equally across all institutions. A further analysis of the data, together with the findings above, indicate that multiple product use is concentrated at local depository institutions, particularly at households' main checking and primary institutions; among nondepository institutions, multiple product use is concentrated at brokerage firms.

At the primary, main checking, and other checking institutions, households on average have two to three accounts (memo, table 12). At these institutions, multiple account usage generally includes checking; at least three-fourths of households having accounts at these institutions have checking accounts there. The other product is most often another liquid asset account or a bank credit card. It is important to note again that primary and checking institutions are almost always local depository institutions.

Table 12 also shows that when an account is held at a nonprimary or nonchecking institution, it is most likely to be some form of credit. The occurrence of IRAs and Keogh accounts at these institutions is also greater than at primary and checking institutions.

These, together with earlier findings, indicate that nonprimary financial institutions, especially finance companies and other nondepository financial institutions, are likely to be single-product institutions, and credit products such as mortgages and vehicle loans appear to be associated with these singleproduct financial institutions. The one nonprimary, nondepository institution that is an exception to this conclusion is brokerage companies. Clustering may occur at brokerage companies, and the products involved are IRAs and Keogh accounts, brokerage services, and, less frequently, a money market account.

CONCLUSION

Local depository institutions, especially local commercial banks, are still the main suppliers for most of the financial services used by households. The savings institutions and credit unions used by households are, like their commercial banks, overwhelmingly local. Nondepository institutions used by households are also mostly local, but not to the same extent as are depository institutions.

Commercial banks are the single largest supplier for most of the financial services. Even so, other depository and nondepository institutions are important for some of the financial services considered. Other depository institutions are important suppliers of checking and other liquid asset accounts (savings, certificates of deposit, and money market accounts), as well as some credit, particularly mortgages. Nondepository institutions are relatively more important for credit products.

Households certainly do not purchase all of their services from a single institution. Rather, households seem to bundle some of their purchases at certain institutions (for example, the household's primary institution, the main checking institution, and brokerage companies), and purchase single products from others (for example, nonprimary institutions, finance companies, and other financial institutions such as mortgage and insurance firms).

Clustering, or multiple service usage, is most often associated with the checking account, and the institution at which clustering occurs is typically a local depository institution. Credit products such as mortgages and vehicle loans are often purchased separately. The institutions from which credit is obtained are mostly local, but somewhat less locally concentrated than suppliers of asset services. The institutions from which credit products are obtained are frequently nonbank and nondepository institutions.

These findings are directly relevant to the definition of banking markets for households. They are consistent with the view that the markets for many of the financial services used by households are local. This is particularly true of asset services. Somewhat surprisingly, credit products are also decidedly local as well. Moreover, the data indicate that there may be validity to the notion that commercial banks and other depository institutions offer a unique set of services and products that are often purchased as a bundle. This bundle tends to consist of a checking account and another liquid asset account or credit, although other liquid asset accounts and credit are also purchased separately.

The findings also suggest that each credit service used by households may belong to a distinct economic market. The geographic dispersion of suppliers differs across products, and the institutions important to each of the credit products vary.

At least for households, these results support the current definition of banking markets used in antitrust analysis, which consists of local commercial banks and, when they provide services similar to those of commercial banks, other local depository institutions. Limiting the product market to depository institutions, does not, however, require acceptance of the notion that all bank products belong to the cluster. The survey results suggest that checking and other liquid asset accounts (savings, certificates of deposit and money market accounts) are probably a distinct product. These accounts clearly are different from the other financial services used by households both in terms of the location and types of institutions supplying them. Moreover, these accounts are important: They are used by nearly every household. This market may not be the "traditional" product market definition used in banking, but it does indeed appear to be a relevant economic market for antitrust analysis.

APPENDIX: DEFINITIONS AND IMPUTATIONS FOR MISSING DATA

The 1989 Survey of Consumer Finances collected data on specific financial institutions used by households and the households' business relationship with these financial institutions. These data included the type of financial institution and the distance between the household's residence or a household member's place of employment and the most frequently used office or branch of the financial institution. Distance was reported as less than one mile, or as the actual number of miles between one and fifty, or as more than fifty miles.

The identity and location of each financial institution used by the household was not ascertained for all financial institutions. By design, this information was collected for only the first six financial institutions identified by the household. This restriction was necessary to prevent the interview from becoming too burdensome for households with complicated finances, but in practice few households exceeded this limit. Also by design, the identity of the institution was not collected if the household only had a bank credit card from the institution. Finally, location information generally was not collected when respondents did not recall specific institutions until they were asked about specific financial products. For these institutions, however, institution type was collected. As a result of these considerations, location of the office of the financial institution used by the household is missing for about one-third of the household-institution pairs.

When location was missing, it was imputed assuming that the locations of the unknown institutions are distributed identically to the locations of the known institutions of the same class and for the same product. The classes of institutions were commercial banks, savings institutions, credit unions, finance companies, brokerage companies, and other financial institutions. The products were checking, savings, money market accounts, certificates of deposit, IRAs and Keogh accounts, brokerage services, trust services, bank credit cards, mortgages, vehicle loans, home equity or other credit lines, and other loans. Aggregate product or institution categories are derived from the distribution of these values.

As is true for any dataset with missing values, the imputation procedure could affect the final results. The institutions for which location is known are probably the most important financial institutions to the household, since they were reported without the stimulus of other questions. There are proportionately fewer missing values for location for depository institutions than for nondepository institutions. Within nondepository institutions, missing values were most prevalent for other financial institutions. Among the products considered, missing data were most prevalent for bank credit cards. As mentioned, this latter result is partly due to the data collection procedure, since location was obtained only for those credit card suppliers which also supplied other financial products. Even for this category, however, location is known for about half of the institutions identified. When credit cards are omitted in calculating the aggregate credit statistics, about the same proportion of local and nonlocal suppliers are obtained as those reported in the tables.

The failure to ascertain the identity of all institutions also affected the computation of the number of financial institutions per household in table 2 and the number of accounts per financial institution in tables 6 and 12. If a financial service was not obtained from one of the first six institutions, the SCF requested that the respondent identify the type of supplier (for example, commercial bank, credit union, automobile finance company). Fourteen of the thirty-seven types of supplier categories were financial institutions, and each of these fourteen institution types was assumed to be a different institution. This assumption may understate the number of institutions per household and overstate the number of accounts per institution. The error resulting from this assumption, however, is likely to be small. When they were used, most of the institutions not included in the first six, especially nondepository institutions, had only one financial service indicated.

All statistics reported in this article were computed using weights to produce estimates that represent the population of U.S. households. The weights are the same as those used in Kennickell and Shack-Marquez, "Changes in Family Finances."