Evidence on the size of banking markets from mortgage loan rates in twenty cities.

Financial innovation and a surge in bank mergers over the past decade have intensified the debate over whether banking markets are still local. That is, have the decade's developments caused banking markets to grow beyond cities and rural localities to become regional or national in scope? This question is particularly important with respect to bank mergers and acquisitions: If banking markets are nonlocal, then supply and demand conditions in a locality should not affect prices in that locality; therefore, such local conditions would not be relevant in the evaluation of the competitive effects of any given merger.

This study conducts an empirical test of the issue by examining whether local market concentration and other local market conditions have a systematic influence on mortgage interest rates, that is, on the price of mortgages. A by-product of this analysis is a test of the price--concentration relationship, which is a hotly debated element in general antitrust analysis. The study analyzes more than 13,500 observations of mortgage interest rate terms and non-interest rate terms covering a sixteen-week period for twenty U.S. cities from late 1987 through early 1988.

It has been argued that, of all banking services, mortgage lending has been the most affected by the developments tending to widen geographic markets. A prominent claim in this regard is based on the fact that such loans represent by far the largest and longest-term financial service purchased by most consumers. Consumers, therefore, presumably would be willing to incur greater transactions costs in searching for a low-cost mortgage than they would in obtaining other financial services because the returns to searching would likely be the greatest in the case of the mortgage. This behavior should generally expand the size of the geographic market and drive mortgage interest rates toward a competitive level. Furthermore, in recent years, mortgages have been packaged as the basis for mortgage-backed securities, which are sold in national markets, as are mortgages themselves--for many years, they have been sold off by local originators to large "permanent investors" located in major cities.

These arguments suggest that mortgage rates should provide a relatively stringent test of the proposition that the market for banking services is local. That is, mortgage loans have probably been subject to more forces conducive to a geographic broadening of the market than most other bank services; consequently, a finding that mortgage rates are still determined by local market factors is probably a good indication that the market for consumer-oriented banking services in general is local in nature. A finding that the mortgage market is local would also be particularly striking because of the nature of this study's statistical sample, which consists entirely of urban areas. Urban locations typically have far more competitors than nonurban areas, and competition is generally one of the factors determining price differences between markets. Thus, a sample characterized by a relatively high degree of competition should bias the result of a test for price differentiation toward uniformity of prices.

Data on the interest rates (prices) banks charge for their loans are not generally available. Listings of mortgage rates, however, are fairly common in newspapers throughout the country. Therefore, to construct a sample, this study recorded mortgage rates for the sixteen weeks from September 20, 1987, through January 9, 1988, as they appeared in weekly listings published in newspapers in twenty cities. The twenty cities were selected more or less randomly, although some choices were made to achieve a degree of geographic diversity.

The listings covered 1,039 lending institutions, which, over the sixteen weeks, produced 15,286 interest rate observations for fixed-rate, conventional mortgages. The sample included all the major types of mortgage originators in the United States--commercial banks, savings and loan associations, savings banks, credit unions, and mortgage banks; 873 observations were not classified by type of firm. Most of the newspaper listings of rates also showed other loan terms: the percentage downpayment required, the length of the loan, points charged (fee in terms of percentage points of the loan amount), and the maximum loan to which the terms apply. The exclusion of observations with missing items left 13,525 observations available for the statistical tests.

Both an ordinary-least-squares fixed-market-effects model and an OLS model containing specific market variables indicate that local market conditions influence mortgage interest rates. Market growth and per capita income are the most important variables, but the Herfindahl index, a measure of market concentration, also plays a role after controlling for costs.

These results for mortgage loans suggest that analyzing bank competition within local geographic markets is still appropriate. That is, local market conditions still make a difference. Moreover, the results suggest that market concentration, as measured by the Herfindahl index, affects the prices charged in local markets for mortgages. This would appear to support the traditional market-power explanation for the market structure-performance relationship rather than the Demsetz efficiency explanation.