The Effects of Past Entry, Market Consolidation, and Expansion by Incumbents on the Probability of Entry in Banking

The actual timing of branching deregulation varied. While the Riegle-Neal Act allowed for interstate branching as of 1997, some states opted out and waited longer before deregulating, and some states deregulated earlier.

The earliest studies of entry in the banking industry, including Hanweck (1971), Boczar (1975), and Rose (1977), covered limited samples and time periods.

Theoretically, growth need not lead to greater profitability and probability of entry; it could lead to a more competitive equilibrium among incumbent firms.

These papers all focus on the effect of mergers on the creation of new charters in urban markets. Seelig and Critchfield (2003) look at entry into over 300 urban markets over 1995–1998 and measure entry by the number of new bank and thrift charters. Berger et al. (2004) also find that past merger activity encourages the creation of new bank charters in all three types of markets that they study: rural markets, small metropolitan areas and large metropolitan areas. Using a measure of merger activity that accounts for the size of the acquired banks and includes only consolidations among unaffiliated firms, Keeton (2000) finds a significant positive relationship between past consolidation and entry. Keeton also finds that new bank formation comes mainly in response to mergers that shift ownership from small, local banks to larger banks that are headquartered at a distance from the market entered.

This work builds on the research of Bresnahan and Reiss (1991) and Cohen (2004).

In rural markets, the share of branches controlled by multi-market institutions increased from 73.5 % in 1995 to 87.6 % in 2008. In urban markets, the branch share of multi-market institutions increased from 80.5 % in 1995 to 88.5 % in 2008.

We also measured market size and growth with either deposits or employment, but we found that these variables are highly correlated with population and population growth and produce similar results if included in place of population measures. Using measures that are based on deposits has the drawback of introducing potential endogeneity into the estimation.

We include both banks and thrift institutions when computing the HHI and other measures of market structure.

The SOD data have been supplemented with thrift information from the Branch Office Survey data that were collected by the Office of Thrift Supervision.

We make multiple checks on potential new branches because branch identifiers are missing for thrifts and can be inconsistent for banks. There are cases where new branches are opened in locations where a branch existed previously. This occurs because the infrastructure for a branch (e.g., a vault) is already in place at that location. These new branches rarely open in the same year in which the old branch was closed. Furthermore, there are cases where branches move locations. Hypothetically, a moved branch could be measured as entry, but we do not treat a relocation of an existing branch by an incumbent bank to another location within the same local market as entry.

In other words, we check to see if the financial institution was reported in the National Information Center data that are maintained by federal bank regulators in the years previous to the appearance of the institution in the SOD.

MSAs are delineated using 2002 definitions; these definitions are held constant throughout the sample period.

The number of new charters that is reported in this paper is substantially less than the number that is reported in some previous research. There are several possible causes of these differences: First, existing institutions can change charters for various reasons, and a charter change may be recorded as an exit and a new entry. Second, for legal or tax reasons, mergers are sometimes structured in a way that results in the elimination of two existing charters and creation of a new charter; this could be recorded as two exits and one entry into a market. Third, some studies appear to designate conversions of already existing financial institutions into institutions that are newly required to report in the SOD as entry.

Entry with new charter creation very seldom occurs in multiple markets for the same charter. In our sample, 97.5 % of new charters enter only one market; about 2.5 % enter two markets; and in only three instances does a new charter enter three markets.

All 318 MSA markets experience entry at some point in our sample, and entry occurs in 310 of these markets in subsequent years. In rural markets, 1332 experience entry, and 652 experience entry multiple times.

A “high-holder” institution is the firm at the top of a banking organization’s organizational chart, typically a bank holding company that may own only one bank or a number of bank and non-banking subsidiaries.

The ordered-probit model is an extension of the simple probit model. The χ² test we employ tests an unconstrained three-way ordered-probit model versus a model in which the two cut points are constrained to be equal, yielding the equivalent of a simple probit model. In this way, we can nest the simple and ordered-probit models within one framework.

The results for the simple probit, Tobit, logit and Poisson regressions are available from the authors upon request.

The χ² statistic is 368.35, which is well above the 1 % threshold.

The χ² statistic is 99.59.

A test of the equivalence of the yearly dummy variables was rejected at the 1 % level in all of the models that were tested.

In both the urban and rural samples, adding a dummy variable for those states that have not yet fully deregulated branching yields an insignificant coefficient and has no effect on the results for other variables.

See, for example, Milligan (2003).

The χ² statistic is 47.16.

The two cases in which the results of Tables 5 and 7 differ are the positive and significant coefficient on per capita income in the regression on new branches in rural markets and the positive but insignificant coefficient on market population in the regression on branch entry into urban markets.

Results for all alternative specifications, robustness checks and marginal effects tests are available from the authors upon request.

An alternative approach would be to evaluate the variable of interest at one standard deviation above and below the mean. We did not take this approach, because, in some cases, the skewed distribution of the variable puts a one-standard-deviation change in the variable outside of its range in the sample.

If the number of past entrants is increased to its maximum value of 36, the probability of single entry actually declines, because multiple entries would be projected to occur 94 % of the time.