Determinants of Entry and Exit: An Application of the Compounded Bivariate Poisson Distribution to U. S. Industries, 1972-1977

Abstract

Numerous empirical studies have applied a variety of approaches and data to examine the market entry of firms across industries. One branch of studies, which is the focus of the present paper, uses industry-level census data to examine the role of industry characteristics as entry barriers. This research features estimation results from very large samples that cover a wide range of industries. Examples include McGuckin [19], Orr [20], Gorecki [9; 10], Duetsch [6; 7], Kessides [14], and Chappell, Kimenyi, and Mayer [5] (henceforth referred to as Chappell et al. [5]). The above studies use net entry (e.g., the number of firms in 1977 minus the number of firms in 1972) as a measure of the number of new entrants in a given industry.' With the exception of Chappell et al., classical regression models of net entry have been the tool of analysis. Chappell et al. point out that because net entry data are integer-valued, and thus deviate from classical regression assumptions, the statistical specification of net entry calls for a discrete probability distribution. Chappell et al. base estimation on a univariate Poisson distribution. In the present paper we consider another aspect of the data, and improve upon the approach of Chappell et al. Observations on net entry actually reflect the difference between unobservable entry and exit flows, and as such contain information on both entry and exit. While this point is well understood, the focus of net entry studies has been entry, and the approach has been to model the data as a function of entry alone. Unfortunately, the failure to incorporate the true informational content of the data into the estimation clouds the results. In particular, neglecting the dependence of the data on exits can produce misleading estimates of the entry parameters. Small values of net entry, for example, might merely reflect a balancing of strong entry and exit flows