TOWARD A POSITIVE THEORY OF EQUITY VALUATION*

Abstract

IN RECENT YEARS academic and business institutions have rapidly expanded their efforts to develop explanatory models of equity valuation. The principal approach has been to postulate theories centered around firm economic activity and then to test these theories by single-equation regression models estimated from financial data for historical firms. The basic conjecture of this dissertation is that such models of equity valuation would fail to meet reasonable statistical tests of reliability, so that any discussion of the implications of specific parameter estimates tends to be spurious. Specifically, it is conjectured that (1) the estimated parameters of the models would not prove consistently significant, (2) the parameters would not exhibit reasonable stability in different cross-section samples, and (3) the parameters would not exhibit stability for the same sample over time. Models selected for testing include formulations suggested by Durand, ModiglianiMiller, Barges, Benishay, and Gordon. These are five widely-known explanatory models that address themselves to various issues in equity valuation. The data used in the tests are financial statistics for firms in four sample groups during the four years, 1956-1959. The sample groups included both itdustry-type risk class samples and a multi-industry sample. The four years selected include different years in a significant stock-market cycle. As a result, the sixteen basic regressions and subsequent pooled regressions computed from these data comprise a tableau of estimated coefficients across groups and across years that provides more extensive results for each model than has heretofore been available. There are four principal findings: