Evidence on a Simplified Model of Systematic Risk

Abstract

ics in finance and accounting. Empirical analyses have examined correlations between accounting variables and beta [1, 2, 3, 9, 11], the association between macroeconomic factors and beta [26], and the relationship between operating and/or financial leverage and beta [5, 10, 12]. More commonly seen in the literature is the multiple regression model of beta as a function of numerous accounting variables and ad hoc proxies for other possible determinants [4, 6, 13, 16, 18, 19, 25, 28, 31]. Unfortunately, the somewhat inconsistent results of those studies are not clearly reconcilable into a unified profile of the determinants of systematic risk. An econometric explanation is not easy. In multiple regression studies such as Breen and Lerer [6], the presence of numerous accounting variables in a single model creates possibilities of obscuring the effects of any appropriately specified variables. Furthermore, it is not at all clear which, if any, of such variables should be tested at all. The firm's systematic risk should be related to the underlying activities of the firm [27], but without a rigorous theoretical model, testing the effects of such variables is not a promising route. For example, Rosenberg and McKibben [28], Pettit and Westerfield [25], and Thompson [31] all developed statistical models relating beta to financial variables, and while their results were interesting, in each case it could be said that the data were in search of a theory. Myers [23] recently attempted to provide a theory in which investor predictions of cash flows were generated by an adaptive expectations model. Solving this multi-period problem confirmed his suspicions that cyclicality, growth, and earnings volatility were determinants of beta. Turbull [32] set up a model where firm value is stochastically generated by a Geometric Brownian motion process. He found that beta was determined by growth of expected cash flows, duration