In keeping wTith the spirit of a factor-based approach, much of our analysis assumes that the information set used by the decision maker consists of histories of factors and stock returns. That is, the decision maker does not make use of firm-specific characteristics, except perhaps in constructing the factors (as in, for example, the Fama-French (1993) model). Previous studies have recommended the use of firm-specific characteristics in estimating the cost of equity (e.g., Elton, Gruber, and Mei (1994) or Schink and Bower (1994)), and the usefulness of various firm-specific characteristics in explaining expected returns has been argued recently by Daniel and Titman (1997).

Another feature of the Bayesian approach is that it allows the decision maker to introduce additional prior information about the firm whose cost of equity is to be estimated, and our methodology allows the decision maker to either ignore or incorporate such prior information. In specifying the prior, the firm can be regarded as a random draw either from the whole cross-section of stocks, when firm-specific characteristics are ignored, or from a group of firms with similar characteristics, when the firm’s characteristics are incorporated. As a simple illustration of the latter case, we include a firm’s industry classification as additional prior information and analyze estimates of expected excess returns on stocks of utilities, which constitute an industry in which estimated costs of equity have clear practical relevance.

The remainder of the paper is organized as follows. The methodology is developed in Section I, wherein we present the general form of the priors used in our Bayesian approach, explain how wTe obtain the resulting posterior distributions of [i and its components, and describe the empirical-Bayes procedure used to obtain parameters in the prior distributions. Sections II and III contain our empirical results. Section II reports and analyzes posterior moments of the expected excess return and its components for individual stocks. Those results include a detailed analysis for one stock as wrell as analyses for two cross-sections: a broad universe of 1,994 stocks and a smaller set of 135 utility stocks. Section III investigates the potential uncertainty about the cost of equity that arises from uncertainty about which pricing model to use. Section IV review’s the conclusions.