The Predictive Value of Analyst Characteristics

Abstract

While numerous intuitively appealing characteristics of individual analyst earnings forecast accuracy have been identified, the extant literature has provided no evidence regarding the out-of-sample predictive validity of these characteristics. We examine whether these characteristics possess incremental predictive ability vis-a-vis forecast age, the control variable of these studies. We compare the predictive ability of a consensus forecast based solely on forecast age (Age) with that of a six-factor model consisting of forecast age plus five analyst characteristics suggested by the literature (Model). We use the dual evaluative criteria suggested by Foster (1977) and employed numerous times thereafter to assess predictive ability: (1) the ability to predict the next number in the time series and (2) the ability to approximate the capital market's earnings expectation.

We show that Age predicts as well as Model using both predictive ability criteria. Because Model consists of Age plus five other factors, it is cost-beneficial to use Age in lieu of Model for predictive purposes. FASB Statement of Financial Accounting Concepts No. 2 (1980) identifies predictive value as a criterion of information relevance. We show that analyst characteristics have no relevance for the purpose of obtaining better predictions of the next quarterly earnings number vis-a-vis a simpler model based on forecast age. There are many reasons for identifying individual analyst-based characteristics, but predictive ability of the next quarterly number is not one of them.

Keywords: Predictive ability, Trading profits, Forecast timeliness, Analyst characteristics

JEL Classification: M41, G14, G29