Earnings Estimates, Predictor Specification, and Measurement Error

Increased use of expectational data for modeling stock returns places a spotlight on the specification of predictor variables. Choices between alternative specifications of a given predictor such as E/P or earnings trend, for example, can have wide-ranging effects on portfolio selection and quantitative modeling. Investigation here indicates that the importance of predictor specification may vary depending upon the predictor and the investment strategy, and that the relationship between predictors and returns may vary across types of stocks