Earnings Quality As Measured By Predictability of Reported Earnings

Abstract

The study selected a single normative criterion related to earnings quality, that current and prior years’ earnings per share (eps) should be reasonable predictors of future eps. Correlations between one year’s eps and the following year’s eps were sometimes much higher than expected. A surprise was that the second prior year’s eps was a slightly better predictor of the current year’s eps than the immediate prior year. Performing multivariate regressions with an additional prior year dependent variable provides an improvement in the residual sum of squares (two independent variables predict better than three, etc). But with the addition of each additional prior year the percentage improvement in the residual sum of squares declines. Further, starting with the sixth prior year, adding an additional dependent variable makes no noticeable impact on prediction.