Critical review of models of earnings mean reversion

Abstract

The hypothesis that earnings are mean reverting was suggested 90 years ago and has been extensively tested since then. Expectations of earnings’ mean reversion (hereinafter “EMR”) significantly influence pricing of shares or earnings forecasts. Despite proposals and testing of numerous models of EMR, there has been very little inquiry into the meaning of those models in corporate and valuation terms in the academic literature. Therefore, we see such an inquiry as highly desirable. The aim of this paper is to critically review the models of transitory earnings (vice versa EMR), their methodology, practical applicability of their results, and their limitations stemming from the characteristics of earnings data. We find that most of the recent models of transitory earnings (EMR) are misspecified in terms of target earnings or reasons of EMR. We also find that EMR is partly caused by cycles in relevant industry or economy, and partly by company-specific processes and accruals. Also, elimination of survivorship bias and use of margins or lower-level profitability like ROI and ROC instead of ROE is worth testing in EMR models.