This paper aims to identify what drives the temporal reduction in the value relevance of earnings documented in the literature. Is it the increasing noise in stock returns over time, noise in earnings, or both?
The authors develop hypotheses from the lead/lag structure between stock returns and accounting earnings and perform empirical tests using data from annual COMPUSTAT and monthly CRSP over the sample period of 39 years (1970‐2008).
The test results show that increasing noise in stock returns over time is primarily responsible for the temporal reduction of R2 in regressions of returns on earnings. Additional analysis shows weak evidence that both the noise in returns and the noise in earnings are responsible for the declining association between earnings and returns in a sub‐period (1970‐1982).
The R2‐based methodology has limitations because, as Gu points out, regression R2s might be incomparable across samples. The findings suggest that future research should control for the effects of the temporal increase in market noise before making value relevance inferences from the declining association between earnings and returns.
The paper contributes to the limited body of research on noise in stock returns as the main driver for the temporal reduction in value relevance of earnings.
