We examine whether time-series attenuation of post-earnings-announcement drift (PEAD) varies with the degree of matching in earnings. While “bottom-line” earnings news exhibits deteriorating serial correlation and a pronounced decline in return predictability over time, we find that “well-matched” earnings news exhibits robust serial correlation and over three times less return attenuation. Notably, well-matched earnings news continues to have strong serial correlation and return predictability for firms with poor matching characteristics, while bottomline earnings news has weak serial correlation and no return predictability for these firms in the post-2000 period. Overall, our results suggest that PEAD attenuation is partially attributable to decaying serial correlation in earnings consistent with declining matching (Dichev and Tang 2008) and that evolving serial properties of earnings merit the attention of market participants and researchers that study earnings-based return predictors.