A Profile of the Return-Suprise Relation

Abstract

The relation of share prices to the information in annual earnings announcements is investigated by measuring return-surprise correlations across partitions of the distribution of earnings or earnings surprise. Correlations in each partition are compared to partitioned correlations from the normal distribution and reveal new insights into the return-surprise relation. In particular, a few extreme events in small firms and value stocks can have a disproportionate influence on estimates of the return-surprise relation. The largest losses and negative surprises are often accompanied by large positive returns during both announcement and pre-announcement periods, resulting in negative and non-normal return-surprise correlations. Large gains and positive surprises also tend to be accompanied by large positive returns, resulting in high and non-normal correlations. Small negative surprises are not highly correlated with return during announcement periods, but are reflected in pre-announcement returns. Although losses are less informative than gains over pre-announcement periods, losses (particularly small losses) are more informative than gains during earnings announcement periods.