Reported Earnings and Analyst Forecasts as Competing Sources of Information : A New Approach

Givoly (1985) provides formal evidence on the relation between the past history of earnings and their own forecast. Our study uses a new methodology, modified Granger causality tests, to further analyze the information flows between earnings and forecasts. Our application of this widely acclaimed time series approach complements existing cross-sectional studies by abstracting from stock market reactions to information, and focusing on the dynamic interaction between reported earnings and analysts' forecasts. Based on long time series of analyst earnings forecasts and reported earnings, we provide formal and compelling evidence of bi-directional causality. Further, we report that the lag structure in information flows is longer than has been documented in the previous literature. This is consistent with our expectation that, in addition to past earnings reports, the forecasts themselves make a significant contribution to the information that is reflected in future earnings. However, the presence of feedback also suggests that past earnings reports as well as past forecasts are incorporated into later forecasts. Collectively, our findings imply that the information in earnings reports has inherent positive value and that forecasts do not fully substitute for earnings releases