Forecast horizon and business cycle effect on analysts' forecast bias and stock returns

The purpose of this dissertation is to investigate whether financial analysts' forecasting behavior is consistent with the prescriptions of the availability heuristic---a behavioral theory of judgment proposed by Tversky and Kahneman (1973). Literature indicates that analysts' forecasts are overall optimistically biased and inefficient. The availability heuristic posits that analysts overweight the information that is easy to use, easy to retrieve, or recent, in making their judgments.; We propose that under this heuristic, analysts' forecasts will not only appear biased, but they will exhibit a "horizon" effect. That is, the extent of this bias will be an increasing function of the forecast horizon. Further, we posit that this horizon effect will bear a systematic association with aspects of the analysts' information set which (supposedly) drives their forecasting behavior. Specifically, we hypothesize that long-term five-year-ahead earnings forecasts issued in expansion (contraction) phases of the business cycle will exhibit a more pronounced optimistic (pessimistic) bias than one-year-ahead and two-year-ahead forecasts. In a similar vein, we hypothesize that forecasts of firms with good (bad) current performance, lower (higher) predictability will exhibit more (less) pronounced horizon effect. Our test results with 9,194 firm-year observations from I/B/E/S forecast and actual EPS data (1982--2003) lend overall support to these predictions.; We use the drivers of the horizon effect to derive adjusted forecasts which we expect will show a less pronounced horizon effect. Our results indicate that adjusted forecasts are overall less biased and show less pronounced horizon effect than raw forecasts. We examine whether the market also makes similar adjustments to analysts' forecasts by testing whether it responds more strongly to adjusted forecast errors than to raw forecast errors. We find that adjusted forecast errors show a stronger association with forecasting period stock returns compared to raw forecast errors only under certain conditions.