Working Paper: CEPR ID: DP15057
Authors: Ling Cen; Yuk Ying Chang; Sudipto Dasgupta
Abstract: We show that when the locations of analysts covering a firm are geographically more diverse, the individual forecasts of the analysts for that firm are less correlated. More geographical diversity of co-analyst locations leads to more accurate individual analyst forecasts. This suggests that analysts assign weights to co-analysts’ forecasts when making their own forecasts, and the individual forecasts become more accurate due to a diversification effect. Moreover, in line with efficient weighted average forecasting, our results indicate that the weights assigned to peer forecasts vary with measures of the precision of the analyst’s signal and those of the peers. Overall, our evidence suggests observational learning in the analyst setting. Our empirical design avoids typical pitfalls of outcome-on-outcome peer effects (Angrist, 2014) by showing that an analyst’s expected absolute forecast error (proportional to standard deviation) is affected by the covariance of co-analyst’s forecast errors (as captured by their locational diversity).
Keywords: information; diversity; learning; herding; analyst forecasts
JEL Codes: G24; D83
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Geographical diversity among analysts (R12) | Less correlated individual forecasts (C29) |
| HHI decreases (L19) | Average covariance between forecast errors (C10) |
| HHI decreases (L19) | Absolute forecast error of the consensus forecast (C53) |
| Analysts assign weights to peer forecasts (G17) | More accurate individual forecasts (C53) |
| Covariance of coanalysts' forecast errors (C51) | Expected absolute forecast error of an analyst (G17) |
| Geographical diversity among analysts (R12) | Forecast errors become more accurate (C53) |