Download This PaperQuantifying the Effect of Spillovers on Analyst Outputs
45 Pages Posted: 9 Feb 2017 Last revised: 1 Jun 2022
Date Written: May 31, 2022
Abstract
Prior research suggests a large and positive spillover effect among sell-side analysts. However, most of the prior work does not account for the feedback loops in spillovers and thus the spillover estimates could be biased. To address this issue, in this study, we use an iterative model that relates mean peer group ability along with the analyst’s own ability and firm-quarter fixed effects to various analyst’s research outputs (forecast bias, accuracy, speed, and the breadth of information production). We show that analyst spillover effect varies by the type of research outputs. First, we find a significantly negative spillover for EPS forecast bias and a small negative spillover for EPS forecast accuracy. This suggests that when an analyst has a peer group that is more optimistic (pessimistic) or accurate on average, an analyst’s EPS forecast will be more pessimistic (optimistic) and less accurate. Second, we find no spillovers for revenue forecast bias but a significantly positive spillover for revenue forecast accuracy. This suggests that whether the peer group’s revenue forecast is more optimistic or pessimistic does not bias an analyst’s revenue forecast in any direction. However, an analyst’s revenue forecast accuracy is positively associated with the peer group’s revenue forecast accuracy. Lastly, we find significant and positive spillovers for forecast revision speed and the number of non-EPS forecast type production consistent with a competition argument that if an analyst knows that the peer group is likely to provide quick updates and more non-EPS forecasts, the analyst will be more inclined to follow as well. This study contributes to the existing literature by presenting a new way to estimate the spillovers and providing new insights on how analysts interact and learn.
Keywords: spillovers; analyst ability; analyst forecast accuracy; iterative model
JEL Classification: D80, G29, M40
Suggested Citation: Suggested Citation