Investor Return Extrapolation and Cross-Sectional Stock Return Predictability

Summary

The article summarizes research using Forcerank, a platform where participants rank stocks by expected performance over roughly a week. Regressions of consensus ranks on past returns show that participants extrapolate recent performance, with more weight on recent weeks. An exponentially decaying weighting model captures this pattern. Extrapolation is asymmetric: negative past returns receive greater and more persistent weight, and nonfinancial participants show stronger tendencies than finance professionals. Estimates also vary with company characteristics.

Consensus ranks are negatively associated with subsequent returns. The article separates the component predicted by historical returns from the remaining ranking signal; both are negatively related to future returns, suggesting the signal contains more than simple return extrapolation. Weekly portfolios sorted on ranks retain significant long-short returns after the reported factor adjustments. Evidence comes from a platform sample of 293 mostly large, growth-oriented stocks, so it may not generalize to all investors or securities. The summary reports an observational study and does not establish that the strategy will remain profitable after implementation costs or outside its sample.

Key ideas

• Forcerank participants tend to expect recent winners to keep outperforming, with stronger weight on more recent returns.
• Participants react more strongly and persistently to negative past returns than to positive ones.
• Nonfinancial participants exhibit stronger return extrapolation than finance professionals in the reported estimates.
• Consensus ranks and both their history-predicted and residual components are negatively associated with subsequent returns.
• The reported weekly rank-sorted strategy retains significant long-short returns after the stated factor adjustments, within a limited platform sample.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.