Differences of Opinion and the Cross Section of Stock Returns
Source: Diether, K. B., Malloy, C. J. & Scherbina, A. (2002) · Journal of Finance 57(5), 2113–2141 · doi:10.1111/0022-1082.00490
TL;DR
Stocks with greater dispersion in analysts' earnings forecasts earn lower future returns. The highest-dispersion quintile underperforms the lowest by 9.48% per year, with the effect concentrated in small stocks and recent past-year losers. This is consistent with Miller's (1977) hypothesis — when opinions diverge and pessimists are kept out by short-sale constraints, prices reflect the optimists and become overvalued — and is inconsistent with treating dispersion as a risk proxy.
What anomaly it documents
Predictor: dispersion in analysts' earnings forecasts (proxy for differences of opinion).
Direction: high dispersion → low future returns (negative predictor); the long leg is low-dispersion stocks.
Shape: monotone across dispersion quintiles; far stronger in small / low-priced / past-loser stocks where short-sale constraints bind.
Sorting variable: DISP = standard deviation of analysts' EPS forecasts, scaled by the absolute value of the mean forecast (require >1 analyst that month).
Each month sort the CRSP/Compustat/I/B/E/S intersection into dispersion quintiles (D1 low … D5 high); equal-weight; rebalance monthly.
Strategy: long D1, short D5 (D1 − D5).
Evidence and replication (IS/OOS)
In-sample, portfolio tests over January 1983 – November 2000 (forecast data from 1976):
D1 − D5 average monthly return = 0.79% (t = 2.88), i.e. ~9.48% per year.
Effect is much larger in the smallest size quintile: D1 − D5 = 1.37% per month (t = 5.98); 68.4% of the spread comes from small stocks.
Robust to Fama–French three-factor and four-factor adjustment: the high-dispersion quintile carries a large negative unexplained alpha (intercept ≈ −0.58% per month); the model does not explain the pattern (GRS rejects).
This is an original-source paper; no separate OOS replication is reported here.
Why it might work
Miller (1977) overpricing under binding short-sale constraints and heterogeneous beliefs: optimists set the price, pessimists are excluded, so disagreement → overvaluation → low subsequent returns.
The data reject the competing view that dispersion proxies for risk (dispersion is positively related to market beta, yet predicts lower returns).
Limitations and risks
Dispersion is confounded with uncertainty, distress, illiquidity, and analyst coverage; identification of the opinion channel is debated.
Concentration in small, low-priced stocks means short legs are costly/hard to short — the very friction the theory relies on, but also a real-world limit to arbitrage.
Key references
Diether, K., Malloy, C. & Scherbina, A. (2002) — Differences of Opinion and the Cross Section of Stock Returns — Journal of Finance
Miller, E. (1977) — Risk, Uncertainty, and Divergence of Opinion — Journal of Finance
Chen, J., Hong, H. & Stein, J. (2002) — Breadth of Ownership and Stock Returns — Journal of Financial Economics
Provenance: verified/generated from the paper's full text.
