Bad News Travels Slowly: Size, Analyst Coverage, and the Profitability of Momentum Strategies
Source: Hong, H., Lim, T. & Stein, J. C. (2000) · Journal of Finance 55(1), 265–295 (NBER WP 6553) · DOI: 10.1111/0022-1082.00206
TL;DR
A direct test of the gradual-information-diffusion theory of momentum (Hong & Stein, 1999). On NYSE/AMEX/NASDAQ stocks 1980–1996, momentum profits have an inverted-U in firm size, are much stronger in low-analyst-coverage stocks holding size fixed, and the coverage effect is driven almost entirely by past losers — consistent with firm-specific, especially negative, information diffusing only gradually. Momentum looks like an information-diffusion phenomenon, not a risk premium.
What anomaly it documents
Predictor: past 6-month return interacted with the information environment — firm size and residual analyst coverage (coverage orthogonalized to size and a NASDAQ dummy).
Direction: standard momentum (winners > losers) that intensifies as the information environment thins.
Shape: profits are an inverted-U in size and decline with analyst coverage; the coverage effect is concentrated in the loser leg (asymmetric underreaction to bad news).
OSAP predictor: related to OSAP AnalystValue/Mom6m family; analyst-coverage-conditioned momentum is not itself a standalone OSAP signal.
How to construct it
Form 6-month/6-month momentum: each month sort into terciles on prior 6-month return; long top 30% (P3) winners, short bottom 30% (P1) losers (the paper's preferred P3–P1 measure has better signal-to-noise than P10–P1).
Compute residual coverage from month-by-month cross-sectional regressions of log(1+ANALYSTS) on log(SIZE) and a NASDAQ dummy; use coverage lagged six months ("stale") to avoid endogeneity.
Drop stocks below the 20th NYSE/AMEX size percentile (they have no analyst variation); double-sort on size and residual coverage.
Evidence and replication
Evidence
Sample
Magnitude (per month)
t-stat
Baseline momentum P3–P1, full sample
NYSE/AMEX/NASDAQ 1980–1996
0.53%
2.61
Peak (3rd size decile, ~$45M cap)
1980–1996
1.43%
6.66
Low-coverage SUB1 momentum
1980–1996
1.13%
—
High-coverage SUB3 momentum
1980–1996
0.72%
—
Coverage spread (SUB1−SUB3)
1980–1996
0.42% (≈60% higher)
3.50
Loser leg P1/SUB1 − P1/SUB3 ("LAST" trade)
1980–1996
−0.70%
5.16
The size effect is non-monotonic: momentum is negative in the very smallest (~$7M) stocks, peaks at 1.43%/mo in the third decile, then declines to ~zero in the largest stocks.
~3/4 of the momentum profit in mid-cap deciles comes from the loser side (P2–P1), holding consistently across deciles 2–8.
Comparison anchor: Jegadeesh–Titman (1993) report ~0.95–1%/mo; the lower 0.53% here reflects the milder P3–P1 cut and the damping effect of tiny NASDAQ firms.
Why it might work
Hong–Stein's heterogeneous-investor model: private information diffuses gradually across investors, so prices underreact and drift. Where information spreads slowest — small firms, thin coverage — drift is largest. The loser asymmetry fits firms' and analysts' reluctance to publicize bad news, so negative information is impounded most slowly. Low coverage is also associated with lower beta (median 0.75 vs 0.95), so the effect is not a hidden risk premium.
Limitations and risks
Analyst coverage is an imperfect, time-varying proxy; the I/B/E/S coverage universe and zero-analyst share (77% in 1976 → 37% in 1996) shift over the sample.
SUB1/SUB3 size matching is imperfect (large low-coverage names like GM land in SUB1).
As with all momentum, the strategy carries crash risk, high turnover, and small-cap trading costs.
Key references
Hong, H. & Stein, J. (1999) — A Unified Theory of Underreaction, Momentum Trading and Overreaction — Journal of Finance
Jegadeesh, N. & Titman, S. (1993) — Returns to Buying Winners and Selling Losers — Journal of Finance
Fama, E. & French (1996) — Multifactor Explanations of Asset Pricing Anomalies — Journal of Finance
Provenance: verified/generated from the paper's full text.
