Lakonishok, Shleifer, and Vishny (1994) provide the canonical behavioral explanation for why value strategies outperform: investors systematically extrapolate past performance too far into the future, overpricing “glamour” stocks and underpricing “value” stocks.
Key results
Using NYSE/AMEX stocks over April 1968 to April 1990:
- B/M deciles: value stocks earned 19.8%/yr vs. 9.3%/yr for glamour stocks (10.5%/yr spread)
- C/P deciles: 20.1%/yr vs. 9.1%/yr (11.0%/yr spread)
- E/P deciles: 19.0%/yr vs. 11.4%/yr (7.6%/yr spread)
- Past sales growth: lowest growth quintile earned 19.5%/yr vs. 12.7%/yr for highest (6.8%/yr spread)
Two-dimensional sorts (combining expected and past growth) produce even larger spreads: C/P + sales growth yields a 10.7%/yr value-glamour gap (cumulative 100% over 5 years).
The extrapolation argument
Investors form expectations by extrapolating past growth rates without adequately accounting for mean reversion (Kahneman and Tversky’s “base rate neglect”). Glamour stocks had high past growth and are expected to continue growing fast; value stocks had poor past growth and are expected to remain poor. When actual growth mean-reverts, glamour stocks disappoint and value stocks positively surprise.
LSV show that the value-glamour return difference is concentrated in periods around earnings announcements, consistent with expectations being corrected. Over the 5 post-formation years, the cumulative earnings surprise (as measured by announcement-day returns) is strongly positive for value and negative for glamour.
The risk argument is insufficient
LSV challenge the Fama-French view that value returns compensate for higher risk:
- Value stocks do not have higher betas than glamour stocks
- Value stocks do not underperform in recessions; their outperformance is remarkably consistent across states of the world
- In the worst 25 months for the market, value stocks decline less than glamour stocks, not more
- Traditional risk measures fail to explain the magnitude of the value-glamour return differential
Significance for the field
This paper launched the behavioral vs. risk-based debate over the value premium that continues to define empirical asset pricing. The risk camp (led by eugene-fama and kenneth-french) responded with the factor model framework, arguing that HML captures a priced distress risk. The behavioral camp (LSV, Daniel and Titman 1997) argues the premium reflects mispricing that is corrected over time.