1. Does Academic Research Destroy Stock Return Predictability. R
Does Academic Research Destroy Stock Return Predictability? R. David McLean (Alberta) Jeffrey Pontiff (BOSTON College) Risk 2015 March 31, Generously supported by the Dauphine-Amundi Chair in Asset Management

2. Our Research Question
Academic research has uncovered many predictors of cross- sectional stock returns
e.g., size, momentum, book-to-market, etc….
Does return-predictability typically persist outside the original sample?
We explore this question using 97 “predictors” documented in published academic studies

3. The Structure of Our Tests
We compare returns over 3 periods:
Sample period from the original study
Post-sample, but pre-publication
Out-of-sample, but paper is not widely distributed
Post-publication
Paper distributed, more people know than before

4. Does Predictability = Statistical Artifact?
If so, predictability should disappear immediately out of sample
Perhaps researchers choose methods and samples that give them their desired result
Data Mining Bias

5. Does Predictability Reflect Risk and Costs?
Then predictability should be similar in-sample, out-of-sample, and post-publication
Sharpe (1964) – Market Risk
Amihud and Mendelson (1986) – Transaction costs

6. Does Predictability Reflect Mispricing?
Then publication should attract arbitrageurs, who correct the mispricing
Costless arbitrage: The effect disappears entirely
Costly arbitrage: Effect is reduced, not eliminated
Pontiff (1996, 2006)

7. Main Findings – Suggest Mispricing
Out-of-sample, pre-publication decline is 26%
1% monthly return declines to 0.74%
Upper bound estimate of statistical biases
Post-publication decay is 58%
1% monthly return declines to 0.42%
Implies a publication effect of 58%-26%=32%

8. Choosing the Predictors
Peer-reviewed academic studies
Primarily in top 3 finance journals
Characteristics that can be constructed with COMPUSTAT, CRSP, and IBES data
Cross-sectional predictors only

9. The Predictors 97 in Predictors in Total
Oldest: Blume and Husic (1972)
Include a few forthcoming papers
We include variables with strong theoretical motivations
Fama and MacBeth (1973)--market beta
Amihud (2002)—illiquidity
Most predictors are not theoretically motivated

10. Creating the Sample Constructing portfolio returns and estimation
Long-short monthly quintile returns
Pooled cross-sectional time-series regressions
Each predictor has 3 distinct periods
In sample — average, 329 months
Mean Return, 58 bp
Out-of-sample but pre-publication — average, 44 months
Mean Return, 40 bp
Post-publication — average, 141 months
Mean Return, 26 bp

11. Main Regression Results Dependent variable is long-short monthly return
Variables
(1)
All 97
predictors
(2)
85 w/ in-sample significance
Post-Sample (S)
-0.150***
-0.180**
(0.077)
(0.085)
Post-Publication (P)
-0.337***
-0.387***
(0.090)
(0.097)
Null Hypothesis: S=P
0.024
0.021
Null: P = -1*(Mean)
0.000
Null: S=-1*(Mean)

12. In-Sample Returns vs. Post-Publication Decay

13. In “Event” Time

14. Do Time Trends or Persistence Explain the findings?No

15. Returns by Predictor Type
We explore returns and decays by predictor type
Event – Corporate events, changes in performance, downgrades
Fundamental – constructed only with accounting data
Market – Constructed only with market data and no accounting data
Valuation – Ratios of market values to fundamentals

16. Returns and Decay by Predictor Type

17. Costly Arbitrage
Costly Arbitrage: Predictors that are less costly to arbitrage have lower alpha, especially post-pub.
Size
Dollar Volume, Turnover
Dividends
Idiosyncratic risk
Principal Component of all four

18. High Arbitrage Costs = More Alpha
Publication Dummy (P)
Index
P x Index
Index +
P * Index
Coefficient
-0.272**
-0.046***
-0.036
-0.082***
**Maximum index value is 3.87
**High index values mean less costly to arbitrage

19. Trading Activity in Portfolio Stocks
If academic research attracts arbitrageurs, then it should cause an increase in trading in predictor portfolios
We test whether trading in the stocks that make up predictor portfolios increases out-of-sample and post-publication.

20. Trading Activity ↑ in Portfolio Stocks
Variables
Variance
Trading Volume
Dollar Volume
Short - Long Short Interest
Post-Sample (S)
-0.054***
(0.007)
0.092***
(0.001)
0.066***
0.166***
(0.014)
Post-Pub.(P)
-0.065***
0.187***
0.097***
0.315***
(0.008)
(0.013)
Observations
52,632
41,026
Time FE?
Yes No
Predictor FE?
Null: S=P
0.156
0.000

21. Correlations Across Predictors
If predictors reflect mispricing, and mispricing has common causes (e.g., sentiment)….
We might expect in-sample predictor-portfolio returns to be correlated
If publication attracts arbitrageurs…..
Then published predictor-portfolios may be more highly correlated with other published predictor-portfolios

22. Publication Affects Correlations—Factor Model
Variables
Coefficients
In-Sample Index Return
0.748***
(0.000)
Post-Publication Index Return
-0.008
(0.243)
P x In-Sample Index Returns
-0.674***
(0.425)
P x Post-Publication Index Return
0.652***
(0.576)
Publication (P)
-0.880
(0.544)
Constant
0.144***
(0.267)

23. Conclusions Evidence suggests markets learn from academic research
Return-predictability falls 58% post-publication
Predictor portfolios that decline the most have the largest in-sample returns
Trading activities increase in predictor portfolios
Predictor correlations change with publication