Anomalies are not anomalous after accounting for trading costs. The
average anomaly has negative returns net of costs, and even the best
anomalies have tiny net returns after publication. Here are the trading costs for 120 anomalies:
Publication Bias and the Cross-Section of Stock Returns
Publication bias among anomalies is surprisingly small, accounting for a
modest 12% of the typical in-sample return. This small bias comes
from estimating a model of biased publication on replications of 156
cross-sectional stock return predictors.
Do t-stat Hurdles Need to be Raised? Direct Estimates of False Discoveries in the Cross-Section of Stock Returns
t-hurdles that control the FDR at 1% or 5% are very poorly identified. Models that imply that t-hurdles should be raised or lowered are observationally equivalent. In contrast, the FDR among published predictors and bias adjustments for expected returns are robust and small.
Suppose asset pricing factors are just noise. How much data mining would be required to generate the literature? I find it would take 10,000 academics 15 million years of full-time data mining.
Full-Information Examinations of Long-Run Risks and Habit
We horse-race long run risks, habit, and a residual in a Bayesian framework. The price-dividend ratio is 75% residual, but stock returns are 75% long run risks.
An Irrelevance Theorem for Risk Aversion and Time-Varying Risk
We prove a theorem that helps explain why decades of advances in risk modeling have had relatively little effect on either asset prices or business cycles in neoclassical models.
Permanent Working Papers
Semi-Parametric Restrictions on Production-Based Asset Pricing Models
Matching
the data on asset prices requires either extremely volatile IST shocks
or huge capital adjustment costs. Those parameters imply a very low
EIS. These restrictions apply regardless of many other details of the
model.
The best parts of this paper are extended and featured in "An Irrelevance Theorem for Risk Aversion and Time-Varying Risk" with Francisco Palomino.