Is There a Replication Crisis in Finance?
Read the paperopens doi.org in a new tab
What they found
Responding to Harvey-Liu-Zhu and Hou-Xue-Zhang, the authors built a public dataset of 153 factors in 93 countries and re-examined replicability using a Bayesian framework that pools information across related factors. By their measure the majority of factors do replicate, both in the U.S. and internationally, and clustering factors into themes shows that most themes are robust. Their key argument is that the earlier pessimistic results came from demanding each factor stand alone rather than recognizing that dozens of published factors are variations on a few underlying ideas.
What you can use
- The core factor themes (momentum, value, quality, low risk) hold up around the world; the noise is in the hundreds of near-duplicate variants.
- Global data is a powerful out-of-sample test: a signal that works only in the U.S. deserves suspicion.
- The authors' factor dataset is free and can be used to check your own ideas against 90+ countries.
Caveats
Bayesian priors and factor-clustering choices drive the conclusions; the paper is part of an ongoing debate rather than the final word. Advanced statistics.
Tags: anomalies, replication, global, factor-zoo
Summaries are our own reading of the paper, not the authors' words. Educational only, not advice. Discuss it in Book Club.