Chen and Zimmermann 2022
“Open Source Cross-Sectional Asset Pricing” by Andrew Y. Chen and Tom Zimmermann (Critical Finance Review, 2022, 11(2), 207–264) provides public data and code that reproduce 319 cross-sectional stock-return predictors drawn from prior meta-studies, comparing replicated t-statistics directly with the original papers. Their finding is two-sided and important for the vault. On one hand, replication succeeds: for the 161 characteristics clearly significant in their original papers, 98% of the replicated long-short portfolios have in-sample t-stats above 1.96 — so the literature is not mostly false positives, against the strongest data-mining-sceptic readings. On the other hand, their companion analyses (“Publication Bias in Asset Pricing Research”, arXiv 2209.13623; “Publication Bias and the Cross-Section of Stock Returns” with Pontiff) confirm the post-publication decay documented by McLean and Pontiff 2016, and find the decay is stronger for predictors that were stronger in-sample.
For this vault the paper sits beside McLean & Pontiff as corroboration of the Backtest-to-Live Performance Gap: even when a backtested edge is real and replicable in-sample, its forward return is materially lower, and the strongest-looking backtests decay the most. Chen and Zimmermann attribute part of the gap to mispricing being arbitraged away rather than pure statistical bias — but the forward-looking lesson is unchanged: published returns overstate what a strategy delivers live. The note carries a negative profitability grade for the same reason as McLean & Pontiff — its finding about profitability is one of decay, not durable alpha — even though its methodological contribution (open, reproducible replication) is exemplary and is the standard the vault’s Replication Crisis in Quantitative Finance note calls for.
Connections
- McLean and Pontiff 2016 — supports, larger replication confirming the post-publication decay, source: https://arxiv.org/abs/2209.13623
- Backtest-to-Live Performance Gap — supports, in-sample-strong predictors decay most out-of-sample, source: https://www.openassetpricing.com/
- Replication Crisis in Quantitative Finance — replication_available, open data and code are the cure it models, source: https://www.openassetpricing.com/
- Data-Snooping Bias — relates, debates how much decay is data mining vs mispricing, source: https://arxiv.org/abs/2209.13623
Chen and Zimmermann 2022 [supports] Backtest-to-Live Performance Gap Chen and Zimmermann 2022 [supports] McLean and Pontiff 2016
Sources
- Chen, A. Y., & Zimmermann, T. (2022). “Open Source Cross-Sectional Asset Pricing.” Critical Finance Review, 11(2), 207–264. https://www.openassetpricing.com/ — SSRN https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3604626
- Chen, A. Y., & Zimmermann, T. (2022). “Publication Bias in Asset Pricing Research.” arXiv 2209.13623. https://arxiv.org/abs/2209.13623