Mulliner et al. 2025
“Regimes” (SSRN 5164863; DOI 10.2139/ssrn.5164863) is by Amara Mulliner, Campbell R. Harvey, Chao Xia, Ed Fang and Otto Van Hemert. It is a Man Group paper — Xia, Fang and Van Hemert are with Man AHL / Man Numeric, and Man Group has published it openly as “Regimes, Systematic Models and the Power of Prediction.” Versions are dated 4 March 2025 (working) and 4 October 2025 (latest). The paper proposes a deliberately simple, non-parametric similarity-based regime model: the user picks a set of economic state variables (seven in the paper — stocks, curve steepness, short yields, oil, copper, VIX, and stock-bond correlation), each transformed to a rolling z-score, and each month is compared by Euclidean distance to every historical month. The aggregate distance is a “Global Score”; the historically closest months define the current regime. The authors stress this avoids “presupposing a set of possible regimes” and contrasts with discrete-regime or fixed-transition statistical models — so it is not a Markov Regime-Switching Model and not a hidden Markov model.
The empirical test times six long-short equity factors — the Fama-French five (Market, Size, Value, Profitability, Investment) plus 12-month Momentum — over 1985-2024. The rule: go long a factor if its average return after the most-similar historical dates was positive, short otherwise. Aggregating across the six timed factors, the headline result is the “difference” portfolio — long the most-similar quintile, short the most-dissimilar (“anti-regime”) quintile — which the paper reports has a 0.82 Sharpe ratio, only 0.37 correlation to a static long-only factor book, and alpha three standard errors from zero. The “anti-regime” finding (months most dissimilar to today perform worst) is presented as the paper’s distinctive contribution.
It appears in this vault as a recent pro-alpha claim for Regime Classification and as the clearest named-author regime publication from Man Group. Two things raise the bar for taking it at face value, in opposite directions: co-author Campbell Harvey is one of the field’s leading critics of data-snooping in strategy evaluation, which lends credibility — yet the same scrutiny he applies elsewhere is exactly what is missing here. It is graded weak for profitability. The firm itself discloses that variable selection “induces look-ahead bias,” that the paper is “not intended to be either complex or comprehensive,” and that it is methodological rather than a finished strategy. No transaction costs or slippage are modelled; only Sharpe ratios are reported, “in a vague sort of way” per an independent quant review; the static long-only benchmark actually beats the timed quintile-one portfolio on raw return (the timed book only wins on crisis drawdown); factor timing is itself contested (see Factor Timing and Asness’s “siren song” critique held in Live Regime-Model Evidence Gap); and there is no independent replication. The similarity model is a genuine, transparent regime-classification method; the standalone-alpha claim is not substantiated after costs.
Connections
- Regime Classification — reports_profitability, source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5164863
- Similarity-Based Regime Detection — proposes_model, source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5164863
- Man Group — relates, source: https://www.man.com/insights/regimes-systematic-models-power-of-prediction
- Campbell Harvey — proposes_model, source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5164863
- Factor Timing — tests_strategy, source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5164863
- Regime-Based Asset Allocation — tests_strategy, source: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5164863
- Transaction Costs and Slippage — excludes_costs, source: https://www.man.com/insights/regimes-systematic-models-power-of-prediction
Ontology
Mulliner et al. 2025 proposes Similarity-Based Regime Detection Mulliner et al. 2025 supports Regime Classification Mulliner et al. 2025 tests_strategy Factor Timing Mulliner et al. 2025 part-of Man Group
Sources
- Mulliner, A., Harvey, C. R., Xia, C., Fang, E. & Van Hemert, O. (2025). “Regimes.” SSRN working paper 5164863. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5164863
- Man Group (2025). “Regimes, Systematic Models and the Power of Prediction.” https://www.man.com/insights/regimes-systematic-models-power-of-prediction