SignalTrace

Factor Timing

John Montgomery4 min read

Factor Timing

Factor timing is the practice of dynamically tilting a portfolio toward the equity style factors — value, momentum, quality, size, low-volatility — expected to deliver above-average returns over the coming period, rather than holding static factor weights. It is also called “smart-beta timing” or “style timing.” The premise is that factor risk premia are cyclical: they vary with valuations, the macro environment, or a detected market regime, and an investor who leans into the currently favourable factors and away from the unfavourable ones can earn more than a buy-and-hold blend. Because factor investing is sold at fees between active management and cap-weighted indexing, factor timing is attractive to providers precisely because it reintroduces a skill-based, “active management” component into an otherwise systematic product.

The literature is broadly skeptical, and the most-cited skeptical voice is Cliff Asness. His 2016 editorial “The Siren Song of Factor Timing” (Journal of Portfolio Management, Special Issue; SSRN 2763956) deliberately frames factor timing as a siren song — alluring and dangerous. Asness examines valuation-based (“factor value”) timing signals and finds them “quite weak historically,” and notes that whatever timing power they do have is “too highly correlated to the simple value factor itself” — meaning a factor-timing overlay is largely a redundant repackaging of the value factor rather than an independent source of return. He further argues that some third-party tests of factor-timing power are “exaggerated and/or inapplicable.” His prescription is the opposite of timing: identify the factors you believe in over the very long haul and diversify aggressively across them; and if you must time, since timing is a “sin,” “sin only a little.” The position is contested — Rob Arnott of Research Affiliates argues factor valuations are a useful timing guide — but the burden of proof sits with the timer.

In this vault factor timing is the broader category that Style Factor Rotation belongs to. Wang Lin Mikhelson 2020 is the vault’s concrete instance: it uses a three-state HMM regime classifier as the timing signal, switching among six style-factor models depending on the inferred regime. Substituting an HMM regime label for Asness’s valuation signal changes the trigger but not the underlying difficulty — a regime-conditional timing rule is still a factor-timing strategy, and it inherits the same skepticism: the signal must be genuinely informative out-of-sample, the implied turnover must be paid for, and the in-sample selection of “which six factor models, classified by which three states” is itself a Data-Snooping Bias surface. Asness’s own caution that more elaborate timing systems “become more like traditional Tactical Asset Allocation” applies directly: a regime-switching factor strategy is best understood as TAA on factor sleeves, with all of TAA’s well-known fragility.

This skepticism is a key reason the vault grades the Style Factor Rotation / Wang-Lin-Mikhelson outperformance claim as weak: a backtest that shows factor timing “worked” must clear an unusually high bar, because the broad evidence is that factor-timing signals are weak, easily overfit, and largely collinear with the value factor. Factor timing is not impossible, but in this vault it is treated as an unproven source of edge, and any HMM-driven instance of it is graded on out-of-sample performance net of costs, not on a cost-free in-sample Sharpe figure.

Factor Timing [contradicts] Cliff Asness Style Factor Rotation [part-of] Factor Timing Factor Timing [relates] Tactical Asset Allocation Factor Timing [relates] Overfitting in Quantitative Trading

Connections

Sources

Graph View

Backlinks