The Case Against Factor Timing

Systematic equity managers commonly apply static factor weights when ranking stocks by multi-factor scores. This paper asks whether that assumption holds in the data. Using Fama–French long-short portfolio returns from July 1963 to December 2025 (750 months), the paper tests whether factor premiums are genuinely time-varying or whether a constant weighting scheme is statistically justified. Two competing estimates of each factor's premium are compared: an expanding-window average (representing the static approach) and a 60-month rolling average (representing a time-varying approach). Structural break tests applied to monthly returns fail to detect any genuine shifts in the average premium for any of the six factors considered — even though the same test wrongly suggests strong shifts when applied to the rolling-average series instead of the raw returns. A separate forecasting test across 690 out-of-sample months confirms this: the rolling estimator does not improve predictive accuracy for the equal-weighted factor composite at any window length tested, and performs significantly worse than the static estimator for the profitability and investment factors, with no factor showing a significant improvement. The practical implication is that strategies which adjust factor weights in response to rolling-premium estimates are responding to statistical noise rather than genuine regime shifts — and incur measurable cost in doing so.