The Risk in Low-Variance Anomaly

The low variance (LV) strategy always bets against the volatile leg of common factor-portfolios. Factor loadings of the strategy are thus perfectly predictable based on the status of factor portfolio variances during the formation period. I find that the strategy earns alpha only when traders have to bear major factor risk to arbitrage it away: LV is an anomaly only when it is expected to bet on factor risk. In other times—when low variance means low factor risk—alpha is exactly zero. My results are consistent with models that rationalize anomalies by arbitrageurs reluctance to eliminate mispricing due to factor risk aversion. I use the findings to develop a trading strategy that uses factor data to time LV.