The Optimal Time-varying Factor Portfolio: A Mean-variance Approach

Abstract

Timing the exposure to risk premia plays a huge role in active asset management as well as in academia in order to achieve the highest return for the lowest risk. This paper studies the optimal time-varying exposure to eight common equity factors from Jul. 1963 to Nov. 2022 via modern portfolio theory and volatility forecasting, while incorporating transaction cost. The thesis finds that factor-timing can yield a higher Sharpe ratio than an equal-weight portfolio, as the means, variances and correlations of the factors vary over time. The effect is largest if the investor takes long-short factor bets besides a full investment in the market portfolio, which leads to a doubling of the Sharpe ratio after transaction cost, compared to the equal-weighted portfolio. For example, the investor benefits from investing in size during economic booms, quality and earnings during downturns, and can use value and investment as an inflation hedge. The culprit is that although factor-timing is beneficial theoretically, forecasting returns and the macroeconomy is extremely tricky, wherefore real-world implementation of factor-timing is hard and likely to result in excess risk. A valid implementation method is adjusting the asset allocation slightly to the underlying economy. The paper also finds that if the goal is to maximize the added alpha per unit of added idiosyncratic risk, the equal-weight portfolio substantially outperforms the factor-timing portfolio. Therefore, the benefit of factor-timing comes down to investor preferences. Estimating the volatility via GARCH-models improves the fit of the volatility and correlations, but still underperforms due to poorly fitted means. Hence, the investor can gain an edge by combining GARCH estimates with a model for mean returns. The paper establishes that the success of factor-timing depends on the assumptions on transaction costs as well as forecasts of means and variances more so than the true link between factor returns and the economy. In theory, factor-timing does pay both over time and across business cycle stages, but the practical implementation is very tricky.