Research

I investigated the predictability of factor returns using common economic and financial variables. Unlike traditional market-timing models, I found that most of economic and financial predictors exhibit low signal-to-noise ratios for factor timing. Using recent advances in Bayesian econometrics with variational inference, I showed that stochastic volatility improves density forecasts, while time-varying parameter models tend to overfit. Forecast combination and economically motivated constraints (e.g., Sharpe ratio bounds) enhance out-of-sample performance in terms of certainty-equivalent returns but not Sharpe ratio. (Link)

I investigated the role of subjective belief on portfolio level. Different from the previous research on subjective belief, I found that the divergence between subjective belief and rational belief may be exaggerated because of the heterogeneity in slopes and the cross-sectional dependence component in the error structure. Using the latest development on panel data econometrics, I found that analysts may underestimate the one-year cash flow growth and overestimate the second-year cash flow growth, and this relationship will change significantly during the recession. This evidence supports the time-varying rationality of expectation and adaptive market hypothesis.

I proposed investigating the factor timing problem using the Bayesian Decision Predictive Synthesis framework, where factor-specific predictors are integrated with a predictive likelihood that is adjusted according to different economic regimes.