This thesis focuses on variance risk and correlation risk in the equity market, and
consists of three essays. The first essay demonstrates that the variance risk, mea-
sured as the difference between the realized return variance and its risk-neutral
expectation, is an important determinant of the cross-sectional variation of hedge
fund returns. Empirical evidence shows that funds with significantly higher loadings
on variance risk outperform lower-loading funds on average. However, they incur
severe losses during market downturns. Failure to account for variance risk results
in overestimation of funds' absolute returns and underestimation of risk. The results
provide important implications for hedge fund risk management and performance
evaluations.
The second essay examines the empirical properties of a widely-used correlation
risk proxy, namely the dispersion trade between the index and individual stock
options. I find that discrete hedging errors in such trading strategy can result in
incorrect inferences on the magnitude of correlation risk premium and render the
proxy unreliable as a measure of pure exposure to correlation risk. I implement a
dynamic hedging scheme for the dispersion trade, which significantly improves the
estimation accuracy of correlation risk and enhances the risk-return profile of the
trading strategy.
Finally, the third essay aims to forecast the average pair-wise correlations between
stocks in the market portfolio. I investigate a comprehensive list of forecasting models and find that past average correlation and the option-implied correlation provide
superior out-of-sample forecasting performance compared to other predictors. I provide empirical evidence showing that the forecasts of average correlation can improve
the optimal portfolio choices and substantially enhance the performance of active
correlation trading strategies.