We propose a refined method to estimate the conditional aggregate Relative Risk Aversion coefficient from equity index option prices. The estimated coefficient is used to translate risk-neutral distribution moments into physical moments, with a focus on forecasting realized returns through conditional expected return estimation. By applying Generalized Methods of Moments to a system of moment conditions and restrictions, we avoid common estimation issues in single-equation approaches. A simulation study shows improved estimation accuracy, and an empirical application to S&P500 options data demonstrates reduced levels of expected variability and expected return, and enhanced return forecasting and stock market timing ability out of sample.