According to the Recovery Theorem (Ross, 2015), options data can reveal the market’s true, contemporaneous expectations about a specific future horizon. We implement empirically the theorem’s approach to separate implied (risk-neutral) volatility into 1) Ross-recovered true expected volatility and 2) a risk preference component, using Optionmetrics Ivy option data for the S&P500 index and four European indices (FTSE, CAC, SMI, DAX). This separation leads to better forecasts of realized volatility for all indexes in our sample compared to a traditional benchmark, implied volatility. The improvements are statistically and economically significant, as increases by up to 8%.
Internationally, Ross-recovered expectations are more highly correlated than is the risk preference component. To better understand the sources of risk, we follow Bollerslev et al. (2014) and compute value-weighted global volatility measures. Models using Ross-recovered global risk preferences generally have the best performance among all models tested based on the , AIC and BIC. In contrast, performance generally is not improved for models using only implied volatility. Although global factors load heavily on the U.S. index, non-U.S. markets matter as global measures improve volatility forecasts for the S&P 500.
The findings suggest that, to obtain the best volatility forecast, risk preferences are best measured globally. Consistent with theory, the evidence shows that the markets in our sample are sufficiently integrated to assume that international market participants have common, globally shared risk preferences, but that a local preference risk factor is still present in the data.
These results suggest that applying the Recovery Theorem empirically to option data using this paper’s methodology successfully captures the market’s true expectations and risk preferences, allowing forward-looking risk preferences to be measured for the first time in the literature. This analysis could also pave the way for more accurate equity premium forecasts.