Option-Implied Forecasts with Robust Change of Measure
Abstract: Option prices can be useful in forecasting due to its forward-looking information, yet the challenge lies in how to transform the risk-neutral measure implied by options to the physical one. While existing papers deal with this issue by assuming an equilibrium model, we introduce a novel approach that does not require any model assumptions by adopting the robustness approach (Hansen and Sargent, 2001). We take the risk-neutral measure as the reference model and consider a set of measures that are “close” to it in the relative entropy sense. The forecast is then determined along with the worst-case scenario, i.e., the probability measure that leads to the maximum expected loss, and we call the solution the robustly-transformed measure. We show that the relative entropy bound, which needs to be fixed by a forecaster, can be determined by exploiting the theoretical connection between the risk-neutral and the physical measures. In an empirical application to S&P500 returns, we document that our approach results in conservative forecasts on market risks, with 5%-value-at-risk 4%/12%/17% (for 1-day/10-day/22-day horizons) larger, and 5%-expected shortfall 5%/15%/24% (for 1-day/10-day/22-day horizons) larger than those under the risk-neutral measure.
Return Predictability and Risk management (joint with Nour Meddahi)
Forecast Comparison Tests Under Fat Tails (joint with Nour Meddahi and Jihyun Kim)
Toulouse School of Economics (2015 - 2020)
Ph.D. in Economics, 2020
M.A. in Economics, 2014
Dissertation title: Three Essays on Financial Risk Management and Fat Tails
Committee members:
Professors Raffaella Giacomini, Andrew Patton, Francis Diebold, Jihyun Kim, Christian Gourieroux and Nour Meddahi