Variance Risk Premia Estimation and Stock Return Predictability

Louis Galowich & Lukas Gejgus

Student thesis: Master thesis

Abstract

In this paper we explore how to derive the variance risk premium which is a premium paid by investors to shield against large price swings. Formally, it is defined as the difference between the expectation of the risk-neutral and physical return variation and represents the payoff of a variance swap. We rely on the VIX volatility index and high frequency realized variance as proxies for risk-neutral and physical variance, respectively. An empirical challenge in the computation of the variance risk premium is modelling a forecast of the actual (or physical) expected volatility accounting for the distribution properties and stylized facts of the realized measure. To this end, we employ forecastingmodels based on a heterogenous autoregressive (HAR) framework to generate the one-month ahead estimate of the realized variance. Lastly, we show that the variance risk premium can be a predictor for stock returns at short horizons (i.e. less than one year). Although significant, the proportion of the variance in future returns explained by the variance risk premium is rather small in a univariate regression analysis. The degree of predictability is more impressive within a multivariate regression including other commonly employed predictor variables.

EducationsMSc in Applied Economics and Finance, (Graduate Programme) Final Thesis
LanguageEnglish
Publication date2021
Number of pages120