Christian Dorion

Volatility Components, Affine Restrictions, and Nonnormal Innovations

Recent work by Engle and Lee (1999) shows that allowing for long-run and short-run components greatly enhances a GARCH model’s ability fit daily equity return dynamics. Using the risk-neutralization in Duan (1995), we assess the option valuation performance of the Engle-Lee model and compare it to the standard one-component GARCH(1,1) model. We also compare these non-affine GARCH models to one- and two- component models from the class of affine GARCH models developed in Heston and Nandi (2000). Using the option pricing methodology in Duan (1999), we then compare the four conditionally normal GARCH models to four conditionally non-normal versions. As in Hsieh and Ritchken (2005), we find that non-affine models dominate affine models both in terms of fitting return and in terms of option valuation. For the affine models we find strong evidence in favor of the component structure for both returns and options, but for the non-affine models the evidence is much less strong in option valuation. The evidence in favor of the non-normal models is strong when fitting daily returns, but the non-normal models do not provide much improvement when valuing options.

Option Valuation with Macro-Finance Variables

I propose a model in which the price of an option is partly determined by macro-finance variables. In an application using an index of current business conditions, the new model outperforms existing benchmarks in fitting underlying asset returns and in pricing options. The model performs particularly well when business conditions are deteriorating. Using the recent financial crisis as an out-of-sample experiment, the new model has option-pricing errors that are 18% below those of a nested 2-component volatility benchmark. Results are robust to using alternative business conditions proxies and comparing to different benchmark models.

Volatility Forecasting and Explanatory Variables: A Tractable Bayesian Approach to Stochastic Volatility

We provide a formulation of stochastic volatility (SV) based on Gaussian process regression (GPR). Forecasting volatility out-of-sample, both simulation and empirical analyses show that our GPR-based stochastic volatility (GPSV) model clearly outperforms SV and GARCH benchmarks, especially at long horizons. Most importantly, our approach enables the straightforward incorporation of arbitrary covariates without requiring the specification of functional forms a priori. Augmenting the GPSV model with exogenous variables increases its performance even further. In particular, a simple set of covariates reduces the error rate on one-year out-of-sample forecasting during the 2007-09 recession by 26% relative to a benchmark range-based SV model.

What Drives Corporate Asset Prices: Short- or Long-Run Risk?

This paper investigates the relative impact of two fundamental types of systematic risk on equity prices. Equity risk premium is determined in a consumption-based corporate finance model with time-varying macroeconomic conditions. We show that long-run risk, which exposes firm profitability to the slowly-moving business cycle, commands most of the risk premium in recessions. However, during periods of expansions, most of the equity risk premium relates to short-run risk, which arises from the instantaneous covariation between firm cash flows and consumption. These findings are robust to accounting for endogenous capital structure and default policies. An empirical analysis with U.S. firms over the period 1952-2016 provides support for the model predictions.This paper addresses this question with an asset-pricing model featuring endogenous corporate policies. Long-run risk reflects a firm’s profit exposure to slowly-moving expected consumption growth, whereas short-run risk captures the exposure to frequent unexpected changes in consumption growth. Long-run risk reduces a firm’s optimal leverage while driving most of the equity risk premium. The contribution of short-run risk increases during expansions and for firms with higher idiosyncratic volatility, but remains similar across levels of default risk and systematic volatility. These findings contribute to understanding the expected return on a stock, both over time and in the cross-section.