Analysis of Linear Factor Models with Multivariate Stochastic Volatility for Stock and Bond Returns
46 Pages Posted: 18 Jul 2003
Date Written: February 18, 2003
We explore high-dimensional linear factor models in which the covariance matrix of excess asset returns follows a multivariate stochastic volatility process. We test crosssectional restrictions suggested by the arbitrage pricing theory, compare competing stochastic volatility specifications for the covariance matrix, test for the number of factors, and analyze possible sources of model misspecification. Estimation and testing of these models is feasible due to recent advances in Bayesian Markov chain Monte Carlo (MCMC) methods. We find that five latent factors with multivariate stochastic volatility best explain excess returns for a sample of seventeen stock and bond portfolios. Analysis of cumulative latent factor shocks suggests that APT pricing restrictions, coupled with constant factor risk premia, do not adequately explain cross-sectional variation in average portfolio excess returns.
Keywords: Arbitrage Pricing Theory, Factor Model, Multivariate Stochastic Volatility, Markov chain Monte Carlo, Gibbs sampling
JEL Classification: G12, C11, C15, C32
Suggested Citation: Suggested Citation