Do macrofactors help forecasting stock market volatility?∗

This study examines several dynamic heteroskedastic factor model specifications to test for the confidence set of model parametrizations that best incorporate economy-wide information for forecasting stock market volatility. To this end, diffusion indices (i.e. factors) are distilled from two large sets of US excess stock returns and macroeconomic variables. Using 40 years of data, the main empirical results show that: (i) suitably selected averages of both data sets are good proxies for the systematic driving forces of excess returns; (ii) the pooling of information gives rise to additional risk factors capturing subtile influences of economic variables and relations which would otherwise have been ignored; (iii) macroeconomic factors indeed help to forecast co-volatilities of excess returns, whereby the macroeconomic information can best be modeled as innovations to factor betas that follow a stochastic drift process; (iv) the factor extraction method is of minor importance for selecting the best volatility forecasting model.