Estimating Volatility Shocks

Abstract

This paper proposes a framework and a model-consistent estimation approach for the analysis of the dynamic consequences of changes in volatility. The proposed model is a Vector Autoregression in which time-varying volatility has a first-order impact on the observable variables. The volatility process is estimated within the model, and therefore, the proposed estimation approach does not rely on proxy measures of aggregate uncertainty as it has been generally done in the literature extant. Estimates of the proposed model using data from the United States show important quantitative and qualitative departures from estimates incorporating non-model-consistent measures of volatility. In particular, an increase in overall volatility similar to the one experienced during the Great Recession is predicted to have a strong negative and persistent impact on key macroeconomic indicators, including output, investment and the unemployment rate, and to worsen financial conditions. Moreover, a decomposition of the estimated volatility time series shows that fiscal volatility shocks are associated with important deflationary pressures, have a strong crowding out effect on investment and increase the cost of borrowing. Finally, the estimated model predicts that volatility has an asymmetric effect on the economy so that only rare shocks matter.