Measuring the Effect of Financial Stress

• Activity Number: 247
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Michael Owyang*+ and Ana Beatriz Galvao
• Companies: Federal Reserve Bank of St. Louis and Queen Mary College, University of London
• Keywords: dynamic factors ; STAR model ; covariate selection ; MCMC

Abstract

In this paper, we construct a measure of financial stress which is defined by its effects on the dynamics of macroeconomic variables. Based on previous work by Hubrich and Tetlow (2010), we create an index of financial stress which has the property that high stress times alter the underlying dynamic responses of macro variables to shocks. This implies a sort of regime switching initiated by a rise in the stress index. We augment the standard logistic smooth transition autoregression with a factor which enters contemporaneously into the VAR and with a lag into the transition function. Estimation of the model is conducted in a Bayesian environment using the Metropolis-Hasting steps to draw the transition function parameters. Because of the nonlinearity in the transition function, the factor must be estimated using a nonlinear filter. We use the unscented Kalman filter as a proposal density in a Metropolis step. We find that nonlinearities produced by increases in financial stress. Estimating the model, we find three periods of high stress. During each of these periods, the responsiveness of macro variables to shocks to financial variables is heightened.