What Happens After A Technology Shock? A Bayesian Perspective

This paper investigates the effect of a positive technology shock on per capita hours worked within the class of Bayesian Vector Auto-Regressive [BVAR] models. Such a framework avoids the current debate regarding the specification issue of per capita hours [level versus first-difference stationary]. Six priors are considered and for each, we examine the impulse responses of per capita hours following a positive technology shock. The marginal posteriors of the VAR parameters are generated using the Markov Chain Monte Carlo (MCMC) Gibbs sampler. We find that the estimation of the VAR yields significantly different estimates under competing priors. Using the Francis and Ramey (2004, UCSD working paper) new measure for per capita hours, and after imposing the identifying restrictions (i.e., Blanchard-Quah and sign restrictions), the results show that per capita hours worked rise following a positive technology shock - if one [objectively] assumes a non-informative prior.