Trend and Cycle Shocks in Bayesian Unobserved Components Models for UK Productivity

Abstract

This paper presents a range of unobserved components models to study productivity dynamics in the United Kingdom. We introduce a set of univariate and bivariate models that allow for shocks between the trend and the cycle to be correlated, and use Bayesian sampling techniques to estimate the models. We show that the size of the priors on the trend and cycle shock has an effect on the results, suggesting that a range of priors need to be considered for policy-making purposes. If the prior is set to a smooth trend, then models with little correlation between the trend and cycle shocks are the likeliest to fit the data. On the other hand, if there is a prior belief that the trend shock is allowed to vary relatively freely, the results suggest that there is a negative correlation between trend and cycle shocks to LIK productivity. This is consistent with real-business cycle type narratives, where trend shocks are the main driver of productivity dynamics. Finally, our evidence suggests that the trend productivity growth rate in the UK has been weaker since the financial crisis. There is also a significant positive correlation between shocks to UK trend productivity and those of other advanced economies.