Econometrics with system priors

A Multivariate Unobserved Components Model to Estimate Potential Output in the Euro Area: A Production Function Based Approach

“…These hyper-parameters imply fairly persistent but stationary processes that are clearly distinguishable from I(1) processes, since close to 90 per cent of the probability mass of these prior distributions fall between 0.5 and 0.95. In case of the output gap that is defined as an AR(2) process as in Clark (1987), the above beta prior is used on the sum of the autoregressive coefficients, following the procedure of imposing system priors by Andrle and Plašil (2018). Given the (0,1) support of the beta distribution, this is sufficient to ensure stationarity of the output gap.…”