In the present paper we propose a method to infer on the different clusters potentially present in a panel data set that is data driven in the sense that the classification of each subject into a specific group is estimated along with the model parameters. The general model allows additionally for time-varying parameters, whereby the timing of the structural changes is also part of the model estimation. The presence of two latent variables, the group- and the state-identifying indicators, calls for Bayesian Markov chain Monte Carlo techniques. An application to individual bank lending data of the US banking sector illustrates the methodology. We obtain results that are broadly consistent with the bank lending view. Moreover, we infer a significant asymmetric effect of monetary policy over time which favors the evidence for models of credit cycles.
30 Okt. 2002
CORE (Center for Operations Research and Econometrics)