Clustering Macroeconomic Variables

Many papers have highlighted that some macroeconomic time series present structural instability. The causes of these remarkable changes in the reduced form properties of the macroeconomy is a debated argument. In literature this issue is handled with three main econometric methodologies: structural breaks, regime-switching, and time-varying parameters (TVP). Nevertheless, all of these approaches need some ex ante structure in order to model the change. Based on the Recurrent Chinese Restaurant Process, I have specified a model for an auto-regressive process and estimated via particle filter using a conjugate prior, which applied the idea of evolutionary cluster to the study of the instability in output and inflation for the U.S. after War World II. This procedure displays some advantages; in particular it does not require a strong ex ante structure in order to neither detect the breaks nor manage the evolution of parameters. The application of the cluster procedure to GDP growth and inflation rate for the U.S. from 1957 to 2011 shows a good ability in fit the data. Moreover, it produces a clusterization of the time series that could be interpreted in terms of economic history, and it is able to recover key data features without making restrictive assumptions, as in âone-breakâ or TVP models