UK macroeconomic forecasting with many predictors: Which models forecast best and when do they do so?

摘要: Block factor methods offer an attractive approach to forecasting with many predictors. These extract the information in these predictors into factors reflecting different blocks of variables (e.g. a price block, housing financial etc.). However, model which simply includes all as risks being over-parameterized. Thus, it is desirable use methodology allows for parsimonious models hold at points time. In this paper, we dynamic averaging and selection achieve goal. automatically alter weights attached evidence comes about has forecast well recent past. empirical study involving output growth inflation using 139 UK monthly time series variables, find that set changes substantially over Furthermore, our results show can greatly improve performance relative traditional methods.