In order to provide medium run forecasts of headline and core HICP inflation for the euro area, we assess the usefulness of dynamic factor models. We use Stock and Watson's (1999) out-of-sample methodology for models estimated over the 1988:1-2002:3 period, with balanced and unbalanced panels. We provide evidence that factors alone or combined with indicators help improve upon the simple Autoregressive (AR) model for forecasting HICP core inflation as well total inflation, if one refers to the usual criterion of "Relative MSE" together with its standard deviation. However, regarding total HICP we do not produce forecasts that are totally satisfactory in the sense of being capable of recognizing the 1999-2000 upturn in inflation in a timely manner. But, from that point of view, the construction of a ''synthetic core'' indicator helps achieve significantly better forecasts over a 12-month horizon than the AR model for total inflation for the final part of the sample. We also show that the results are rather robust to potential data-snooping.
Catherine Bruneau, Olivier De Bandt and Alexis Flageollet
Classification JEL : C33, C53, E37.
Keywords : inflation, out-of-sample forecast, indicator models, dynamic factor models, Phillips curve, euro area, data snooping
Updated on: 06/12/2018 10:59