Inflation forecasting and the crisis: assessing the impact on the performance of different forecasting models and methods

This paper analyses how euro area inflation forecasts have been affected by the financial and economic crisis. Its first objective is to evaluate the accuracy of three representative groups of inflation forecasting models (rules of thumb and benchmark models; autoregressive moving average models; autoregressive distributed lag models) under a direct and an indirect approach, respectively. The second objective of the paper is to study how the absolute and relative forecasting performances of the models and approaches have been impacted by the economic and financial crisis.

The paper finds that direct forecasting models selected on the basis of a penalty function generally dominate simple benchmark models. The analysis furthermore suggests that when an appropriate specification for the component-specific models is found, indirect forecasts outperform the corresponding direct forecasts. Nonetheless, in line with the findings from earlier studies, there are insufficient elements to assert a systematic superiority of one of the two approaches. Concerning the second objective, the across-the-board rise in the forecast errors of all models considered, confirms that inflation forecasting has become substantially more difficult after the onset of the crisis. However, the deterioration of the different models has been uneven: indeed, direct autoregressive distributed lag models and indirect models improved in relative terms during the crisis.