An Inflation Forecasting Model for the Euro Area

With the European economic integration, the understanding of inflation and inflationary pressures requires to analyse both the national level and the whole Euro area level. This is true in particular for the inflation forecasts that are carried out within the Eurosystem and published four times a year in the ECB Monthly Bulletin. For that purpose, the Banque de France is currently building tools for the Euro area in addition to those already in use for France. The present study puts forward a simple model of short-term developments (one year ahead) in inflation, as measured by the Harmonized Index of Consumer Prices (HICP) of the Euro area. This model does not take into account the feed-back effect of prices on activity, which should be considered in order to analyse medium-term price developments. It could hence be improved along these lines in the future. The model includes seven equations, explaining the total HICP of the Euro area and some of its sector-based sub-indexes (services, manufacturing sector, unprocessed food, processed food, energy and underlying inflation, defined as HICP inflation excluding unprocessed food and energy prices). It uses exogenous variables such as unit labour cost, import deflator, indicators of tightening in the labour market, or in the goods market, and indirect tax indicators. We have favoured an empirical approach rather than a strict compliance with theoretical models, paying particularly attention to the fit of the equations to the data. However, this model is able to provide relevant economic interpretations of recent price developments. Finally, we assess the forecasting performance of the model in traditional in-sample and out-of-sample rolling event evaluations. To do so, the forecasts were compared to the ones obtained from simple autoregressive equations, which are also commonly used to forecast short-term price developments. On the whole, the model provides more accurate forecasts than those provided by the autoregressive model, and a sector-based disaggregated approach outperforms a single equation to forecast total HICP. Part of this result may come from dummy variables that correspond to well identified shocks that improve both the econometric characteristics and forecast performance of the equations of our model.