Forecasting Inflation in Argentina: A Comparison of Different Models

In general, central banks are concerned with keeping the inflation rate stable while also sustaining output close to an efficient level. Under “inflation targeting”, forecasts of the evolution of the general price level are an essential input for policy decisions and these are usually released in quarterly “Inflation Reports”. The costs and benefits of transparency in monetary policy are widely debated, but the need for a central bank to incorporate forecasts of future inflation is broadly agreed. In short, forecasting inflation is of foremost importance to households, businesses, and policymakers. In 2016, the Central Bank of Argentina began announcing and inflation targeting scheme. In this context, providing the authorities with good estimates of relevant macroeconomic variables turns out to be crucial to make the pertinent corrections to reach the desired policy goals. This paper develops a group of models to forecast inflation in Argentina and conducts a comparison of their predictive ability at different horizons. Our variety of models includes: (i) univariate time series models, (ii) VARs, Bayesian VARs and Time-Varying Parameter VARs, and (iii) conventional New Keynesian Phillips Curves including one that incorporates money to evaluate its information content as a predictor of inflation. We compare the predictive performance of the different methods using the Giacomini-White test over the relevant horizons for monetary policy decisions.