Forecasting a monetary aggregate under instability: Argentina after 2001

This paper compares different forecasting approaches for the Argentine monetary aggregate M2, which is a key variable for monetary policy. First, we estimate a conditional equilibrium-correction model of money demand, which is theory consistent and accounts for the main features of the data. Next, we compare its forecasts with those obtained by other methods: a VAR in differences, naïve models, robustified devices, forecasting aggregates using disaggregates, and pooling of forecasts using different models and windows. They are evaluated over an unstable period in which there was often uncertainty about the economic regime. For forecasting the growth rate of M2, it can be useful to complement the equilibrium-correction model with other approaches like univariate AR models, either individually or by pooling.