Short-term inflation forecasting in Argentina with Random Forest models

This paper examines the performance of Random Forest models for forecasting short-term monthly inflation in Argentina, especially for the current month or the following. Using a database with indicators on a monthly basis since 1962, it is concluded that these models achieve a forecasting accuracy statistically comparable to the consensus of market analysts surveyed by the Central Bank of the Argentine Republic (BCRA) and to traditional econometric models. One advantage of Random Forest models is that, as they are non-parametric, they allow for the exploration of nonlinear effects in the predictive power of certain macroeconomic variables on inflation. It is found, among other things, that: 1) the relative relevance of the exchange rate gap for forecasting inflation grows when the gap between the parallel and official exchange rates exceeds 60%; 2) the predictive power of the exchange rate on inflation increases when the BCRA's net international reserves are negative or close to zero (specifically, less than USD 2 billion); 3) the relative relevance of lagged inflation and the nominal interest rate to forecast inflation for the following month increases when the level of inflation and/or the level of the interest rate rise.