Effects of contemporaneous aggregation on the predictive power of information variables

Abstract

Obtaining reliable forecasts of the future path of inflation is crucial for inflation targeting. The importance of information variable approach in forecasting inflation is widely stated. Studies that explore the information content of money basically rely upon the Granger-causality tests and use contemporaneously aggregated time series. For a potential information variable such as a monetary variable, failing to find predictive power for a contemporaneously aggregated series such as the inflation level does not necessarily imply that the potential information variable lacks predictive power. This letter shows that, provided that such a causal relationship is valid for, at least, one of the components of the aggregate, then it is possible to obtain better forecasts of the aggregate using a disaggregated model.