Probability Forecasting for Inflation Warnings from the Federal Reserve

Forecasting inflation constitutes a primary task of monetary policymakers. The US Federal Reserve and other central banks occasionally publish verbal warnings about extreme in ation events. In this paper, we set out a formal framework in which monetary policymakers communicate in ation event warnings to the public. We show that these warnings require probabilistic forecasts in the presence of an asymmetry between the costs of anticipated and unanticipated in ation. In our application, we combine the predictive densities produced from Vector Autoregressive (VAR) models utilizing prediction pools and evaluate our ensemble probabilistic forecasts for quarterly US data from 1990:1 to 2012:1. We adopt a cost-loss ratio for forecast evaluation consistent with our policymaking communication framework. The incidence of low in ation events has increased during the Great Recession and, for these events, economic loss could be reduced by up to 30% relative to a benchmark model in which in ation follows an integrated moving average process. Calibrating our framework to the actual verbal statements from the Federal Reserve during the Great Recession implies a performance advantage of around 5% to 10%. Our findings indicate considerable scope for using formal forecasting methods to forewarn the public of in ation events during the Great Recession.