Predicting relative forecasting performance : An empirical investigation

Abstract

The relative performance of forecasting models changes over time. This empirical observation raises two questions: is the relative performance itself predictable? If so, can it be exploited to improve forecast accuracy? We address these questions by evaluating the predictive ability of a wide range of economic variables for two key US macroeconomic aggregates, industrial production and inflation, relative to simple benchmarks. We find that business indicators, financial conditions, uncertainty as well as measures of past relative performance are generally useful for explaining the relative forecasting performance of the models. We further conduct a pseudo-real-time forecasting exercise, where we use the information about the conditional performance for model selection and model averaging. The newly proposed strategies deliver sizable improvements over competitive benchmark models and commonly used combination schemes. Gains are larger when model selection and averaging are based on financial conditions as well as past performance measured at the forecast origin date.