Forecasting Low Frequency Macroeconomic Events with High Frequency Data

Abstract

High-frequency financial and economic activity indicators are usually time aggregated before forecasts of low-frequency macroeconomic events, such as recessions, are computed. We propose a mixed-frequency modelling alternative that delivers high-frequency probability forecasts (including their confidence bands) for these low-frequency events. The new approach is compared with single-frequency alternatives using loss functions adequate to rare event forecasting. We provide evidence that: (i) weekly-sampled spread improves over monthly-sampled to predict NBER recessions, (ii) the predictive content of the spread and the Chicago Fed Financial Condition Index (NFCI) is supplementary to economic activity for one-year-ahead forecasts of contractions, and (iii) a weekly activity index can date the 2020 business cycle peak two months in advance using a mixed-frequency filtering.