Nowcasting

Nowcasting estimates the present or very near-term state of an economic variable, such as GDP growth or inflation, before the official statistic is published. Government economic data is typically released with a lag of weeks or months and is often revised afterward, leaving a gap nowcasting tries to fill with a timely, if less precise, estimate.

Nowcasting models combine faster-arriving indicators — such as purchasing managers' index surveys, retail card spending, freight and shipping volumes, or other alternative data — using statistical or machine-learning methods to produce a running estimate that updates as new data arrives throughout the quarter or month, rather than waiting for the full official release. This differs from ordinary time series forecasting, which typically projects a known series forward in time; nowcasting instead estimates a value for a period that has already started, or just ended, but has not yet been officially measured.

Central banks, investors, and forecasters use nowcasts to react to economic conditions in real time rather than months later, and several central banks publish their nowcasting models openly, including estimates feeding into readings such as the consumer price index. The main pitfall is that nowcasts are built on proxy indicators whose relationship to the true underlying figure can shift over time, so a model that tracked well historically can drift out of line during unusual periods, such as sudden supply shocks or policy changes.