Can Macroeconomists Get Rich Nowcasting Output Gap Turning Points with a Simple Machine-Learning Algorithm?

Abstract

To nowcast output gap turning points, probabilistic indicators are created from a simple and transparent machine-learning algorithm known as Learning Vector Quantization. The real-time ability of the indicators to quickly and accurately detect economic turning points in the United States and in the euro area is gauged. To assess the value of the indicators, profit maximization measures based on trading strategies are employed in addition to more standard criteria. When comparing predictive accuracy and profit measures, the bootstrap based model confidence set procedure is applied to avoid data snooping. A substantial improvement in profit measures over the benchmark is found: macroeconomists can get rich nowcasting output gap turning points.

Keywords: Learning Vector Quantization, Economic turning points detection, Profit maximization measures, Model Confidence Set

JEL Classification: C32, C45, C53, E32, E37, G11, G12