Macroeconomic Information and the Relative Predictability of U.S. Sector Returns

Source abstract

A long-standing finding in empirical asset pricing is that equity-return magnitudes are difficult to predict out of sample. But can macroeconomic information still identify which sectors will outperform others and the market? This study examines whether macroeconomic levels and announcement surprises forecast next-month U.S. sector ETF returns relative to SPY, with forecast performance evaluated separately for return magnitude, cross-sectional ordering, and direction relative to the market. A reduced-rank Ridge regression (RRR-Ridge) forecasting system selects its penalty and rank in each rolling window and is typically low-dimensional. The results show little magnitude predictability and modest evidence for relative performance: two of the six primary comparisons are supported after multiplicity adjustment. Announcement-surprise RRR-Ridge improves ordering relative to a rolling-mean forecast and direction relative to sector-by-sector Ridge. Two simple sector-rotation rules, unoptimized translations of the supported comparisons, outperform their matched benchmark’s net of transaction costs. These results are specific to the baseline design: they are not maintained with longer estimation windows or with tuning of ordering or directional accuracy. Within the design, macroeconomic information carries a modest but usable signal about the relative ordering and direction of sector returns.