Measuring the Predictable Variation in Stock and Bond Returns

Recent studies show that when a regression model is used to forecast stock and bond returns, the sample $$R^2$$ increases dramatically with the length of the return horizon. These studies argue, therefore, that long-horizon returns are highly predictable. This article presents evidence that suggests otherwise. Long-horizon regressions can easily yield large values of the sample $$R^2,$$ even if the populations $$R^2$$ is smaller or zero. Moreover, long-horizon regressions with a small or zero population $$R^2$$ can produce t-ratios that might be interpreted as evidence of strong predictability. In general, the analysis provides little support for the view that long-horizon returns are highly predictable.