The Impact of Uncertainty on Forecasting the US Economy

ABSTRACT This paper examines the predictive value of uncertainty measures for key macroeconomic indicators across multiple forecast horizons. We evaluate how different uncertainty proxies—economic policy uncertainty (EPU), VIX, geopolitical risk, and measures of macroeconomic and financial uncertainty—enhance forecast accuracy for industrial production, consumer price index, and the federal funds rate using both traditional vector autoregressive models and machine learning approaches. Our results reveal that the marginal predictive power of uncertainty is heterogeneous, varying substantially across variables, forecast horizons, and model specifications. The incremental forecasting value of uncertainty measures that is visible only in a couple of instances becomes statistically insignificant when forecasts are combined.