We study the implications of flexible capacity utilization for firms’ investment and risk in an RBC framework. Capacity utilization improves the model’s fit for numerous investment-related moments. Specifically, flexible utilization reconciles the positive sign of firm-level investment skewness with the data. Without utilization, the model-implied skewness bears the wrong sign. The model predicts an annual risk premium spread between low and high utilization firms of about 5%. We confirm this novel utilization spread in the data. Empirically, utilization explains premia beyond other intensive-margin characteristics. Our results demonstrate the importance of utilization for the joint dynamics of firms’ production and valuation.
Using a sample of the 48 contiguous United States, we consider the problem of forecasting state and local governments' revenues and expenditures in real time using models featuring mixed-frequency data. Among the single-equation models we consider, we find that ADL-MIDAS regressions combining high-frequency economic variables together with low-frequency fiscal data yield forecast gains over traditional models in which all data are included at the same (low) sampling frequency. Among the multi-equation models we consider, we find that low-frequency Bayesian vector autoregressions (BVARs) typically outperform both mixed-frequency Bayesian vector autoregressions (MF-BVARs) and low-frequency vector autoregressions (VARs). When we directly compare the forecasts from ADL-MIDAS regressions to those from MF-BVARs and BVARs, we find that ADL-MIDAS models typically produce the most accurate forecasts of state-level revenues and expenditures. A simulation exercise confirms this conclusion and shows that ADL-MIDAS models typically produce accurate forecast in many empirically realistic settings. Overall, our results show that ADL-MIDAS regressions provide policy makers and market participants with a simple yet reliable tool for forecasting fiscal variables in real time.