Investment Dispersion and the Business Cycle
We document a new business cycle fact: the cross-sectional standard deviation of firmlevel investment (investment dispersion) is robustly and significantly procyclical. This makes investment dispersion different from the dispersion of productivity and output growth, which is countercyclical. Investment dispersion is more procyclical in the goodsproducing sectors, for smaller firms and for structures. We show that a heterogeneous-firm real business cycle model with countercyclical idiosyncratic
... ical idiosyncratic firm risk and non-convex adjustment costs calibrated to match moments of the long-run investment rate distribution, produces a time series correlation coefficient between investment dispersion and aggregate output of 0.58, close to the 0.45 in the data. We argue, more generally, that cross-sectional business cycle dynamics impose tight empirical restrictions on the physical environments and the structural parameters of heterogeneous-firm models. JEL Codes: E20, E22, E30, E32. Investment at the micro level is lumpy and infrequent (Doms and Dunne, 1998, Cooper and. The distribution of investment rates is positively skewed and has excess kurtosis (Caballero et al., 1995) . Motivated by these facts about the long-run cross-sectional distribution of micro-level investment, researchers have studied models with non-convex capital adjustment costs and asked whether "getting the micro facts right" matters for aggregate (investment) dynamics. 1 In this paper we show that lumpy investment models have important and thus far unexplored implications for the dynamics of the cross-section of investment rates. Conversely, we show that the joint dynamics of the cross-section of firm-level investment, output growth and productivity growth provide important parameter restrictions for heterogeneous firm models. We start with the data and establish a new business cycle fact: the cross-sectional standard deviation of firm-level investment rates is procyclical. 2 Figure 1 and Table 1 on the next page illustrate this new cross-sectional business cycle fact. The scatter plot reveals a positive correlation between the cross-sectional investment rate dispersion and the cyclical component of aggregate output. Relatedly, a recent literature has documented that, across different countries and data sets, the dispersion of changes in firm-(or plant-) level output, productivity and prices is countercyclical. 3 We find the same negative association with the cycle also for the dispersion of firmlevel employment growth. Table 1 shows that the signs of the correlation between aggregate output and the cross-sectional dispersion of core economic variables are statistically significant at conventional levels (see Appendix A.2 for extensive robustness checks and a time series graph of the investment rate dispersion). These findings together are incompatible with a simple frictionless model of the firm. Such a physical environment implies that the distributions of firms' decision variables comove over the cycle. We propose a heterogenous-firm real business cycle model with fixed capital adjustment costs to explain qualitatively and quantitatively the procyclicality of investment dispersion, even when the dispersion of firm-level productivity growth is countercyclical.