This paper demonstrates that correlated idiosyncratic fluctuations among firms are an important source of macroeconomic volatility. Standard methods based on cross-sectional demeaning obscure this channel by mechanically driving average pairwise correlations to zero asymptotically. We develop a nonparametric bounds approach quantifying the contribution of cross-firm comovement within industries directly from firm-level data. Applied to the United States economy, we find that such clustered comovement explains 10-15% of GDP volatility in normal times, 15-30% in downturns, and up to 45% following the Great Recession, while also helping account for the Great Moderation and its undoing.
This paper studies variation in healthcare expenditures across time and space using a model of structural transformation tailored to the healthcare sector. It addresses a key tension: cross-country and time-series data show the health spending share rising with overall consumption, suggesting an aggregate income elasticity above one, yet within countries health spending is flat across household incomes, and quasi-experimental spending responses to oil-driven income shocks rise less than one-for-one with income. The calibrated model reconciles this tension in two ways: First, health spending is equalized across incomes by the government or insurers. Second, with an own-price elasticity greater than one in absolute value, positive shocks outside healthcare such as oil income shocks generate price effects that drive the health share down. I find that rising incomes are indeed the dominant driver of healthcare spending, and that reforms that raise the productivity of the healthcare sector will drive its spending share up.
A structural model featuring ICT as a distinct type of capital is used to study how AI-driven surges in ICT investment could affect the U.S. natural interest rate, R∗. The analysis reveals three key insights. First, estimates show that labor and ICT capital have historically been gross substitutes, and their elasticity of substitution has increased over time. Second, expansionary ICT investment-specific technology shocks affect R∗ through intertemporal resource reallocation: the natural rate initially falls and then rises as future income increases. This mechanism is amplified when ICT capital and labor are more substitutable. Third, the analysis considers plausible scenarios for ICT investment under permanent AI-related shocks. ICT deepening in line with its historical trend would raise R∗ only modestly, whereas an ambitious AI scenario adding stronger productivity growth and an upward shift in ICT-labor substitutability generates a much larger medium-to-long-term increase in R∗.
