Working Papers

Business Cycle, R&D, and Hysteresis: Searching for Asymmetries with Hedieh Shadmani, [Updated: February 2026]
In endogenous growth models where the research and development (R&D) level determines the growth rate of productivity, temporary changes in R&D cause permanent level effects, i.e. hysteresis. When positive and negative shocks move R&D symmetrically, an equally sized positive and negative shock produce R&D booms and busts that generate equal and opposite permanent level effects, so their long-run impacts cancel out. Asymmetries break this offsetting, allowing temporary shocks to influence long-run growth. Because private R&D is well known to be pro-cyclical, understanding whether these asymmetries exist is a necessary first step toward assessing how short-run fluctuations in R&D might translate into long-run growth effects. This paper tests for asymmetries in U.S. aggregate private R&D by applying a McKay-Reis duration and steepness analysis to characterize expansions and contractions in R&D, and endogenous threshold regressions to examine state-dependent pro-cyclicality. The main findings are duration asymmetries — longer expansions, although with low statistical significance —, and dependence on labor-market and financial conditions: R&D’s pro-cyclicality intensifies when unemployment is low or credit is abundant. These patterns highlight the importance of boom periods in shaping R&D dynamics and point to channels through which business cycle shocks could generate growth effects.

Super-Robust Endogenous Growth: Theory and Estimation with Pietro Peretto, R&R at Journal of Economic Behavior & Organization [Online presentation]
We propose an endogenous growth model that accommodates increasing, constant, or decreasing aggregate returns to scale with respect to the growth-driving factor: quality-improving knowledge accumulated by firms in-house. When aggregate production is non-linear in firm knowledge, the profitability of firms reflects that property, and entry (new product creation) responds accordingly. The consequent changes in market share offset the non-constant aggregate returns to scale and deliver constant firm-level returns to innovation in steady state. Because returns to innovation are constant, the steady-state growth rate of income per capita is constant and fully endogenous (i.e., dependent on policy parameters). The non-linearity with respect to the growth driving factor has testable implications for convergence dynamics. Specifically, the speed of convergence is decreasing (increasing) in the distance from the steady state when aggregate returns to firm knowledge are increasing (decreasing), causing asymmetric convergence dynamics. This propagation mechanism ensures that, subject to symmetric shocks, the average growth rate across shocks differs from the steady state rate, implying that these shocks’ frequency and magnitude are a determinant of long-run growth. The model reduces the growth dynamics to a single quadratic differential equation in the growth rate of GDP per capita. The quadratic term captures the non-linearity in convergence and uniquely identifies the aggregate returns to firm knowledge. We estimate this equation on a panel of countries in the post-industrial revolution era finding evidence of increasing aggregate returns to firm knowledge.

Turbulent Growth: Business Dynamism and Aggregate Productivity, [Updated: January 2026] R&R at European Economic Review
Turbulence refers to the endogenous reallocation of resources (such as jobs) across firms due to entry, exit, and churning (movements within the firm-size distribution). The paper develops a model of turbulent endogenous growth in which firms invest in in-house innovation to cut costs and gain market share. As firms grow, the marginal return to market share declines due to downward-sloping demand, weakening the incentive to innovate. This mechanism, combined with idiosyncratic shocks, generates endogenous churning while preserving a stationary firm-size distribution. The results are robust to introducing entry and exit, which amplify churning and affect growth through selection. In a counterfactual exercise, I model the observed decline in high-growth startups as a thinning of the right tail of the R&D productivity distribution. While eliminating skewness can generate large reductions in aggregate outcomes, matching its decline explains only 15% of the post-2000 slowdown, suggesting a limited aggregate role for fast-growing startups.