Working Papers

Entrepreneurial Investment Dynamics and the Wealth Distribution (Job Market Paper) [Link to most recent draft

Abstract: This paper studies how entrepreneurial capital illiquidity and incomplete financial markets interact to influence entrepreneurial savings and investment behavior, and in turn shape the wealth distribution. To that end, I construct an endogenous occupational choice model, in which an entrepreneur holds two assets: liquid risk-free assets and illiquid capital. I calibrate the model to match salient facts about entrepreneurial investment drawn from a nationally representative panel survey on startups, and find that an entrepreneurship model disciplined only by microdata on investment can broadly match the high wealth dispersion in the data. I also find that entrepreneurs face severe illiquidity risk, which leads to an increase in precautionary liquid savings and an allocation of illiquid capital toward rich entrepreneurs who have low productivity. Moreover, illiquidity risk depresses overall wealth dispersion due to its distortion of the ability of high productivity entrepreneurs to efficiently accumulate wealth. Consequently, I find that a policy that taxes returns to liquid assets to provide partial insurance against illiquidity risk can be welfare improving, but also increases wealth dispersion.

“Capital-Reallocation Frictions and Trade Shocks”, joint with Andrea Lanteri and Pamela Medina [Link to most recent draft]

Abstract: What are the short- and medium-run aggregate effects of an international-trade shock that increases competition for domestic manufacturing industries? In this paper, we address this question by combining detailed firm-level investment data from several manufacturing industries in Peru, data on the import penetration of Chinese manufacturing goods in Peru, and a quantitative general-equilibrium model of trade with heterogeneous firms subject to idiosyncratic shocks. We find evidence of large frictions in disinvestment (i.e., capital reallocation), that induce high persistence of low returns from capital at the firm level, and dampen the empirical response of reallocation to Chinese import competition. In our calibrated model, these frictions play a key role in the dynamic response of the economy to an import-competition shock, inducing slow transitional dynamics and several years of low aggregate productivity, while the distribution of firm-level capital and productivity adjusts to the new steady state.

Uncertainty Shocks and Entrepreneurship” [Link to most recent draft]

Abstract: Aggregate startup rates are strongly and negatively correlated with economic policy uncertainty, while exit rates are only very weakly and positively correlated with uncertainty. This observation can be reconciled within a model of endogenous entrepreneurial choice, where entrepreneurial capital is illiquid and thus partially irreversible. Heightened uncertainty leads to a delay in entry due to a real options effect, but simultaneously raises exit propensities for poorly performing entrepreneurs and lowers exit propensities for others. At the aggregate level, this interaction of lowered entry propensities and slightly higher exit propensities depresses the size of the entrepreneurial sector while increasing aggregate labor supply for many years. This in turn leads to recessions that are deeper and more prolonged relative to firm-dynamics models where the extensive margin of adjustment did not exist. The results suggests that business-cycle models that abstract from the extensive margin might substantially under-estimate the impact of uncertainty shocks.

Work in Progress

“Illiquidity Risk, Partial Risk Sharing, Entrepreneurship, and the Wealth Distribution”

 “Should Local Regions Promote Large Firms or Small Startups? Trade-offs Between Scale and Market Power in Local Labor Markets”

Technical Notes

“A Fast and Low Computational Memory Algorithm for Computing Distributions in Heterogeneous Agent Models” [Link to paper]

Heterogeneous agent models in macroeconomics generally require numerical computation of the cross-sectional distribution of agents, either in steady state or across time. One standard approach is to utilize “non-stochastic simulation”, as described in Young (2010). This method requires an approximation of the Markov kernel that iterates the distribution forward in time. The textbook method calls for a full approximation of the Markov kernel as a Markov transition matrix, which can be costly in terms of computational time and memory when the state space is large. This paper proposes an alternative simple method that requires much less computational memory, and is substantially faster than the standard method in the case of models with large state spaces.