Benjamin Moll

In his earlier work, Moll showed that an important driving factor in determining the aggregate effects of poorly functioning credit markets is the persistence of idiosyncratic productivity shocks hitting producers. Higher persistence leads to smaller steady-state productivity losses and slower transition dynamics. He later reveals with David Lagakos, Tommaso Porzio, Nancy Qian and Todd Schoellman that wages increase twice as much with experience in rich countries compared to poor countries, supporting the claim that human capital accumulation plays a significant role in explaining cross-country income differences.[4]

In 2018, Moll together with Greg Kaplan and Gianluca Violante coin the term HANK (Heterogeneous Agent New Keynesian) model to describe the rising literature incorporating household heterogeneity into New-Keynesian models. They argue that monetary policy operates mostly via general equilibrium effects on the labor market, instead of the standard intertemporal substitution channel. This is due to a sizable share of households exhibiting high MPCs, whose spending behavior reacts strongly to changes in disposable income. As Ricardian equivalence fails in HANK models, the reaction of the fiscal authority to a monetary shock is key to determine the overall macroeconomic response.[5]

Moll developed and popularized a number of continuous-time methods for solving heterogeneous agent models. Together with mathematicians Yves Achdou, Jean-Michel Lasry and Pierre-Louis Lions, he recasts general equilibrium models in continuous time using mean-field game theory. The standard Aiyagari Model is boiled down to a system of two main equations: a Hamilton-Jacobi-Bellman equation associated with the optimal decision of the household, and an associated Kolmogorov Forward equation (Fokker-Plank equation) governing the dynamics of the wealth distribution. This recasting allows for analytic solutions when the model is parsimonious enough. Moll and his coauthors popularized finite difference methods for solving numerically those continuous time models, which allows for gains in speed compared to discrete time models.[6]