Abstract: Benchmark models of structural transformation focus on the reallocation of employment across sectors while assuming that overall employment stays constant. We show that this assumption does not match facts for developing economies. We study a panel of 48 mostly developing economies over the period 1990--2018 and document a strong positive relationship between the share of the population employed in agriculture and the overall employment rate. That is, the early part of the development process is associated with a substantial decline in the total employment rate. Motivated by this finding, we extend a benchmark model of structural change featuring Stone-Geary preferences to allow for endogenous labor supply. We show that this model can account for the patterns we document in the data both qualitatively and quantitatively. We use a calibrated version of our model to study the employment dynamics in several developing economies and show that structural change is a quantitatively important source of employment changes during the early stages of development.
Temptation, Self-Control, and the Design of Optimal Unemployment Insurance, with Pei-Cheng Yu
previously titled "Optimal Unemployment Insurance with Costly Self-Control"
Abstract: This paper examines optimal unemployment insurance design when individuals prefer immediate rewards due to temptation and self-control problems. In addition to moral hazard, unemployment benefits also affect search effort through self-control costs. The trade-off between insurance and incentives is weakened by costly self-control. In certain cases, the incentive constraint may not bind, potentially avoiding immiseration and justifying the need for a basic income. Compared to setups without costly selfcontrol, the optimal unemployment insurance features front-loaded consumption upon employment—lending theoretical support to back-to-work bonuses—while consumption for the unemployed is more back-loaded. Our quantitative evaluations of the optimum highlight its potential to significantly enhance welfare.
A Quantitative Analysis of Multidimensional Changes in Unemployment Insurance Policies, with Lei Fang and Jun Nie
Paper download (Updated November 2023; First version July 2020) Reject & Resubmit at JEDC
Abstract: Unemployment insurance (UI) policies along three different policy dimensions—weekly benefit amount, benefit duration, and benefit eligibility–have been used in the United States during past economic downturns. This paper studies the effect of these three types of policies and their interaction on the labor market. While higher weekly benefit amounts and longer benefit duration reduce the individual unemployed worker’s work incentive, expanding the benefit eligibility increases UI recipient rates. At the disaggregated level, the impact of each policy is hump-shaped over a worker’s wage income. We quantify the impact of the CARES Act UI, which combined all three policies during the COVID-19 pandemic. The CARES Act UI package raised the average unemployment rate by 1.61 percentage points during April–December 2020, with two-thirds of the effect coming from the amplification of the economic shutdown policy and the COVID-19 infection risk. Decomposing the CARES Act UI package, the interaction between the three policies was quantitatively important and accounted for one-third of the total effect.
The Short and the Long of It: Stock-Flow Matching in the US Housing Market, with Eric Smith and Lei Fang
Paper download (Updated June 2025; First version October 2022) Conditionally accepted at International Economic Review
Abstract: From 2006 until 2020, the probability of selling a house in the US housing market declined sharply after listing for two weeks. Moreover, sales within the first two weeks of listing ("quick sales") and sales happening afterward ("slow sales") behaved differently over the housing cycle. The probability and associated price of a quick sale recovered from the housing slump sooner, faster, and more prominently than a slow sale. This paper demonstrates that a calibrated stock-flow matching model not only generates quantitatively consistent sales, prices, listings, and time on the market but also captures the distinctions between fast and slow sales over the housing cycle.