working papers
Changes in Employment and Wages in the United States, 1980–2019: Skills, Occupations, and Technology
NBER Working Paper 23283, 2017
Changes in U.S. employment and wages since 1980 pose a joint puzzle: routine employment contracted, but relative wage growth was strongest among high-skill workers and stronger still among managers. Unlike other high-paying worker groups, managers experienced rising relative wages alongside a declining employment share after 2000. We construct a tractable model of skills, occupations, and sectors in which individuals differ in worker and managerial skills, and an additive occupational requirement governs comparative advantage among workers. The equilibrium connects worker assignment, wage differences among workers, managerial earnings relative to workers, and sectoral reallocation. Under explicit sufficient conditions, we derive analytical comparative statics for these margins. In the 1980–2019 quantitative exercise, task-productivity changes account for much of employment reallocation but produce too little high-skill and managerial wage growth and too much managerial employment growth. Changes in requirements and skill supply help reconcile employment and wages. Occupational task characteristics support the distinction between productivity and effective-skill changes. An illustrative AI scenario shows that abstract relative-wage growth can be higher even when abstract employment grows more slowly than under a continuation benchmark.
*html and slides are based on earlier versions
Transforming Institutions: Labor Reallocation and Wage Growth in a Reunified Germany
Working Paper, 2021
How do institutions affect economic performance? We exploit a unique historical episode, the German Reunification, to investigate how this radical change transformed East Germany’s labor market allocation, igniting wage growth in the early years after reunification. Using matched employer-employee data constructed from the universe of German social security records, we show that the sharp growth in East German wages strongly correlates with a rapid reallocation of workers across plants within East Germany. Moreover, reallocation was disproportionately larger among older cohorts, suggesting that longer exposure to communist institutions led to more severe misallocation: In a competitive market, these older workers would have switched jobs or been fired at a younger age. By the same token, only East German plants that already existed at the time of reunification display different reallocation patterns compared to their Western counterparts: Large plants downsize, indicating that they had previously been inefficiently large, while all plants experience significant levels of worker turnover. We find that plants with larger levels of reallocation experience larger wage growth. This provides rare, direct empirical evidence that the reallocation of workers-both within and across plants-spurred by new labor market institutions, was consequential for wage growth.
The Option Value of Human Capital: Higher Education and Wage Inequality
NBER Working Paper 21724, 2017
Going to college is a risky investment in human capital. However, two options inherently embedded in college education mitigate the risk: (i) college students can quit without completing four-year degrees after learning about their post-graduation wages and (ii) college graduates can take jobs that do not require four-year degrees (i.e., underemployment). These options reduce the chances of falling in the lower end of the wage distribution as a college graduate, rendering standard mean-variance calculations misleading. We show that the interaction between these options and the rising wage dispersion, especially among college graduates, is key to understanding the muted response of college enrollment and graduation rates to the substantial increase in the college wage premium in the United States since 1980. Furthermore, expanding subsidies to induce more students to attend college has a negligible net benefit: Due to selection at the enrollment stage, marginal students who only enroll thanks to larger subsidies are far more likely to drop out of college or become underemployed even with a four-year degree, implying small wage gains from college education.
published papers
The Causal Effect of Parents' Education on Children's Earnings
International Economic Review, 2026
We develop and estimate a model of endogenous schooling and earnings to isolate the causal effect of parents' education on children's outcomes. Identification uses earnings differences among children with equal schooling whose parents have different schooling. Health and Retirement Study data indicate that the observed intergenerational schooling correlation is largely driven by correlated unobserved preferences for schooling. This is partly offset by a slightly negative structural relationship between parents' and children's schooling choices. Nevertheless, an exogenous one-year increase in parents' schooling raises children's lifetime earnings by 2 percent on average.
COVID-19 Doesn't Need Lockdowns to Destroy Jobs: The Effect of Local Outbreaks in Korea
Labour Economics, 2021
Unlike most countries, Korea did not implement a lockdown in its battle against COVID-19, instead successfully relying on testing and contact tracing. Until the summer of 2020, only one region, Daegu-Gyeongbuk, had a significant number of infections, traced to a religious sect. This allows us to estimate the causal effect of the outbreak on the labor market using difference-in-differences. We find that a one per thousand increase in infections caused a 2 to 3 percent drop in local employment in the early spring. We also find that employment losses caused by local outbreaks in the absence of lockdowns were (i) mainly due to reduced hiring by small establishments, (ii) concentrated in the accommodation/food, education, real estate, and transportation industries, and (iii) worst for economically vulnerable workers who are less educated, young, in low-wage occupations, and on temporary contracts, even controlling for industry effects. These patterns are similar to what we observed in the US and UK: The unequal effects of COVID-19 were the same with or without lockdowns. Our findings are consistent with the lifting of lockdowns having led to only modest employment recoveries in the US and UK, absent larger drops in infection rates.
Inequality of fear and self-quarantine: Is there a trade-off between GDP and public health?
Journal of Public Economics, 2021
We construct a quantitative model of an economy hit by a pandemic. People choose occupations and make work-from-home decisions to maximize income and minimize their fear of infection. Occupations differ by wage, infection risk, and the productivity loss when working from home. The model is calibrated to South Korea (SK) and the United Kingdom (UK) to compare SK's intensive testing and quarantine policy against UK's lockdown. We find that SK’s policies would have worked equally well in the UK, dramatically reducing both deaths and GDP losses. The key contrast between UK’s lockdown and SK's policies was not in the intensity of testing, but weak restrictions on the activity of many (UK) versus strict restrictions on a targeted few (SK). Lockdowns themselves may not present a clear trade-off between GDP and public health either. A premature lifting of the lockdown raises GDP temporarily, but infections rise over time and people voluntarily choose to work from home for fear of infection, generating a W-shaped recession. Finally, we find that low-skill workers and self-employed always lose the most from both the pandemic itself and containment policies.
Who is afraid of machines?
Economic Policy, 2020
We study how machines, such those that embody information and communication technologies (ICT), software, and industrial robots, affect the demand for workers of different education, age, and gender. We do so by exploiting differences in the composition of workers across countries, industries and time. Our dataset comprises 10 high-income countries and 30 industries, which roughly span their entire economies, with annual observations over the period 1982–2005. The results suggest that software and robots reduced the demand for low and medium-skill workers, the young, and women, especially in manufacturing industries; but raised the demand for high-skill workers, older workers and men, especially in service industries. These findings are consistent with the hypothesis that automation technologies, contrary to other types of capital, replace humans performing routine tasks. We also find evidence for some types of workers having shifted away from such tasks.
Entrepreneurs, managers and inequality
Income concentration in the U.S. rose sharply since the 1970s. But the share of wealth held by the top 1 percent increased less. This can be partially accounted for by a quantitative model of occupational choice, in which rich individuals choose to become entrepreneurs or managers. Collateral constraints induce entrepreneurs to hold more wealth, while managers earn higher wages as a result of competitive assignments to firms. Declining tax progressivity from 1970 to 2000 replaces top entrepreneurs with top managers, which can account for 65% and 30% of the increase in the share of wages and income earned by the top 1 percent, respectively. At the same time, the share of wealth held by the top 1 percent remains stable, as entrepreneurs decumulate but managers accumulate wealth.
On the Intergenerational Transmission of Economic Status
Journal of Political Economy, 2019
We present a model in which human capital investments occur over the life-cycle and across generations, à la Becker and Tomes (1986), also featuring incomplete markets and government transfer programs. The human capital technology features multiple stages of investment during childhood, a college decision, and on-the-job accumulation. The model can jointly explain a wide range of intergenerational relationships, such as the intergenerational elasticities (IGE) of lifetime earnings, college attainment and wealth, while remaining empirically consistent with cross-sectional inequality. Much of life-cycle inequality is determined early in life, which in turn is explained in large part by parental background. The model implies that this is mainly due to early investments in children made by young parents, so life-cycle constraints these parents face are important for understanding the persistence of economic status across generations. Education subsidies, especially early on, can significantly reduce the intergenerational persistence of economic status.
Computerizing industries and routinizing jobs: Explaining trends in aggregate productivity
Journal of Monetary Economics, 2018
Complementarity across occupations and industries implies that the relative size of those with high productivity growth shrinks, reducing their contributions toward aggregate productivity growth and thereby resulting in its slowdown. This force, especially the shrinkage of occupations with above-average productivity growth through "routinization," was present since the 1980s. Through the end of the 1990s, it was countervailed by the extraordinary productivity growth in the computer industry, of which output became an increasingly more important input in all industries ("computerization"). It was only when the computer industry's productivity growth slowed that the negative effect of routinization on aggregate productivity became apparent.
Economic Policy and Equality of Opportunity
The Economic Journal, 2018
We employ equality of opportunity (EOP) definitions from the literature on distributive justice to a quantitative model featuring intergenerational human capital investments and luck. When calibrated to the U.S., the model-implied degree of EOP differs substantially depending on whether one consider it ethical to reward offspring for the effort of previous generations. Despite reducing intragenerational inequality, education subsidies do little to promote EOP. This is because if one thinks intergenerational investments should be rewarded, there is little room for improvement to begin with; In the opposite case, much stronger redistribution is needed for the policies to have a quantitative impact.
other work
EEA Professional Climate Survey Report
EEA, 2025
The European Economic Association (EEA) Minorities in Economics (MinE) Committee, together with the German Economic Association (VfS), conducted a professional climate survey in 2023. The survey targeted current and former members of the EEA with a focus on equality, diversity, and inclusiveness, particularly across the dimensions of gender, ethnicity, LGBTQ+ status, and disability. The aim was to assess the professional climate for minorities within the economics profession across Europe. The effort is comparable to the American Economic Association (AEA) professional climate surveys, against which we also compare our results below.
The Political Economy of Early COVID-19 Interventions in U.S. States: Comment
Who Should Work from Home during a Pandemic? The Wage-Infection Trade-off
Federal Reserve Bank of St. Louis Review, 2022
Shutting down the workplace is an effective means of reducing contagion, but can incur large economic losses. We construct an exposure index, which measures infection risks across occupations, and a work-from-home index, which gauges the ease with which a job can be performed remotely across both industries and occupations. Because the two indices are negatively correlated but distinct, the economic costs of containing a pandemic can be minimized by only sending home those jobs that are highly exposed but easy to perform from home. Compared to a lockdown of all non-essential jobs, the optimal policy attains the same reduction in aggregate exposure (32 percent) with one-third fewer workers sent home (24 vs. 36 percent) and with only half the loss in aggregate wages (15 vs. 30 percent). A move from the lockdown to the optimal policy reduces the exposure of low-wage workers the most and the wage loss of the high-wage workers the most, although everyone's wage losses become smaller. A constrained optimal policy under which health workers cannot be sent home still achieves the same exposure reduction with a one-third smaller loss in aggregate wages (19 vs. 30 percent).
Hit Harder, Recover Slower? Unequal Employment Effects of the Covid-19 Shock
Federal Reserve Bank of St. Louis Review, 2021
The destructive economic impact of the Covid-19 pandemic was distributed unequally across the population. Gender, race and ethnicity, age, education level, and a worker's industry and occupation all mattered. We analyze the initial negative effect and the lingering effect through the recovery phase across demographic and socio-economic groups. The initial negative impact on employment was larger for women, minorities, the less educated, and the young, even after accounting for the industries and occupations they worked in. By November 2020, however, the differential impact between men and women, and between education and age groups has vanished. Across race and ethnic groups, Hispanics and Asians were the worse hit but made up for most of the lost ground, while the initial impact on Blacks was smaller but recovery slower.
Industrial and Occupational Employment Changes During the Great Recession
Federal Reserve Bank of St. Louis Review, 2017
The United States labor market contracted sharply during the Great Recession. The ensuing recovery has been sluggish and, by some measures, still incomplete. In this paper, we break down aggregate employment during the Recession and the recovery into changes across industries and occupations. There is a clear asymmetric pattern: The contraction is driven by sectors, and the recovery by occupations. In particular, the contraction between 2008 and 2010 primarily reflects a steep decline in construction employment, partially mitigated by expansions in the food services, education, and health industries. The recovery first came from a gradual increase in low-skill occupation employment across all sectors, but after 2012, from a pronounced increase in high-skill occupation employment across all sectors. This pattern of recovery is a continuation of the underlying trend of polarization across occupations, which commenced in the 1980s.