Latest version, October 2021
We quantify the unemployment-risk channel in business-cycle fluctuations, whereby an initial contractionary shock is amplified through workers reducing their demand in fear of unemployment. We document two stylized facts on how unemployment and unemployment risk respond to identified demand and supply shocks in US data. First, separation and job-finding rates play similar important roles in accounting for the overall unemployment response. Second, separations are more important early on, while job-finding rates respond with a lag. We show how a tractable heterogeneous-agent new-Keynesian model with a frictional labor market matches both facts once we include endogenous separations and sluggish vacancy creation. Relative to a model with exogenous separations and free entry, our framework attributes almost twice as large a share of output fluctuations to the inefficient unemployment-risk channel, and thus gives a larger role to stabilization policy.
First version, December 2020
NBER working paper, December 2020
We formulate an economic time use model and add to it an epidemiological SIR block. In the event of an epidemic, households shift their leisure time from activities with a high degree of social interaction to activities with less, and also choose to work more from home. Our model highlights the different actions taken by young individuals, who are less severely affected by the disease, and by old individuals, who are more vulnerable. We calibrate our model to time use data from ATUS, employment data, epidemiological data, and estimates of the value of a statistical life. There are qualitative as well as quantitative differences between the competitive equilibrium and social planner allocation and, moreover, these depend critically on when a cure arrives. Due to the role played by social activities in people's welfare, simple indicators such as deaths and GDP are insufficient for judging outcomes in our economy.
CBS working paper, December 2020
I introduce a simple model which endogenously generates a Pareto distribution in top earnings, consistent with empirics. Workers inhabit different niches, and the earnings of a worker is determined by the niche-specific supply of labor and a constant-elasticity labor-demand curve. The highest paid workers are the ones that inhabit a niche with few other workers. A Pareto tail in earnings emerges as long as the distribution of workers over niches satisfies a regularity condition from extreme-value theory, satisfied by virtually all continuous distributions in economics.
(R&R at the International Economic Review, previously circulated as Harmenberg and Sievertsen)
Latest version, November 2017
I use Danish administrative data 1980-2013 to study the underlying mechanisms generating fluctuations in income-shock moments. I partition the population into 37 narrowly defined educational categories and document the cyclicality of labor-income shock moments for each category separately. For the individual educational categories, mean income growth is strongly correlated with income-growth skewness, with an average correlation of 0.87 − 0.88. The connection between income-growth skewness and mean income growth is not only strong in the time dimension, but also in the cross section. Across the 37 educational categories, the correlation between mean income growth and income growth skewness is 0.93 − 0.96. Labor-market frictions together with variations in productivity growth can generate the relationship between mean income growth and income growth skewness. In a quantitative job-ladder model, variations in productivity growth quantitatively capture both the time-series and cross-sectional relationship.