Job Market Paper

  • Growth and Labor Reallocation: Vertical versus Horizontal Innovation [paper] [slides]

Presented in Society for Economic Dynamics Meeting 2021 (Minneapolis), Chicago-Fed Fellow Seminar 2021, and European Winter Meetings of the Econometric Society 2021 (Barcelona)


Labor reallocation is often regarded as an essential component of productivity growth, based on the thought that innovation is made by firms expanding to new business areas with better technology (horizontal innovation). I study the relationship between long-run growth and labor reallocation by introducing innovation made by incumbent firms previously operating in the product market (vertical innovation). I use novel data from a German establishment-level panel survey to distinguish innovation types from different survey questions about innovation activities and construct an endogenous growth model with different types of innovation to decompose the relative contribution of innovation types on long-run growth. I calibrate the model with the survey data merged with the administrative employer-employee matched data and use the calibrated model to quantify the aggregate effect of the Hartz reform (which increased labor supply and reduced labor market rigidity in Germany during the mid-2000s). I find out that vertical innovation that does not cause direct creative destruction accounts for more than 90% of productivity growth on the balanced growth path. Moreover, vertical innovation generates 85% of dynamic gains from faster productivity growth after the reform. Since vertical innovation has a smaller effect of reallocating labor inputs toward innovating establishments, despite the sharp increase in output growth, labor reallocation rates, especially job destruction rates, do not increase but decrease over the transition path.

Working Papers

  • Rising Skill Premium and the Innovation on Tools [paper] [slides] joint with Younghun Shim


This paper measures innovation on tools used by different occupations and studies its impact on the increasing skill premium. First, we match the description of tools from Wikipedia with patent text data using textual analysis to measure the innovation on tools. Then, we study its relation with the labor market variables at the occupation level. We find 1) innovation on tools grew more in skill-intensive occupations. 2) it is positively associated with wage and employment growth across occupations. 3) it is positively correlated with the skill premium and skill intensity growth within each occupation. Motivated by this reduced-form evidence, we build a model where tool innovation increases the demand of occupations, potentially more for skilled workers. Parameters are estimated through the Generalized Method of Moments. We find that tool innovation accounts for 61% of the total demand factor that contributed to the skill premium increase in 1980-2015.

  • Labor Market Sorting over the Business Cycle [paper] [slides]

Presented in Midwest-Macro Fall 2019 (Michigan State) and Perspectives on (Un-)Employment (IAB)


I show selective separations from the job match at the onset of the recession affect the job creation decision of employers. Since vacant jobs are more likely to be matched with unemployed workers, cyclical variation in the composition of unemployed workers changes the expected return to post vacancies. Using data from the Current Population Survey, I first show that workers with short tenure and high wages are more likely to lose their jobs and become unemployed in recession periods. Moreover, the weekly history of workers from National Longitudinal Surveys suggests that workers having an experience of separation into unemployment for more than four weeks have larger separation probabilities later in their career. Moreover, the gap in separation probabilities between workers with and without separation experience becomes wider in recessions, which suggests an abundance of workers with small expected match surplus, or high separation rates, in the unemployment pool in bad times. I introduce worker heterogeneity of outside values into the Diamond-Mortensen-Pissarides model with endogenous separation to see the macroeconomic implications of worker sorting through employment status over the business cycle. The key mechanism is that the firms tend to separate from their low surplus employees more frequently during recessions, entrants take this into account, and they reduce the number of vacancies created. Although the compositional change amplifies cyclical fluctuations in job-finding rates and unemployment rates, its quantitative importance is small, compared to underlying forces to create cyclical fluctuations.

Work in Progress

  • Age Sorting in the Labor Market [slides] joint with Toshiaki Komatsu


We examine the interaction between the aging labor force and declining business dynamism. Using matched employer-employee data from Germany, we document the patterns of age sorting between firms and workers, where younger firms hire disproportionately more younger workers. We find that such patterns are not fully explained by sorting based on firm productivity or other observables, such as industry and geographic locations. We then construct a directed search model that features both worker life-cycle and firm life-cycle. The model generates age sorting when younger firms face higher uncertainty about their future productivity. As a result, younger firms are more likely to direct their search toward younger workers who have a higher uncertainty tolerance with longer life spans. Lastly, we take the calibrated model to quantify the impacts of the aging labor force and firm on the labor market outcomes of the current young population. We find that changes in firm entry rates and population structure generate a significant decrease in employment rates among the younger population while increasing employment rates among the older population.