Best paper award at the 19th GEP/CEPR Annual Postgraduate Conference
Robotics has been substituting or complementing workers in a wide range of occupations. To examine the strength of this substitutability, I match unique data on imported robot prices with the occupational task information to measure the cost of using robots by occupation. The data show that a 10% reduction in the cost is associated with a 1.2% reduction in wages for production and transportation occupations in the US, suggesting strong substitutability in these occupations. This finding motivates the development of a model in which robots are traded and can substitute for labor with different elasticities of substitution across occupations. Using a model-implied optimal instrumental variable, I estimate a higher elasticity of substitution between robots and workers than that of general capital goods in production and transportation occupations. These estimates imply that the adoption of industrial robots significantly affects wage polarization in the US.
We study the role of multinational enterprises (MNEs) on the labor share in the source country. A unique natural experiment from the 2011 Thailand Floods, which forced Japanese-MNEs plants to halt operations, is employed. This foreign productivity shock leads to a relative decrease in domestic employment and fixed assets of the MNEs affected by the flood, with a stronger effect on the latter. We propose a heterogeneous firm GE model that features a production function with offshore factor inputs and an ``extensive margin hat algebra'' method to solve the model quantitatively without observing the cost savings of marginal offshorers. We estimate the elasticity of substitution between home labor and foreign inputs by relating the home and foreign factor demands to the flood shock. The estimated model indicates that foreign factor productivity growth increased capital demand in Japan more than labor demand, reducing the labor share in Japan by 2.26 percentage points from 1995-2007.
We construct commuting zones (CZs) in Japan using the inter-municipality commuting patterns observed in the 1980-2015 Population Census. We employ the hierarchical agglomerative clustering method adopted by Tolbert and Sizer (1996), who defined the standard CZs in the US. As a result, for example, in 2015, from 1,736 municipalities, we construct 265 CZs that are mutually exclusive and collectively exhaustive. We discuss the properties of economic variables within and across the CZs and find that CZs are feasible to capture the heterogeneity that exists across labor markets.
We study the role of an investment promotion policy in adopting industrial robots and firm performances, notably employment. Combining the policy variation in the Tax Credit for Promoting Productivity-Enhancing Equipment Investment (TC-PPEI) in Japan and a newly collected Japanese firm-level longitudinal data on robot adoption, we find that the firms eligible for the TC-PPEI increased the adoption of robots. Our event-study analysis reveals that when firms adopt robots, they do not decrease the total number of workers but significantly increase it after 1-3 years of adoption event as well as sales. Our results suggest that adopting robots can be employment creating instead of destroying at the firm level.
Media Coverage: VoxEU, World Economic Forum
This paper studies the relationship between industrial robots and employment in Japan based on a unique dataset that allows us to calculate the unit price of robots. Our model combines standard factor demand theory with a recent task-based approach to derive a simple estimation equation between robot prices and employment, and our identification strategy leverages heterogeneous applications of robots across industries and heterogeneous price changes across applications. We find that the decline in robot prices increased both the number of robots and employment by raising the productivity and production scale of robot-adopting industries.
This paper is the first empirical study on adverse selection and moral hazard in the corporate disaster insurance market. By constructing and examining a unique plant-level panel dataset on the 2011 Thailand floods, we overcome the general lack of data that has previously prevented a systematic study on the issue. By exploiting unexpected, large losses caused by a severe disaster, we find evidence of adverse selection for both property and business interruption insurance. Moral hazard, measured by impacts on recovery efforts, is also found for both types of insurance, albeit more salient effects for business interruption insurance.