xuanli zhu

Abstract: This study examines the effects of automation on labor demand, focusing on the power loom adoption in Japan’s early 20th century silk-weaving industry. Exploiting plant-level panel data, we find that, compared to non-adopted plants, power loom adaption increased the employment and wages for adult male workers likely engaged in engineering tasks. Female adults, the main manual workforce, experienced stable employment with wage increases despite displacement and task transitions. However, equilibrium spillover effects from imperfect labor market competition led to a decrease in overall female adult employment in highly mechanized areas, primarily driven by the exit of low-wage plants.

Abstract: This paper introduces a novel method to study the determinants of wage inequality in the labor market, utilizing online job posting data and machine learning algorithms. This method offers new insights by opening the black box of the multidimensional worker and job heterogeneity, which also enables to identify the most important skills and tasks that account for sorting between worker/job and firm components. Applying our method to the posting data of a Chinese online job board, we estimate different wage dispersion components and find clear evidence of firm wage premiums and positive firm-job sorting, with the shares consistent with those observed in many developed countries. During the estimation procedures, our machine learning approach reveals a data-driven skill and task structure featured by different levels of specificity. Variations in occupation-specific skills and tasks are the primary determinants of wage variances, contributing through both channels of job differences and sorting with firm pay policies. This is especially prominent in high-skill occupations, which are characterized by large wage dispersions. Meanwhile, in low-skill occupations, education- and experience-related skills and tasks, which are relatively less specific, play an equal or sometimes more important role. In contrast, variations in the most general skills, whether cognitive or interpersonal, have minimal impact on posted wage inequality.

Abstract: This paper studies the supply of non-wage compensations across different types of firms and jobs, as well as their impact on wage determination. Taking advantage of the data from a Chinese online job board where most firms document pecuniary and nonpecuniary amenities to attract applicants, we extract a large set of non-wage compensations that employers and jobseekers recognize as essential for their matching in the labor market. We find that different firms in different jobs also provide different non-wage compensations in a systematic way. In particular, high wage-premium firms sorted with high skill jobs also more likely provide advanced insurance packages, backloading wage and stock options, professional coworkers, and flexible work-time, and such amenities are positively correlated with posted wage. In contrast, low wage-premium firms sorted with low skill jobs more likely to offer weekend, holiday, and regular work-time, and such amenities are strongly subject to compensating differential. To account for these findings, we propose a new theory that combines the compensating differential with two new forces: efficiency compensation and (productivity-based) firm-worker sorting. This framework not only reconciles the inconsistency between our data and the incumbent theories but also offers important general insights into the labor market inequalities.

Abstract: We study the long-run evolution of establishment dynamism in post-war Japan and document three previously unreported trends. First, establishment entry rates declined persistently from the late 1950s to the late 1990s, while exit rates remained low and stagnant, producing a pronounced aging of Japanese business units. Second, average establishment size fell sharply during the 1960s and 1970s—most notably in manufacturing and construction—before partially recovering for younger establishments in subsequent decades. Third, the average lifecycle growth of cohorts shifted markedly downward in the same period, further dampening market dynamism. Using a standard firm dynamics model calibrated to Japan’s data, we test a range of potential drivers. We find that changes in labor supply growth account for much of the long-term decline in entry rates, albeit via a largely direct mechanism with minimal compositional feedback. Moreover, a moderate reduction in fixed operation costs or in the dispersion of ex-ante productivity helps generate the observed declines in both entrant and incumbent sizes. In contrast, altering entry costs, exit values, or labor market distortions fails to yield realistic predictions in an economy with limited ex-post heterogeneity. Our findings suggest that long-running declines in labor supply growth, falling fixed costs, and narrowing entrant heterogeneity together offer the most plausible explanation for Japan’s enduring transition toward lower market dynamism.

Abstract: This paper investigates the role of labor market institutions in shaping human capital investment and technology adoption by focusing on the IT sectors of Japan and China. Leveraging online job posting datasets, we find that Japanese IT firms are more inclined to offer on-the-job training and rely on older technologies, while Chinese firms prioritize hiring skilled workers and adopting newer technologies. We develop a two-period model that encapsulates the interactions between human capital investment, technological shifts, and labor market frictions. Contrary to conventional models positing that firms and workers are perfect substitutes in investing in human capital, we argue that their roles are asymmetrical and influenced by the technological regime and labor market structure. The model reveals that the rigidity of Japan’s labor market suppresses workers’ incentives but enhances firms’ incentives for human capital investment and technology adoption, despite its technical efficiency, while the fluidity of China’s labor market does the opposite. We further utilize our model to explain how labor market institutions can be both cause and consequence of technology adoption and economic development by suggesting that labor market structures could be aligned with the dominant technological regime for optimal human capital investment, and that a misalignment due to technological change afterwards could lead to significant economic inefficiencies.