Essays on Labor and Health Economics: The Effect of Amazon Warehouses on Employment and Housing and the Effect of Managed Care for Long-Term Care

The rise and decline of industries can reshape local economies. The recent growth of e-commerce has affected many businesses and localities, with store-based retail operations giving way to warehouse-based online shopping. In the first two chapters of my dissertation, I study how the expansion of the warehousing industry, fueled by the shift toward online retailing, affects local economies.

First, I examine the effects of Amazon warehouses (known as fulfillment centers) on local labor markets. Between 2011 and 2017, the number of Amazon warehouses across the U.S. quadrupled. I focus on the subset of U.S. counties in which the online retailer has either opened or announced its plan to open warehouses since 2011, and use the gradual roll-out of warehouse openings as a source of variation within this sample. Matching information about warehouse openings with county-level employment and earnings from the Quarterly Wage Indicators series, I estimate that opening an Amazon warehouse increases employment in the warehousing industry by about 2,000 jobs. This number is larger than the average number of workers hired by a warehouse in my data, suggesting a small spillover within the industry. For total county employment, I do not find any significant increase except in the fourth quarter of the year. My findings also show average monthly earnings in the warehousing industry decline at least by 2%. Together, these results suggest Amazon warehouses generate low-paying seasonal jobs and have limited ability to boost local employment.

In the second chapter, I explore whether warehouse openings have any general equilibrium effects on local housing markets. I use data from the Zillow Home Value Index to determine the effect of warehouses on housing prices. My results show warehouse openings increase median housing values by 1.5%-2%. Using the ZIP-code-level data on home values, I show that the increase in housing values is concentrated within 11-20 miles of warehouses. I discuss mechanisms that could propagate local shocks to the housing market, especially in the absence of any strong employment effect. One possibility is that Amazon’s entry drives up the demand for industrial space for warehousing and logistics operations, which, in turn, increases land prices in these markets. I provide evidence that an Amazon warehouse increases the number of large warehousing establishments in a county by more than one.

The first two chapters of my dissertation make two contributions. First, I provide causal estimates of the effects of Amazon warehouses. The endogeneity of firm entry makes studying its effect on local markets difficult. To get around this problem, I use the variation in entry timing across locations chosen by Amazon. I also provide insight into Amazon's location strategy, which makes this variation conditionally exogenous. Second, the findings of this paper have important implications for policymakers negotiating with Amazon on subsidy deals for opening warehouses. The main argument in favor of providing subsidies to businesses is the direct and indirect jobs that they could add to the local economy. However, the nature and magnitude of this spillover effect is likely to be different across industries. My work shows the warehousing industry, although an emerging one, has limited scope for boosting local employment.

The third chapter of my dissertation is a joint work with Leora Friedberg and Anthony Webb. We investigate the impact of managed care organizations in the context of long-term care for the elderly population. State Medicaid programs have increasingly contracted with managed care organizations for the efficient delivery of care. We exploit this variation in program roll-out—across and within states—to examine its impact on Medicaid take-up, nursing home use, and health outcomes. We find important shifts in Medicaid use and care use, with sharp declines for single individuals needing assistance with activities of daily living. For this group, the decline in nursing home use is larger than the decline in Medicaid use, whereas for married individuals, Medicaid use does not change and nursing home use rises in some cases. These results are consistent with cream-skimming by managed care organizations, because they may screen out individuals who are likely to be high-cost patients, while also reducing the intensity of treatment for those who remain on the Medicaid rolls. We find a possible moderate reduction in health in the most affected group, but we do not find overall evidence of dire health consequences. Given the increasing need for long-term care, and Medicaid being the chief source of payments, these findings improve our understanding of the role of managed care organizations.