Abstract: What determines locally traded consumption bundles? Utilizing rich data on cell-phone based residence-to-establishment flows and establishment-match electronic transactions, I address this question with unsupervised machine learning and a novel quantitative model. The machine learning exercise identifies local consumption heterogeneity using over 100 ``in-person" industries, finding a subset of urban neighborhoods have consumption bundles matching remote rural communities. I then augment a standard commute-based general equilibrium model with: discrete, possibly infinite varieties; firm shutdown from entry costs; inferior and luxury goods from nonhomothetic preferences; and consumption-related travel. Investigating local product variety determinants, I estimate two demand-side and two supply side policy counterfactuals using the machine learning identified neighborhoods to target place-based policy. Across multiple intervention intensities, I find reducing income inequality (altering consumer preferences) improves consumption equality for most locations while least welfare improving, however, subsidizing entry costs is more welfare improving and has a ``Goldilocks" zone for consumption access equality.
Abstract: I analyze neighborhood-level per-capita consumption bundles using the highly granular details of cellphone flow data. I show that while cities do have more consumption varieties available than rural communities, several urban neighborhoods have consumption bundles as if they were located in remote, rural region. Neighborhood groupings are identified using an unsupervised machine learning clustering algorithm on per-capita consumption patterns on over 100 consumption industries. The "correct" number of neighborhood groupings is determined using a statistic-based selection procedure. I then utilize establishment-matched electronic transaction data to provide new consumption-price elasticity of substitution estimates that utilizes establishment distance variation for identification for the industries of study. I conclude the paper with a discussion on when granularity, heterogeneity, and the spatial dimension play an important role using a model, patterns in the data, and borrowed estimates of income elasticity and consumption travels sensitivity from my job market paper and the literature.
Abstract: I examine the relationship between the United States' labor market shift towards high-skilled workers and the increase in city-level income inequality that started in the 1980s. I use the Dotcom bubble as a natural experiment, which is a shock to a subset of high-skilled workers. Using a differential exposure difference-in-differences model, I show that a one percent increase in exposure to the bubble led to a 6% increase in wage inequality and a 9% increase in total personal income inequality at the metropolitan-level and a within metropolitan 5% increase in household inequality at the census tract-level. 50% of the estimate disparity can be explained by capital income. A decomposition suggests that the increase in income inequality was driven more by computer-related employment than wages and that there were spillovers to other industries. Tightly estimated null migration results suggest occupational switching occurred. The null migration result likely explains the tightly estimated null housing market results. The last two findings contradict the literature, which instead uses general shocks to high-skilled labor demand, providing a useful avenue for future research.
(Previous name: John Julian Patrick Wade)
Wade, John J. P. (2016). "The Effect of Health Insurance Expansion on Mortgage Performance," (Resting Paper).
Gilbert, Ben and Wade, John J. P. (2016) "Health Care and the Housing Crisis," Social Science Research Network. available at ssrn.com