I study whether the Opportunity Zones (OZ) program, the largest place-based developer subsidy in U.S. history, generated net new housing supply or merely redistributed construction across space and time. Using a quarterly panel of residential address counts for all U.S. census tracts from 2012 to 2022, I estimate inverse-probability-weighted difference-in-differences models using nearby non-OZ tracts following Abadie (2005). A naive baseline comparison yields a marginally positive treatment effect, consistent with headline estimates in prior work, but reflects two redistribution mechanisms rather than genuine new construction. First, residential address growth slowed in the window preceding QOF opening, as developers apparently waited for formal eligibility before breaking ground, an anticipation effect that artificially depresses the pre-treatment baseline. Second, non-OZ tracts within ten miles of a designated zone experienced a symmetrical decline in address growth after OZs opened, indicating that investment relocated across census-tract boundaries rather than expanding net supply. The net effect of OZ designation on housing supply is statistically and economically indistinguishable from zero once both mechanisms are accounted for, casting doubt on the efficacy of the proposed expansion of OZ-style incentives in the One Big Beautiful Budget Act for stimulating new housing production absent complementary zoning or infrastructure reform.
I examine whether the Opportunity Zones (OZ) program affected rental housing affordability in designated tracts. Place-based development subsidies can affect rents through competing channels: increased investment that expands rental supply and dampens rents, or increased local demand that bids them up. Using a propensity-score-matched difference-in-differences design, I compare rents in OZ tracts to observationally similar non-designated low-income communities (LICs) before and after the program's 2018 opening, controlling for pre-period differences in poverty, employment, rent growth, and home values. Drawing on tract-level data from the American Community Survey, I find no statistically or economically meaningful change in median rents attributable to OZ designation over the 2018–2022 period. The null result is consistent with a literature finding that OZ investment concentrated in high-end real estate rather than expanding affordable rental supply, and suggests the program's tax incentive structure — which rewards capital gains deferral rather than housing output — is poorly suited to improving affordability for zone residents.
This paper investigates the capacity of large language models (LLMs) to approximate the elasticity of taxable income (ETI), a central parameter in public finance. Using simulations of controlled experiments and replications of studies using observational data, we evaluate how LLMs respond to tax schedule changes. Our results show that LLMs reproduce key behavioral patterns observed in human studies, though often with heightened responsiveness, suggesting limited recognition of real-world frictions. Findings suggest LLMs' potential as low-cost, flexible complements to traditional methods for analyzing taxpayer behavior and policy design, although further validation across models and contexts is needed.