Regulators often delegate enforcement to third-party firms that sell compliance services. Competitive markets for these services can improve buyers’ choices but encourage sellers to help buyers evade regulations. I quantify this tradeoff in the Texas market for passenger vehicle emissions tests, using administrative data from 2002 – 2024 to detect more than 9 million (7%) falsified tests at thousands of firms. An entry event study shows that a 47% increase in nearby competitors raises incumbent cheating by 6.3%. Cheating drivers skip repairs that would prevent $27–$54 worth of annual air pollution per car. An equilibrium model evaluates welfare, capturing bribe setting by spatially-differentiated firms and consumers’ dynamic choices of repairs, cheating, and fair tests. A counterfactual monopolist eliminates cheating but wishes to close locations if it faces economies of scale. Fixing location density near the level regulators typically choose in monopoly markets lowers drivers’ utility by more than 100 times the social benefit of reduced pollution. Doubling firm-level cheating penalties in the status quo enforcement regime reduces cheating by 60.3% at much lower cost to drivers. 

Dockless, shared e-bikes and e-scooters are a rapidly-growing segment of the local transportation market despite being taxed at higher rates per mile of travel than many conventional transportation modes, including motor vehicles. This paper asks what the optimal tax rate for shared bikes and e-scooters would be when accounting for environmental externalities, internalities caused by biased beliefs about accident risk, and redistribution. We derive a sufficient-statistics model that describes how optimal tax rates depend on both extensive-margin (number of trips) and intensive- margin (duration of trips) demand responses to price changes, as well as the income of bike and e-scooter users, internalities, and net externalities relative to other travel modes. We estimate the parameters of this model using trip-level data from a large bike and e-scooter company, and find that in several U.S. cities, use of its vehicles is concentrated among low-income individuals. A nationwide dataset of emergency department admissions shows that e-scooter use generates substantial social costs in terms of healthcare expenditures and lost quality of life due to injury. The optimal tax weighs these costs against environmental benefits and redistributive concerns. 

Asymmetric information between home buyers and sellers may reduce owners’ incentives to invest in energy-conserving upgrades: when buyers cannot perfectly differentiate high- and low-efficiency buildings, sellers are unable to fully pass on the up-front cost of energy-related upgrades, and so face weaker incentives to make these upgrades. Local governments in many large U.S. cities have correspondingly passed legislation requiring public disclosure of energy use for large buildings. Do these laws improve energy efficiency, and if so, why? This paper first simulates a model of building owners’ dynamic decision about when and how much to invest in upgrading their building’s energy efficiency to show that steady-state building energy use is sensitive to information provision only when the extent of information provision about building efficiency affects the capitalization of energy efficiency in unit prices. I therefore investigate whether energy disclosure laws in New York City have affected unit prices, sale frequencies, and efficiency-related renovations in regulated multifamily buildings, using a regression discontinuity design. The effect of the policy on unit prices and sales is null and imprecise, which makes it difficult to conclude that the policy provided new information to buyers, but disclosure caused a 60% increase in boiler-related renovations in between 2016 and 2020.

Conservation easements protect open space by letting landowners cede development rights in perpetuity, either as a donation against tax liability or as a sale to a public or private buyer. The opt-in mechanism introduces inefficiencies. Landowners may protect land to prevent socially valuable development and over-ease. On the other hand, open space is a public good and thus faces private underprovision. Further, landowners may select land based on characteristics different from those a planner would use. We provide evidence of these inefficiencies. An increase in nearby development is correlated with more easing, suggesting a distaste for density drives easing. Using a close elections design on conservation bond levies, we find publicly funded acquisitions are concentrated on expensive land, but the marginally funded acquisition is cheap. The marginally procured easement is on private land, closed to the public, and has no biodiversity protection mandate, suggesting over-easing could dominate on the margin. States which pass legislation permitting ADUs and lot splits see a rise in easing, confirming these are substitutes.

Measuring pollution is central to enforcing many environmental regulations, but accurate tests are costly. Are inexpensive screening tests an effective replacement? We study this question in the context of vehicle emissions testing, where some regulators measure the emissions of passing cars with roadside remote sensing technology and use these readings to exempt low-emissions vehicles from costly in-person testing. Unlike traditional applications of screening tests in public health, where tested individuals want to learn their true disease status, screening in a regulatory context introduces an adverse selection problem because higher-polluting individuals may use the noisy screening measurements to conceal their true state. We find that remote sensing emissions tests are affected by adverse selection in Virginia: drivers of older, dirtier cars are more likely to seek out and drive by remote sensing devices on days when weather conditions are favorable for testing. In ongoing work, we use an instrumental variables strategy to estimate the effect of a marginally more generous screening threshold on air pollution and use un-selected data from Texas, where remote sensing is only used for program evaluation and not for screening, to explore how the policymaker can improve test performance by accounting for selection.