Environmental regulation requires difficult targeting decisions. Regulators cannot inspect every factory at every moment, and therefore must decide which firms to monitor, inspect, and penalize with limited enforcement capacity. In practice, these decisions are often based on a factory’s measured pollution levels, past violations, industry, or perceived compliance risk. These are important inputs, but they do not necessarily capture the social harm caused by a given factory’s emissions. The same amount of pollution can generate very different damages depending on the factory’s location, prevailing wind patterns, atmospheric transport, nearby population density, and the vulnerability of exposed populations. This project explores a unique opportunity to study whether these inefficiencies can be reduced through algorithmic regulation.
