Environmental externalities—uncompensated damages imposed on others—lie at the root of climate change, pollution, deforestation and biodiversity loss. Empirical evidence is limited, however, as to how externalities drive private decision making. We study one such behaviour, illegal tropical forest fires, using 15 years of daily satellite data covering over 107,000 fires across Indonesia. Weather-induced variation in fire spread risk and variation in who owns surrounding land allow us to identify how far externalities influence the decision to use fire. Relative to when all spread risks are internalized, we find that firms overuse fire when surrounded by unleased government lands where property rights are weak. In contrast, and consistent with the Coase Theorem, firms treat risks to nearby private concessions similarly to risks to their own land. Government sanctions, concentrated on fires spreading to populated areas, also deter fires, consistent with Pigouvian deterrence.

Many social insurance programs have low take-up, but it is unclear whether this is due to administrative barriers, information, or low insurance valuations. We study a Thai policy that offered large incentives for informal workers in selected provinces to enroll. The incentives increased insurance coverage by 67 percentage points- from 6 percent of informal workers to 73 percent- within two months. However, 12 months later, only 13 percent remained insured. Using choices among insurance tiers to back out revealed valuations, we find that low social insurance enrollment may be due to low ex-ante valuations of insurance, rather than administrative barriers.

Designing public transport networks involves trade-offs between coverage, service frequency, and direct service. We use the expansion of the bus system in Jakarta, Indonesia, to study these trade-offs. We analyze how new direct connections, changes in bus travel time, and wait-time reductions affect bus ridership and aggregate flows and estimate a transit network demand model by matching the route launch events. Commuters in Jakarta are 2–3 times more sensitive to wait time than bus time and inattentive to long routes. We develop a flexible framework to characterize optimal networks. A less concentrated network would increase ridership and commuter welfare.