UC Berkeley

Economists have long prescribed taxing externality-generating goods according to their marginal damages to internalize the social costs of economic activity. Real-world policies, however, often differ significantly from this prescription. In some settings, practical constraints render perfect Pigouvian taxation infeasible. In other settings, political systems favor imperfect externality regulation, even when first-best policies are possible. In this dissertation, I study externality regulation in the context of environmental pollution and traffic congestion, with an eye on designing optimal policies in the face of political and practical constraints on policymakers.

In the first chapter, I study cordon zones, regions in the center of cities where cars are charged a toll for entering during peak hours. These policies are blunt approximations for Pigouvian taxes: uniform prices cannot reflect all of the heterogeneity in congestion and pollution damages associated with urban driving, and the discrete spatial and temporal cutoffs commonly used in cordon zones invite externality leakage. To understand the implications of these imperfections for setting optimal cordon prices, I first extend existing models from public economics to characterize optimal cordon tolls. This exercise yields a set of parameters necessary for calculating optimal road prices. I then recover estimates of these parameters using a natural experiment where bridge tolls were adjusted in the San Francisco Bay Area.

Using these parameters, I estimate optimal cordon tolls for three US cities: New York, San Francisco, and Los Angeles, each of which is currently planning on implementing cordon pricing. Across these three cities, I find that because drivers are willing to substitute to unpriced alternative routes or travel times in response to tolls, optimal peak-hour cordon prices are below the average social damages associated with trips passing through the cordon zone. Counterfactual simulations suggest that while cordon pricing would generate substantial welfare gains in each of the above cities, the imperfections in these policies mean that a majority of the theoretically possible welfare gains are left unrealized even under optimal peak-hour cordon pricing. To conclude this chapter, I discuss the prospects of improving the performance of cordon zones by allowing for granular time-of-day tolls, or by redesigning the cordon zones themselves.

In the second chapter, I investigate why voters support inefficient (standard-based) environmental policies over efficient (price-based) environmental policies. Using an information provision experiment conducted before and after voting on an environmental ballot initiative in Nevada, I estimate a model that maps voters' beliefs about the attributes of externality- regulating policies (effectiveness, cost, regressivity) to their support for these policies. I find that voting behavior is relatively unresponsive to perceived cost and perceived regressivity, but responsive to perceived policy effectiveness. Using this model, I decompose differences in support for a performance-based policy (Nevada’s Renewable Portfolio Standard) and a hypothetical price-based policy (a carbon tax). Oaxaca-Blinder decompositions imply that differences in perceptions of policy attributes explain just 23% of the gap in support be- tween renewable portfolio standards and carbon taxes, suggesting a significant role for “tax aversion.” To the extent that misperceptions of policy attributes do explain differences in support for these two policies, the explained gap results from overly optimistic beliefs about the attributes of renewable portfolio standards. To conclude, I predict voting behavior under several counterfactual scenarios. I find that in this setting, targeting revenue toward “swing” voters is unlikely to significantly improve support for carbon taxes. Instead, the results of this pilot experiment highlight the importance of communicating to voters the efficacy of price-based policies.

In the third chapter, I leverage a natural experiment to study whether Transportation Network Companies (TNCs) like Uber and Lyft worsen traffic congestion, and discuss the policy implications for cities considering regulating or taxing these companies. Applying difference in differences and regression discontinuity specifications to high-frequency traffic data, I estimate that Uber and Lyft together decreased daytime traffic speeds in Austin by roughly 2.3%. Using Austin-specific measures of the value of travel time, I translate these slowdowns to estimates of citywide congestion costs that range from $33 to $52 million annually. Back of the envelope calculations imply that these costs are similar in magnitude to the consumer surplus provided by TNCs in Austin, meaning that the entrance of these companies can be thought of as a transfer of welfare from incumbent road users to TNC customers.

Because it is less costly for cities to track the movements of ridesharing vehicles than it is to build the infrastructure for cordon pricing, TNC taxes are often proposed as an alternative to conventional congestion pricing. Anecdotally, taxing Uber and Lyft also enjoys a political advantage over cordon-based pricing (for example, New York, Chicago, and San Francisco all passed the TNC taxes, but are each engaged in political debates over cordon pricing). The results in this paper suggest that even if TNC taxes or restrictions are politically expedient, they are unlikely to generate large net welfare gains through reduced congestion.