Thanks, Tiebout. You’ve fixed the EU.

Opponents of the European Union’s jurisdictional expansion cite numerous reasons. “Brussels is too opaque,” some say. “Fiscally responsible countries will have to further subsidize irresponsible ones.” Yet, few posit the Tieboutian argument: at status quo, the EU enables the European continent to approach an optimal allocation of public goods. Consider the scholar’s assumptions. The Schengen Area removes international mobility barriers for individuals and businesses, with regional subsidies further lowering relocation costs. The internet, artificial intelligence, and the EU’s language-translation requirements foster the proliferation of information about government services, and the free-movement area includes 29 member states from which to choose. The number of Europeans working remotely has doubled since the pandemic, reducing career-related constraints. Centuries of national division and conflict has limited externality spillover of some public goods. And each nation’s immigration policy suggests there is an optimal population size to which each government aspires. Per Tiebout’s model, a European—his “consumer-voter”—can select the EU member state with the government services which most closely align with their preferences. A health-conscious Hungarian might seek out Finland's air quality, or family-oriented Romanian might enjoy Austria's high law-enforcement efficacy.

Of course, in reality, cultural barriers and job constraints prevent perfect application of Tiebout’s model, which was also designed for local public good allocation. However, never before has the European continent been closer to the model’s assumptions, and largely due to the EU’s integration efforts. At a time in which European public sentiment oscillates between the EU’s dissolution or federalization, perhaps the Tieboutian model can serve as a guiding force. The EU could be an institution which solely seeks to maximize the model’s assumptions, while leaving decisions regarding government resource allocation to its member states. Some public goods, such as defense, may be ceded to the EU, especially to prevent the free-rider problem NATO faces. Yet, Tiebout offers a middle ground which ensures economic efficiency—the goal of programs such as the Schengen Area—while maximizing national sovereignty.

AI Risk: A Coasian Solution?

The recent attack on Hugging Face by rogue OpenAI agent swarms has encouraged discussion of AI risk. "Doomers," or those who favor a shutdown of AI development, fear that continued development poses an existential risk that human civilization will be destroyed as AI agents spiral out of control. Let us assume that these doomers do not use AI, as they believe their usage would create incentives for labs to continue development. If we assume that development does in fact increase existential risk, developers (the producers) and users (the consumers) are imposing an uncompensated cost on the doomers, creating a negative externality. 

Thanks to Coase, we know that there exists a potential solution if transaction costs are sufficiently low and property rights are defined. An individual doomer with sufficient funds or a collective can bargain with the leaders of AI development, and if they can agree on a price to shutdown, the externality can be internalized. However, this solution seems relatively implausible for several reasons listed below. 

• The value that labs demand to shut down may be prohibitively large for a non-government entity. Consider Google's market cap of roughly $4 trillion, SpaceXAI's present valuation of $2 trillion, and Anthropic's and OpenAI's suggested IPOs at roughly the same value. And the heads of these labs believe strongly in the value of continued development. Dario Amodei suggests a future with 10-20% annual GDP growth! 
• There is a potential holdout problem. Even if Anthropic and OpenAI agree to terms with the doomers, Google, SpaceXAI, or Chinese developers like Moonshot may hold out for a higher dollar value or on principle, making bargaining difficult. 
• There is a free-rider problem. Each doomer fearing catastrophic risk benefits from the reduction in risk regardless of their contribution to the proposed payment. 

Is Voting Rational

Voting exhibits many traits of a public good. It is non-rival; my vote does not impede your ability to enjoy the right of voting, and I cannot exclude you from voting if you meet the appropriate qualifications. Therefore, since voting is a public good, it ought to display the trends of other public goods. Namely, that as the population increases, so does the free rider problem. As the number of people in the electorate increases, the power of any individual vote decreases. As a result, there would be less incentive in a large election to bear the cost of voting, as the likelihood of your vote deciding the election of a particular candidate or policy drops. 

But the numbers seem to tell a different story. Take the state of Virginia. In the 2024 presidential election, 72.9% of eligible voters in Virginia turned out to vote. In the following year, only 54.9% of eligible voters voted in the gubernatorial election. This does not make sense from a pure public goods framework. In the presidential election, the population of voters is vastly bigger, and so it would follow that more people would not vote and act as a free rider. However, 18% more Virginians voted in the presidential election than in the gubernatorial election the following year. Clearly there are other factors at play, but why do these other factors cause people to deviate from what the public goods framework would suggest so drastically?