Tyler Cowen came to UATX to give a talk. It will be on line at some point, but the UATX media production team is quite overwhelmed these days.
Tyler has a vision for higher ed that is heavily influenced by AI. Perhaps he came hoping that as a new university UATX would be receptive to his vision. Or perhaps he knew that it would not.
The thing is, UATX was launched not to bury the traditional higher ed model but to save it. The new university’s connections with the Austin tech community influence it in the direction of innovation. But its main approach to fixing higher education owes more to the old-time religion of dead white males in the curriculum, meritocratic admissions, and political incorrectness.
I remain AI-pilled. I wanted to schedule one-on-one meetings with students, and they suggested using a calendar app. I bet myself that I could develop an app faster than I could install, learn, and adapt someone else’s app. Using Claude Opus 4.5, I had the meeting scheduler app vibe-coded and running with less than half an hour of my time put into it.
We also needed to have more clarity concerning the schedule for when material will be covered and which students will be responsible for leading it (I am borrowing from the Swarthmore Honors Program the system of having each class period discussion guided by a student paper.) So I developed an app for that, also.
That planning app took longer, because I wanted to use a free-form, natural-language interface. Instead of the user seeing a menu, I wanted the user to be able to query the data however they choose. “When is my next paper due?” or “What’s the schedule in Public Choice for the next three weeks?”
I think that the user interface for most software needs to move in that direction. I cannot stand bloated “courseware” in general, because pulling data out of such a system ends up being a game of hide-and-seek. The courseware that UATX uses is typical, in that someone can upload a document to share without another user knowing where to find it.
I wanted to demonstrate that an AI interface could expose the data to the user without the user having to play guessing games with a menu. You just say what you want and the AI finds the information in the data tables. I think that if you gave me the data tables for our courseware, if I wanted to I could have a free-form, natural language interface up and running in less than a week. Die, courseware, die.
For my student paper administrative system, Claude knew what I wanted and quickly developed the code. The only challenge with it was configuration. I was impatient to get it up and running, and I didn’t want to go through the process of using Github. That probably was a mistake. But after some painful trial-and-error Claude and I did get the app working. [Both of my new apps need to stay private, because I did not have time to work on issues related to security and confidentiality.]
A valid criticism of my courses so far is that the structure is far from transparent. For the Public Choice course, I tried to address that with a single prompt to Claude. It took my minimal syllabus outline and came up with what I will paste in below.
As a student, you work with a mentor. At the beginning of each term, you and your mentor decide which courses you will take. If there are other students on campus taking them, great. If not, maybe you can take them with students at other schools, meeting remotely.
For each course, an AI can design the syllabus. Tyler gave an example of a syllabus generated by ChatGPT for a course on Tudor England. If you can find a qualified teacher for that course, great. If not, you could try learning it from ChatGPT, which would provide lessons, conversations, and learning assessments (tests).
Tyler thinks that 1/3 of higher ed right now should consist of teaching students how to work with AI. I do that by assigning a vibe-coding project, and by encouraging “vibe reading” and “vibe writing.”
The reason for proposing such a high proportion of effort to learning to work with AI is because we are in a transition period, where the capabilities of AI are changing rapidly. Once capabilities settle down, best practices will become established, and knowledge of how to use AI will be ingrained. For now, it is very hard to keep up.
It is possible, of course, that Tyler and I could be wrong. It could be that the best approach for higher ed is to keep students as far from AI as one can. I can respect someone who favors an anti-AI approach.
But I am disturbed by the lack of humility that often accompanies the anti-AI position in higher education. I have difficulty comprehending how faculty, at UATX and elsewhere, can express their anti-AI views with such vehemence and overconfidence. They come across to me like dinosaurs muttering that the meteor is not going to matter to them.
Public Policy and Public Choice
A Detailed Course Outline
This four-week course introduces students to the economic analysis of government and political decision-making. Beginning with the standard textbook rationales for government intervention in markets, the course progressively builds a more sophisticated framework for understanding when and how governments succeed or fail in addressing social problems.
The course follows a deliberate intellectual arc: Week 1 establishes the conventional economic case for government action, Week 2 introduces alternative private and communal solutions that challenge the necessity of government intervention, Week 3 examines the fundamental information problems that constrain all economic planning, and Week 4 analyzes the political incentives that shape actual government behavior. By the end, students will have a nuanced understanding of both market failures and government failures, enabling them to think critically about policy proposals.
This week establishes the theoretical foundation that dominates most introductory economics courses. Students will learn the standard economic justifications for government intervention in otherwise free markets. These theories, developed primarily in the mid-twentieth century, provide a coherent framework explaining when competitive markets may fail to produce efficient outcomes and how government action might theoretically improve upon market results. Understanding these arguments thoroughly is essential before we can meaningfully critique them in subsequent weeks.
A good is non-rival when one person’s consumption does not diminish another person’s ability to consume the same good. National defense is the classic example: protecting one citizen from foreign invasion automatically protects all citizens within the same borders, regardless of how many people benefit. Knowledge and information often exhibit non-rivalry as well—my learning a mathematical theorem does not prevent you from learning it too. This characteristic creates a wedge between private incentives and social optimality: private providers may undersupply non-rival goods because they cannot capture the full social value of their production.
A good is non-excludable when it is impossible or prohibitively costly to prevent non-payers from consuming it. A lighthouse’s beam guides all ships within view, whether or not their owners have paid for the service. When exclusion is impossible, markets face a free-rider problem: rational individuals have no incentive to pay voluntarily since they will receive the benefits regardless. The combination of non-rivalry and non-excludability defines a ‘pure public good,’ which standard theory suggests will be undersupplied without government provision or subsidy.
An externality exists when a transaction between two parties imposes costs or benefits on third parties who did not choose to participate. Pollution is the paradigmatic negative externality: a factory’s emissions harm nearby residents who receive no compensation and had no say in the production decision. Education and vaccination represent positive externalities, where the benefits extend beyond the direct consumer to society at large. When externalities are present, market prices fail to reflect true social costs and benefits, leading to overproduction of goods with negative externalities and underproduction of goods with positive externalities.
Named after economist Arthur Cecil Pigou, a Pigovian tax is designed to correct negative externalities by imposing a tax equal to the marginal external cost at the socially optimal level of output. By forcing producers to internalize the external costs they impose on others, the tax theoretically aligns private incentives with social welfare. A carbon tax is perhaps the most discussed contemporary application: by charging emitters for each ton of carbon dioxide released, the tax makes polluters bear the climate costs they would otherwise impose on society. The elegance of the Pigovian approach is that it achieves efficiency without dictating specific behaviors—firms can choose to pay the tax, reduce emissions, or exit the market, whichever is least costly.
Richard Musgrave’s influential framework divides government’s economic role into three distinct functions: allocation, distribution, and stabilization. The allocation function addresses market failures such as public goods and externalities, seeking to direct resources toward their most efficient uses when markets fail to do so. The distribution function concerns the division of income and wealth across society, recognizing that efficient outcomes are not necessarily equitable ones. The stabilization function involves macroeconomic policy—using fiscal and monetary tools to maintain full employment, stable prices, and economic growth. This tripartite division remains the organizing framework for most public finance textbooks and provides the intellectual justification for government’s economic role in mixed economies.
Cowen and Tabarrok, Modern Principles of Economics, Chapter 19
Musgrave, The Theory of Public Finance (selected chapters)
Kling, ‘Naive Economics’ from Social Code (soccode.vercel.app)
How do we determine in practice whether a good is sufficiently non-excludable to justify public provision?
What assumptions are embedded in the Pigovian solution that might not hold in practice?
Can Musgrave’s three functions be kept analytically separate, or do they inevitably overlap and conflict?
Having established the standard case for government intervention, this week challenges the assumption that government is the only or best solution to market failures. We examine three alternative approaches: private bargaining when transaction costs are low, community-based institutions for managing common resources, and constitutional frameworks for collective decision-making. These alternatives suggest that the choice is not simply between unregulated markets and government intervention, but includes a rich variety of institutional arrangements that may outperform both extremes.
Week 1 treated government intervention as the natural remedy for market failure. This week complicates that picture by showing that private parties and communities often solve externality problems without government involvement—and sometimes more effectively. The key insight is that identifying a market failure does not automatically justify government action; we must compare imperfect markets to imperfect alternatives, including imperfect government solutions. This comparative institutional perspective will deepen throughout the course.
Ronald Coase’s revolutionary insight was that externalities do not necessarily require government correction. When transaction costs are sufficiently low, private parties can bargain their way to efficient outcomes regardless of how property rights are initially assigned. Consider a factory polluting a river used by a downstream fishery. If transaction costs are low, the factory and the fishery can negotiate: either the factory pays to reduce pollution, or the fishery pays the factory to reduce output—whichever is cheaper. The efficient outcome emerges from bargaining, not regulation. Crucially, Coase’s point was not that transaction costs are always low, but that the policy question should focus on which institutional arrangement minimizes transaction costs, rather than assuming government intervention is the default solution.
Common pool resources (CPRs) are goods that are rival in consumption but difficult to exclude users from—fisheries, irrigation systems, forests, and groundwater aquifers are classic examples. The conventional ‘tragedy of the commons’ narrative suggests such resources will inevitably be overexploited without either privatization or government regulation. Elinor Ostrom’s groundbreaking empirical research demonstrated that this dichotomy is false: communities around the world have developed sophisticated institutional arrangements to manage common resources sustainably, often for centuries. These arrangements typically involve locally devised rules, monitoring by community members, graduated sanctions for violations, and low-cost conflict resolution mechanisms. Ostrom’s work earned her the Nobel Prize and fundamentally challenged the assumption that only states or markets can solve collective action problems.
James Buchanan and Gordon Tullock’s ‘calculus of consent’ applies economic reasoning to the design of political institutions themselves. Rather than taking government as given, they ask: what decision rules would rational individuals choose behind a ‘veil of uncertainty’ about their future positions in society? Their analysis reveals a fundamental tradeoff between two types of costs. External costs are imposed on individuals by collective decisions they oppose—majority rule generates lower external costs than dictatorships. Decision costs are the time and resources required to reach collective agreement—unanimity has high decision costs, while majority rule economizes on them. Optimal decision rules balance these considerations. Buchanan and Tullock argue for constitutional unanimity (everyone must agree to the basic rules) combined with less restrictive rules for ordinary legislation, providing an economic foundation for constitutional democracy.
Coase, ‘The Problem of Social Cost,’ Journal of Law and Economics (1960)
Ostrom, Governing the Commons (selected chapters)
Buchanan and Tullock, The Calculus of Consent (selected chapters)
Under what conditions are Coasean bargaining solutions likely to succeed? What makes transaction costs high?
What features distinguish communities that successfully manage common pool resources from those that fail?
Does the calculus of consent suggest any reforms to existing democratic institutions?
This week examines a fundamental challenge facing any attempt at economic coordination, whether by markets, governments, or communities: the problem of dispersed and often tacit knowledge. Friedrich Hayek’s classic argument demonstrates that the knowledge required for rational economic planning is not concentrated in any single mind or institution but scattered across millions of individuals in forms that cannot be easily communicated or aggregated. This epistemological challenge poses severe constraints on what any central authority—whether a government regulator or a corporate planner—can hope to achieve.
Week 1’s Pigovian taxes and Week 2’s Coasean bargaining both require knowledge that may be difficult or impossible to obtain. How does a regulator know the marginal external cost of pollution at each facility? How do bargaining parties value harms they have never experienced? The information problem suggests that the textbook solutions of Week 1 face implementation challenges that go beyond political will—they may be epistemically impossible to execute correctly. Meanwhile, Ostrom’s community institutions from Week 2 gain new appeal: local knowledge may be precisely the tacit, contextual information that central authorities cannot access.
Michael Polanyi distinguished between explicit knowledge (which can be articulated, codified, and transmitted) and tacit knowledge (which cannot). Tacit knowledge is the ‘know-how’ that skilled practitioners possess but cannot fully explain—the experienced manager’s sense for when a project is going wrong, the entrepreneur’s feel for market opportunities, the farmer’s intimate understanding of local soil conditions. This knowledge resists formalization not because of laziness or secrecy, but because it is embedded in practice and context. Central planners cannot collect what cannot be articulated, which means that any system relying on centralized information will systematically exclude crucial knowledge that only decentralized actors possess.
Hayek’s insight was that prices in a market economy serve as a telecommunications system, conveying information about relative scarcity across the entire economy without any central processor. When a resource becomes scarcer—perhaps due to a natural disaster or new use—its price rises, instantly signaling all users to economize while signaling all producers to supply more. No single person needs to know why the scarcity occurred; the price signal alone provides sufficient information for rational response. This spontaneous coordination mechanism economizes on knowledge in a way that no central planning bureau could replicate. Regulators who override price signals—through price controls, subsidies, or mandates—suppress this information and must somehow replace it with their own inevitably incomplete understanding.
Building on Hayek’s framework, contemporary analysts have identified specific ways the information problem manifests in regulatory settings. Regulators face a calculation problem analogous to the socialist calculation problem identified by Ludwig von Mises: without market prices for regulatory benefits and costs, how can they know whether their interventions create more value than they destroy? Cost-benefit analysis attempts to answer this question, but it requires regulators to estimate values that only market transactions could reveal. The regulator’s calculation problem is not that regulators are stupid or corrupt (though they may be), but that they lack access to the dispersed, tacit, local knowledge that would be required to regulate efficiently.
Hayek, ‘The Use of Knowledge in Society,’ American Economic Review (1945)
Kling, ‘The Regulator’s Calculation Problem,’ EconLib
Kling, ‘Designing a Better Regulatory State,’ National Affairs
Can advances in information technology and artificial intelligence overcome the knowledge problem, or do they face fundamental limits?
How should policymakers respond to the information problem—with humility, with improved measurement, or with alternative institutional designs?
Are some policy domains more susceptible to the knowledge problem than others?
The previous weeks established that markets sometimes fail and that government intervention faces serious knowledge constraints. This final week completes the picture by examining the incentive structures that shape actual government behavior. Public choice theory applies the same self-interest assumptions economists use to analyze markets to the analysis of politics. Politicians, bureaucrats, interest groups, and voters all respond to incentives—and those incentives often diverge dramatically from the public interest. Understanding these incentive problems is essential for predicting which government interventions will improve outcomes and which will make things worse.
Week 1 presented government as a solution to market failure, implicitly assuming benevolent and competent policymakers. Week 2 showed that private and communal institutions sometimes outperform government. Week 3 revealed that even well-intentioned regulators face knowledge constraints. Week 4 removes the assumption of benevolence itself: what if policymakers pursue their own interests rather than the public interest? The course thus traces an intellectual arc from naive faith in government solutions through increasing skepticism, arriving at a mature public choice perspective that treats government failure as seriously as market failure.
Rent-seeking refers to the expenditure of resources to capture transfers rather than to create new value. When government has the power to grant monopolies, restrict entry, or award contracts, private parties will invest in lobbying, campaign contributions, and political influence to capture these government-created rents. These expenditures are socially wasteful: they produce nothing of value and simply redistribute existing wealth. Worse, the possibility of rent-seeking induces government to create more opportunities for transfer, establishing a destructive equilibrium where resources flow into political competition rather than productive activity. Gordon Tullock’s key insight was that the social cost of monopoly includes not just the deadweight loss from restricted output but all the resources squandered in the competition to obtain and defend monopoly privileges.
The ‘resource curse’ describes the paradoxical finding that countries rich in natural resources often experience slower economic growth and worse governance than resource-poor countries. The mechanism involves rent-seeking: when enormous wealth can be captured by controlling resource extraction, political competition becomes a zero-sum fight over resource rents rather than a positive-sum effort to grow the economy. Foreign aid can create similar dynamics—the ‘aid curse.’ When a government’s revenue comes from aid donors rather than domestic taxation, it becomes accountable to donors rather than citizens, and domestic political competition focuses on capturing aid flows rather than providing services. Both phenomena illustrate how wealth that arrives as a windfall, rather than being created through productive activity, can distort political incentives catastrophically.
Mancur Olson’s distinction between roving and stationary bandits provides a framework for understanding the incentives of predatory rulers. A roving bandit—a warlord who raids a territory and moves on—has no incentive to leave anything behind; plundering completely maximizes his return. A stationary bandit who expects to rule the same territory indefinitely has different incentives: excessive plundering today reduces what can be extracted tomorrow. The stationary bandit therefore has a self-interested reason to limit predation, provide basic public goods, and foster economic growth that enlarges future tax revenues. This framework explains why stable autocracies often outperform anarchic failed states and why the time horizons of rulers profoundly affect policy outcomes.
Mancur Olson’s ‘logic of collective action’ explains why concentrated interests systematically triumph over diffuse ones in political competition. The benefits of protective tariffs, occupational licensing, or agricultural subsidies are concentrated among small groups who receive large per-capita gains. The costs are spread across millions of consumers or taxpayers who each bear only trivial losses. Small groups can organize effectively: each member has strong incentives to contribute, and free-riders can be identified and sanctioned. Large groups face crippling collective action problems: no individual’s contribution matters much, and free-riding is easy. The result is systematic bias in democratic politics toward policies that benefit organized interests at the expense of the general public.
A revealing pattern emerges across policy domains: governments simultaneously subsidize demand while restricting supply. Higher education features generous student loans and grants (demand subsidies) alongside accreditation barriers that limit new entry (supply restrictions). Housing policy combines mortgage interest deductions and down-payment assistance (demand subsidies) with zoning laws that constrain construction (supply restrictions). Healthcare shows the same pattern with insurance subsidies alongside licensing and certificate-of-need laws. This combination is economically perverse—it inflates prices while doing little to expand access—but politically rational. Demand subsidies benefit visible recipients while supply restrictions protect incumbent producers. The costs fall on future consumers and never-born competitors who lack political voice.
Bruce Yandle’s ‘bootleggers and Baptists’ theory explains why regulations often persist despite serving neither the public interest nor pure rent-seeking. The name comes from alcohol prohibition: Baptists supported Sunday closing laws on moral grounds, while bootleggers supported them because banning legal competition increased their profits. Neither group alone could have prevailed, but together they formed an unspoken coalition. This pattern repeats across policy domains: environmental regulations may be supported both by genuine environmentalists and by incumbent firms who benefit from compliance costs that burden competitors. The theory explains the durability of inefficient regulations and the strange coalitions that defend them.
Olson, The Logic of Collective Action (selected chapters)
Kling, ‘Political Realism’ from Social Code (soccode.vercel.app)
Can democratic institutions be designed to overcome the logic of collective action, or is political bias toward concentrated interests inevitable?
How can citizens identify bootlegger-and-Baptist coalitions, and does this knowledge enable better policy evaluation?
Does the public choice perspective counsel despair about government, or does it point toward institutional reforms that could align political incentives with the public interest?
This course has traced an intellectual journey from the textbook case for government intervention through increasingly sophisticated challenges to that framework. The result is not a simple rejection of government but a mature comparative institutional analysis that takes both market failure and government failure seriously.
Week 1 established that markets can fail when goods are non-rival, non-excludable, or generate externalities. Pigovian taxes and Musgravian public finance provide theoretical solutions. Week 2 demonstrated that government is not the only alternative to unregulated markets—Coasean bargaining, Ostromite communities, and constitutional design offer competing approaches. Week 3 revealed that all attempts at economic coordination face fundamental knowledge constraints—the dispersed, tacit nature of economic knowledge limits what any central authority can achieve. Week 4 showed that even if regulators possessed sufficient knowledge, their incentives often diverge from the public interest—rent-seeking, collective action problems, and political distortions systematically bias policy outcomes.