In 1954, the economist Paul Samuelson — already on his way to becoming the most influential American economist of the twentieth century — published a three-page paper, The Pure Theory of Public Expenditure, defining a public good by two crisp properties: non-rival (one person's consumption does not reduce another's) and non-excludable (you cannot keep non-payers from benefiting). National defence is the textbook example, but lighthouses, basic scientific research, clean air, herd immunity, the legal system, and freely-broadcast information all have the public-good structure to varying degrees. Samuelson's mathematical contribution was to show that efficient provision required summing — not averaging — the marginal benefits across all consumers, a profoundly different aggregation than the one private markets perform. The result is the foundational market-failure argument for public provision: leave it to the market and everyone free-rides, hoping someone else pays, so the good is systematically underprovided.
Free-riding is what gives public-goods analysis its political teeth: if the good is non-excludable, no individual has any incentive to pay for it, and the unconstrained-market result is zero or near-zero voluntary provision. The cleanest experimental demonstration is the public-goods game in laboratory economics, where subjects given tokens to contribute privately to a multiplied common pot have a Nash equilibrium of contribute-zero, and observed contributions in repeated rounds decline toward it. Real-world manifestations are everywhere — PBS pledge drives, open-source software (the xz utils near-disaster of 2024 was the systemic version), vaccination where high-coverage communities tempt free-riders, climate cooperation where every country gains from emissions cuts but individually prefers to free-ride on others'. The standard solutions each carry characteristic costs: public provision financed by taxation underwrites defence, basic research, and infrastructure but introduces tax distortions and government-failure risk; excludability conversion (cable TV, toll roads, paid software) creates deadweight loss by excluding people who would have benefited at zero marginal cost; voluntary provision via norms sustains academic and open-source communities but scales poorly; Lindahl pricing — each person pays in proportion to valuation — is theoretically efficient but defeated by preference revelation, since no one will admit how much they value the good. The textbook's binary non-rival / non-excludable distinction is really a continuum, with club goods (excludable but non-rival — Netflix, country clubs) and common-pool resources (rival but non-excludable, like fisheries and the atmosphere) as the intermediate cases, and Elinor Ostrom's 2009 Nobel work showed that communities can self-govern many common-pool resources without state coercion. The framework's enduring usefulness is as a diagnostic: when a good looks non-rival and non-excludable, suspect underprovision and ask which solution's failure mode is the lesser evil.
Climate cooperation is the central public-goods problem of the present century: atmospheric stability is non-rival and non-excludable on a global scale, and the Westphalian state system has no enforcement mechanism that can compel contributions, so the Paris Agreement relies entirely on voluntary national commitments — a public-goods regime by aspiration. Open-source software (Heartbleed 2014, Log4Shell 2021, xz utils 2024) has revealed that the digital infrastructure rests on volunteer maintainers funded by no one, a public-goods underprovision visible only when it breaks. Basic scientific research has the strongest economic case for public funding (every empirical study finds R&D spillovers far in excess of private returns) but the framework is routinely threatened by political budget cycles. Wikipedia — a donation-funded global encyclopedia that by Samuelson's logic should not exist — remains the standing demonstration that public-goods provision can sometimes be sustained outside the state via norms, identity, and the strange durability of a small motivated contributor base.