The Myth of Underprovision

Public goods and institutional choice

Standard public-goods theory identifies a coordination problem, not an automatic mandate for the state. A public good is non-rival in consumption and difficult to exclude non-payers from using. Those features can weaken private funding incentives. Whether provision is too low depends on a criterion; whether taxation improves it depends on institutional knowledge, administrative cost, coercion, capture, and available alternatives.

The canonical lighthouse case illustrates why institutional detail matters. British lights were sometimes built or operated by private parties, but light dues were collected through legally backed port systems and monopoly grants. That is neither pure voluntary provision nor simple tax-funded bureaucracy. It is a mixed arrangement that bundled a hard-to-exclude service with an excludable transaction. The example defeats categorical impossibility, not the existence of a free-rider problem.

That inversion is not an isolated embarrassment; it is the shape of the entire argument. Of the standard indictments of markets I answer in Capitalism on Trial, the public-goods charge has the most elaborate technical case and therefore warrants the fullest reply: the underprovision claim is not a discovery about the world. It is a model output whose assumptions contain its conclusion, and both history and technology testify against it.

A Model, Not a Finding

The underprovision result is model-dependent, and empirical work must estimate its magnitude in particular settings. A simple model commonly relies on conditions such as:

1. Exclusion is costly relative to the good.
2. Beneficiaries can free-ride on contributors.
3. Decentralized contributors cannot capture enough of the benefit to fund the chosen benchmark.

Technology, bundling, repeated interaction, clubs, philanthropy, prizes, and assurance contracts can relax these conditions. Scale, anonymity, diffuse benefits, high transaction costs, and strategic delay can strengthen them. A model that identifies the problem still owes a comparison of remedies. A planner need not know a perfect optimum to improve a severe failure, but public provision can also overshoot, misallocate, crowd out alternatives, or become captured.

In other words, the inevitability of the state is baked in from the start. The model does not discover that coercion is necessary; it assumes a world in which nothing else could possibly work, and then announces the necessity as a theorem. That is not economics finding a market failure. That is statism doing arithmetic on its own premises.

History Doesn’t Agree

History supplies counterexamples to universal impossibility and examples of persistent coordination failure. The cases should identify which mechanism did the work.

Exclusion evolves. The British lighthouses are the cleanest case: you cannot bill a ship at sea, but every ship eventually docks, and light dues collected at port funded the beams for generations. Roads and bridges tell the same story — historically toll-based, privately built and maintained turnpikes, exclusion accomplished by nothing more exotic than a gate.

Voluntary cooperation scales. Education flourished through churches, guilds, and subscription schools long before governments monopolized it. Law and order — the good statists reach for when all others fail — emerged in frontier zones and in medieval Iceland from voluntary associations and competing legal systems; the Icelandic commonwealth ran on chieftains whose adherents could leave for a rival, and it kept the peace for centuries without anything a modern textbook would recognize as an executive state. Scientific discovery, the paradigm non-rival good, was propelled by prizes, patrons, and voluntary societies — knowledge bought into existence by people who wanted it found, not extracted from people who never asked.

And nobody planned the optimum. No welfare function dictated how many lighthouses Britain needed or how much science the prize societies should fund. The provision that occurred was whatever the people who valued these goods were willing to pay for — which is not a defect of the record; it is the record’s lesson.

Far from being anomalies, these cases show people consistently solving so-called public goods problems without the state. Where exclusion was feasible, it was invented. Where it was not, cooperation found other channels. The three assumptions do not describe constraints on human societies; they describe the modeler’s refusal to look at them.

The Sleight of Hand

Underprovided relative to what? Here is the pivot on which the whole doctrine turns. When economists say markets “underprovide,” they mean: less than the quantity a planner’s social welfare function demands. That is not an objective analysis. It is a political judgment smuggled in under the banner of science. It presumes that more of the good is always better, and — the deeper smuggle — that coercion has no real cost.

But coercion is never costless. Taxation requires threats. It warps incentives, crowds out the voluntary alternatives that would otherwise evolve, and entrenches monopolies in exactly the goods it claims to rescue — the nationalized lighthouse, the compulsory school. The tidy diagram in the textbook, with its shaded triangle of recovered welfare, leaves out the destruction of agency: every unit of “corrected” provision is financed by payments that were not choices. And a payment extracted under threat reveals nothing about what the payer values — that is the value-theory half of the free-rider argument, and I have made it in full in Value as Sacrifice: benefit is not valuation, and taxed “demand” is the state’s preference wearing the taxpayer’s name. What remains for this chapter is the economic half.

What Markets Actually Show

Voluntary payment reveals what people with resources and a capture mechanism will fund under the available rules. Nonpayment is ambiguous when contributors expect to benefit regardless, cannot coordinate, lack purchasing power, distrust the institution, or reasonably wait for others. The funding test remains powerful for ordinary private goods; it is precisely what the public-goods conditions complicate.

Meanwhile the supposedly fixed wall of non-excludability keeps eroding, because non-excludability was never a metaphysical property of goods — it is a snapshot of current technology and institutions, and the snapshot keeps changing. Micropayments make it economical to charge fractions of a cent for marginal use. Assurance contracts let contributors pledge conditionally, binding only when enough others join, dissolving the free-rider’s advantage by design. Cryptographic exclusion — encrypted signals, access keys, blockchains — makes it trivial to gate goods that were open to all comers a generation ago. The first assumption of the model is not merely false about the past; it is being falsified again every year.

So what looks like underprovision is usually provision in unfamiliar forms: messy, decentralized, diverse, voluntary. The complaint was never about scarcity. It is about a refusal to conform to the state’s preferred template — one uniform good, one funding stream, one administrator.

Owing at the Margin

The marginalist revolution that McCloskey rightly celebrates¹ already contains the refutation of totalist accounting. You can only ever owe at the margin, never for the whole past — no one presents you a bill for the entirety of civilization and demands your gratitude in tax. Likewise, markets only ever provide at the margin, according to voluntary demand, one willing payment at a time. Labeling marginal, voluntary provision “insufficient” against a totted-up social aggregate is not economics. It is ideology — the pre-marginalist habit of pricing wholes, revived exactly where it flatters the state.

Underprovision must not become a blank check for coercion. Nor should the existence of voluntary provision become proof that no important benefit was left unrealized. The disciplined sequence is: define the good and benchmark, identify the free-rider mechanism, inventory voluntary and bundled alternatives, estimate the remaining gap, then compare public and private remedies under the same assumptions about knowledge, capture, enforcement, and error correction.