The lighthouse and the public good

For a century, when economists wanted to explain a good that markets would fail to supply, they reached for the lighthouse. It is almost too perfect an example. A light thrown across a dangerous coast warns every vessel within sight of it, and the keeper has no way to warn only the ships that paid. The beam does not dim for the freeloader. It cannot be metered, invoiced, or withheld; one more ship taking comfort from it costs the light nothing and subtracts nothing from the comfort of the others. John Stuart Mill named it, Henry Sidgwick returned to it, and Paul Samuelson made it the textbook illustration of the concept economists now call a public good. The lighthouse became shorthand for a whole class of things that everyone benefits from, no one can be excluded from, and — precisely for those reasons — no ordinary market will trouble to build.

The two properties are worth naming plainly, because the argument turns on them. A public good is non-excludable: once it exists, you cannot practically stop anyone from enjoying it. And it is non-rival: one person's use does not diminish what is left for the next. Ordinary private goods are the opposite on both counts — a loaf of bread can be withheld from those who do not pay, and eating it leaves less for everyone else. The trouble with the lighthouse is that these two virtues, from the buyer's side, become a defect from the builder's. If I cannot charge the ships that benefit, and my charging one does not use up the benefit available to another, then my private calculation never justifies the cost of the tower. Each shipowner, reasoning the same way, waits for someone else to build it. And so, in the pure version of the story, the light that would save them all is never lit.

This is not an exotic case. It is the shape of a great deal of what makes exchange possible in the first place: the shared standard, the common defense, the clean harbor, the body of law that everyone relies on and no one can be billed for by the drink. These are the goods that sit underneath the market rather than inside it, and the market that depends on them is chronically bad at producing them, because the benefit spills over every fence you could build to charge for it.

The beam does not dim for the freeloader — which is exactly why the freeloader's logic, followed by everyone, leaves the coast dark.

The good you cannot sell

What makes the lighthouse illuminating, so to speak, is that it isolates a failure that has nothing to do with malice or incompetence. No one has behaved badly. Each shipowner has reasoned correctly about his own position: the light, if it exists, will warn me whether or not I paid, so I would be a fool to pay when I might free-ride on the payment of others. The defect is structural. It lives in the good itself, in the impossibility of tying the benefit to a bill. Multiply that reasoning across every party who would gain, and you get the signature result — a thing that would make everyone better off, that no one will build, not because they misjudge its value but because they judge their own share of it exactly right.

The lesson usually drawn is that such goods require the state: only a body that can tax has the reach to charge the beneficiaries who would otherwise slip the net. That is one answer, and often the right one. But it is not the only answer, and the real history of the lighthouse is more instructive than the tidy parable, because it shows that the underprovision is a problem of collection, not of nature — and problems of collection can be solved by institutions cleverer than a simple market and more various than a single state.

How the light actually got built

In 1974 Ronald Coase published a short, deflating paper called The Lighthouse in Economics, and pointed out that the textbook writers had never actually looked. The English lighthouse system, it turned out, was not a monument to the impossibility of private provision. For long stretches it was substantially private. The Crown granted patents to individuals to build lights and, crucially, to collect light dues — tolls levied on ships according to tonnage as they passed or called at nearby ports. Trinity House, the chartered corporation that came to superintend English lighthouses, itself operated within this system of dues, sometimes building lights directly, sometimes licensing and later buying out private ones. The benefit that seemed impossible to charge for was charged for after all, not at the beam but at the harbor, where the ship had to come to shore and could be billed for the passage it had made.

Coase's point was not that lighthouses are ordinary market goods; the toll was collected under a franchise backed by law, hardly the free market of the parable. His point was subtler. The good that theory declared unfundable was funded, for centuries, by an institutional arrangement that found a chokepoint — the port — where the diffuse benefit could be reattached to an identifiable beneficiary and a bill presented. The lesson, read honestly, is not that shared safety infrastructure builds itself, and not that only the state can build it. It is that such infrastructure gets built when someone designs the institution that funds it: a franchise, a levy, a charter, a rule that says those who benefit shall help pay, and a mechanism that makes the saying stick. The commons does not appear by miracle or by market. It is engineered, deliberately, by people who refuse to let the collection problem have the last word.

Trust is a lighthouse

Now consider the thing a world of machine-made decisions most needs and is least equipped to buy on its own. Not any one company's model, and not any one firm's assurances about it, but the shared infrastructure that makes machine decisions trustable at all: common standards for what a decision record must contain to count as an account; open tools for verifying that a record is what it claims to be; the public capacity — in regulators, in courts, in journalists, in the affected parties themselves — to actually audit a decision after the fact. This infrastructure has the exact shape of the lighthouse. It is non-excludable: once a credible standard for a Decision Receipt exists, once the means to check provenance and replay a decision are in the open, every participant in the decision economy benefits from the trust they create, including those who paid nothing to build them. And it is non-rival: my verifying a decision against the shared standard does not use up your ability to verify yours. The benefit spills over every fence.

Which means it will be underprovided for precisely the lighthouse's reason. Each firm, reasoning correctly about its own position, would rather others bear the cost of building the standard, the verifier, the audit capacity — and then enjoy the general climate of trust those things produce. Left to private calculation, everyone waits, and the coast stays dark: a decision economy full of confident machine outputs and no shared means of holding any of them to account. The credibility that would make the whole system navigable is exactly the good no single participant is motivated to build alone.

The lighthouse tells us this is not a counsel of despair; it is a design brief. The light-dues insight is the operative one: the diffuse benefit of trustworthy decisions can be reattached to the parties who gain from it — the institutions that deploy machine decisions and want them believed, the markets that want to trade on them, the public authorities that would rather audit a standard record than reconstruct a lost one. Accountability infrastructure is a commons, and like every commons it needs an institution built on purpose to exist at all — a standard someone convenes, a verification anyone can run, an audit right the affected can actually exercise, and some arrangement by which those who draw comfort from the light help keep it lit. The alternative is not a market that quietly supplies the good anyway. The alternative is the dark coast, and the wrecks we will insist, afterward, that no one could have foreseen.

— Dispatches · Summit Cognitive

---