Alliance free-riding

Why does every member of an alliance underspend on a defence all of them want?

Every few years the same argument returns: the big ally complains that the small ones are not paying their share, and the small ones reply that they are doing what they can. Both sides treat it as a quarrel about character — generosity against stinginess.

But notice something odd before we take a side. If we could somehow read the private preferences of all thirty-odd governments, we would probably find that each of them wants more collective deterrence than the alliance currently buys. Not one of them is getting what it wants, and yet none of them buys more. What kind of arrangement produces an outcome that nobody in it prefers?

Let us start with the thing being bought. What exactly does a dollar of your ally's defence spending purchase for you?

If it purchases deterrence — an adversary's belief that attacking the alliance would not pay — then ask two questions about that product. Can it be sold to one member and withheld from another? And does your consumption of it use it up? The answer to both is no. A deterred adversary is deterred against everyone under the treaty, and your being protected does not leave less protection for the next member. Those two properties — non-excludable and non-rival — are what economists mean by a public good, and they are the whole engine of this problem.

Now put yourself inside one member's budget meeting. You are deciding how much to add. What do you weigh? Your own cost, in full — you pay every cent. And your own benefit from the increment — a small share of the total safety it creates, because the rest of the benefit lands on twenty-nine other capitals that pay you nothing for it. So your calculation stops early. You buy up to the point where the marginal benefit to you equals the marginal cost to you, and the benefit you confer on everyone else never enters the sum at all.

Do you see that this is not cheating? Each member is being perfectly rational and perfectly honest about its own accounts. The shortfall is structural: the good is jointly consumed and separately paid for, so every member ignores most of the value its own spending creates. Add up thirty such decisions and you get less than the alliance, considered as one body, would want to buy. That is the underprovision, and it applies to the largest member too.

Then why does the largest member still end up carrying the heaviest load? Follow the same logic one step further. The bigger a member is — the more it has at stake in the alliance's survival — the larger the slice of the total benefit that lands on it, so the further along the curve it keeps buying before its own marginal benefit falls below its own marginal cost. And once it has bought that much, what is the small member's best remaining move? Very little more is needed for the small member's own purposes, because the giant has already supplied most of what it wanted. Mancur Olson and Richard Zeckhauser set this out in 1966 and named the result the exploitation of the great by the small — the disproportion falls out of the arithmetic, with no bad faith required anywhere.

One honest complication, because the theory has been tested and does not fully survive. Defence spending is not purely a public good. A country's forces also buy things only that country consumes: control of its own territory, the ability to act alone, industrial jobs, political standing. Economists call this the joint product view, and where those private returns are large the free-riding prediction weakens sharply — which is why frontline states facing a direct threat often spend heavily despite being small. The pure public-good model is a floor of understanding, not the whole account.

Three refinements are worth pinning down.

First, "free-riding" is a description of an incentive, not an accusation. The word invites us to look for a shirker, when the mechanism predicts underprovision even in an alliance of saints who all sincerely want more defence.

Second, spending targets such as NATO's two percent of GDP guideline — agreed at the Wales summit in 2014 — are attempts to convert a public good back into something with a private price: a visible commitment, publicly scored, so that failing to pay costs a member something it cannot share with anyone. Whether that works depends on whether the reputational cost is real.

Third, the cure is not simply "make the small ones pay". Because everyone underspends relative to the joint optimum, redistributing the same total differently does not fix the shortfall — it only moves it.

• The joint product model, and why forces that yield private national benefits weaken the free-riding prediction.
• Whether public spending targets change behaviour or only reporting.
• The same structure elsewhere: climate agreements, vaccination, and shared fisheries.

Every term the collection defines is gathered in the glossary.