Anarchy and the security dilemma

There is no world government. Every state is finally responsible for its own survival, and there is no authority that can make another state keep a promise. That is what people mean by anarchy in this context, and it does not mean chaos. Most of the time it means an extremely orderly world in which everybody is arming.

The trap

A state buys weapons because it wants to be safe. Its neighbor cannot tell a defensive purchase from a preparation, because they look identical and the difference lives in an intention that can change with the next election. So the neighbor buys weapons too. Both are now poorer and neither is safer, because safety here is relative and relative positions do not move when both sides move together.

Nobody in that paragraph is aggressive, nobody has misunderstood anything, and nobody has made a mistake. Below are the two states as arithmetic, with no intentions available to either of them.

Each curve says what one state should spend given what the other spent. They cross once, at a point where both have spent heavily and neither has gained anything, and there is nowhere else on the chart where both sides are content to stand still.

The term is John Herz's, from 1950, and the fullest statement is Robert Jervis, "Cooperation Under the Security Dilemma", World Politics, 1978, which is where the offense-defense argument below comes from. The contest function used here is standard in the conflict literature and the figures come from the model on this page, which is unit tested.

ConventionJervis's refinement is the part worth carrying. Two questions decide how bad the spiral gets. Can a defensive weapon be told apart from an offensive one? And does attacking or defending have the advantage? When defense is distinguishable and cheaper, states can buy safety without frightening anybody, and the dilemma nearly disappears. When it is not, every purchase is a threat. Walls, mines, and coastal defenses sit at one end. Long-range strike and first-strike weapons sit at the other.

What actually gets states out of it

• Verification. Not trust. Inspection regimes, satellite monitoring, and notification agreements change what each side can observe, and observability is what the spiral runs on. This is why the arms control that worked was technical rather than rhetorical.
• A big enough asymmetry. The best response to an overwhelming rival is to spend nothing, which the model gives directly. Hopeless positions are peaceful, and that is not a recommendation.
• Making the stake smaller. Everything scales with what is being fought over. Trade, shared institutions, and settled borders shrink the prize, and shrinking the prize shrinks the spending.
• Somebody above both of them. An alliance, a guarantor, or a hegemon can supply the enforcement that anarchy lacks. That works and it moves the dilemma up a level rather than removing it, which is the subject of the hegemony lesson.

What to do with this

When you read about a military buildup, ask what the other side's buildup looks like first. Almost every account is written from inside one of the two chairs, and the sequence reads completely differently from the other one.

Then ask what would make the spending stop being worth it. Verification, a settled border, a guarantee somebody believes, or the stake going away. Those are the levers. Asking each side to be less suspicious is not a lever, because suspicion is the correct response to a situation in which nobody can be certain and nobody can be held to a promise.