Deterrence is a claim about credibility, not capability

Assessments of deterrence tend to become inventories. Platforms, warheads, divisions, basing, readiness rates: all countable, all comparable, and all only half the problem.

Deterrence is a statement about someone else's belief. It holds when an adversary concludes that acting would cost more than not acting, and that conclusion depends on two independent things: whether you can impose the cost, and whether they think you will. The inventory answers the first. It says nothing about the second, and the second is where deterrence actually fails.

Capability is necessary and rarely decisive

The historical record is unkind to the inventory view. Deterrence failures overwhelmingly do not occur because the deterring side lacked the means. They occur because the challenger did not believe the means would be used, and was often enough right.

The reason is structural. A capability that is obvious is also easy to discount, because the challenger knows it is being shown the capability deliberately. What the challenger is actually trying to assess is whether the political system standing behind it will accept the cost of using it, in the specific circumstance, at the moment it matters. That is a judgement about politics, and it is what the challenger is really studying while everyone else counts hardware.

What credibility is built from

Four things, roughly, and none of them are on the inventory.

Demonstrated cost tolerance. The most credible signal is a past instance of accepting real cost to honour a commitment. It is expensive and it is not manufacturable on demand. This is why credibility built over decades can be spent in a single episode of visible reluctance.

Interest that is legible to the adversary. Commitments the challenger can see a reason for are believed; commitments that appear to rest only on the deterring state's word are not. The question the challenger asks is not "did they promise" but "would I, in their position, actually do it." Which is why proximity, alliance structure and material stake do more work than declarations.

Automaticity. Arrangements that remove the decision at the moment of crisis, such as forward-deployed forces that are unavoidably involved, tripwire deployments and integrated commands, are credible precisely because they take the choice away from a government that might hesitate. The point of the deployment is not what it can do. It is that its presence makes non-response impossible.

Domestic political capacity. A commitment that requires a legislature to authorise it under time pressure, in a polity where that is genuinely uncertain, is discounted accordingly. Adversaries read the domestic politics of the deterring state closely, and they are not wrong to.

The adversary is not counting your platforms. They are estimating whether your politics can survive using them.

The question that can be observed

The inventory is attractive because it is measurable. Credibility is a belief held by a small number of people in another government, and it cannot be observed directly. The temptation is to model what is countable and quietly assume the rest.

The reframing that helps is to abandon "is deterrence holding", which is unanswerable until it fails, in favour of what would have to be true for a challenger to conclude the commitment will not be honoured. That second question resolves into observable things: whether the deterring state recently declined a comparable commitment; is the domestic coalition behind the commitment intact; are the forces that create automaticity actually still in place, or have they been quietly drawn down; has the adversary's own cost tolerance changed.

Deterrence erodes gradually and then fails suddenly, and the gradual part is visible in exactly those variables. A commitment can look identical on the inventory for a decade while the belief underneath it drains away.

Why it resists measurement

The inventory is attractive because it is countable. Credibility is a belief held by a small number of people in another government, and it cannot be observed directly at all. It can only be inferred, from behaviour that is itself partly performance.

So the honest position is that the most important variable in the system is the one nobody can measure, and the measurable ones are proxies of uncertain quality. That is not a reason to abandon the exercise; it is a reason to be clear about which half of the analysis is doing the work.

The uncomfortable corollary is about timing. Deterrence erodes gradually and fails suddenly, and the gradual part is visible in political signals rather than in the order of battle. A commitment can look identical on the inventory for a decade while the belief underneath it drains away, which means the period in which the analysis is most reassuring and the period in which it is least reliable can be the same period.