Ambiguity is not noise
in the measurement

01Where the question comes from

A genesis problem, not a performance problem

The standard account of state capacity runs through prerequisites: a fiscal base, a professional bureaucracy, rules applied consistently to cases defined in advance. Most of the states this lab cares about never had those in deployable form, and some had them destroyed. Several of them govern anyway.

Explaining how they do that, without either apologising for what it costs or pretending the standard route was ever available to them, is the problem underneath everything else we do.

02The finding that organises our thinking

Constitutive ambiguity

In one ethnographic survey of Rwanda's gacaca courts, over seventy per cent of participants said that people lied in the proceedings. Over ninety per cent said gacaca would bring peace.

Once you have seen that, you cannot go back to treating underspecification as a defect to be scored down. You have to ask what it is doing.

03The question we ask instead

The bidirectionality condition

If ambiguity can be the medium of governance, what separates it from the medium of predation? Not the amount. Authoritarian collapse is not short of ambiguity.

The difference is direction. A gray zone that extracts and also delivers — labour for infrastructure, confession for release, performed allegiance for civic standing — sustains governing. A gray zone that only extracts becomes predation. So for any institution operating through strategic underspecification, our first question is not whether it is clear. It is: who is getting anything out of it?

This is the most portable tool we have, and we point it at ourselves as often as at the states we study. It is also what we point at the funders — see below.

04From negotiating to anticipating

Attributive ambiguity

Disaster governance is where this family of ideas leaves home. A flood or a dam failure can be truthfully described both as a climatic extreme and as a failure of governance, and which description settles is decided afterwards. Decision-makers know this in advance.

So the actor is no longer negotiating inside a gray zone. It is anticipating an attribution that has not happened yet — and building accordingly. Visible response capacity grows because it will be credited. Quiet prevention lags because its success looks like nothing happening. We call the result deformed specialisation.

Climate change sharpens this. It raises real risk and simultaneously hands every governance failure an alibi.

05What this implies for measurement

Two commitments that follow directly

• Do not aggregate across a distinction that carries the signal. When a capacity index folds a prevention component and a response component into one score, it destroys the only thing the score needed to show. Existing instruments often construct the distinction and then remove it in the last step.
• Report the confidence, not only the value. In a plural world, honest comparability comes from documenting the limits of comparison rather than pretending them away. We would rather publish a wide interval that is true than a point estimate that is tidy.

Both commitments make our output less convenient to use. That is the cost of the position, and we would rather pay it than produce a number that flatters whoever commissioned it.

06Who benefits from the confusion

The bidirectionality condition, turned on the funders

It would be comfortable to say that funders simply cannot distinguish prevention from response. Our own framework does not let us stop there. Ask who the gray zone delivers to, and the answer includes them.

• Response projects disburse faster and hit annual deployment targets.
• Response produces visible results that can be shown to a legislature.
• Prevention's success is an absence, creditable to no particular disbursement.

07The trap we expect to fall into

Substitution, applied to ourselves

One more idea from the same family. Institutions that succeed by substituting — community courts standing in for an independent judiciary, mandatory labour standing in for a tax base — tend to occupy the ground where their own successor would have grown. The better the substitute works, the harder it is to leave. We call this the substitution trap.

It applies to us. If a diagnostic we build becomes the accepted way to assess adaptation capacity, it will occupy the ground where a better instrument would have appeared, and our incentive to say so will fall to zero. We do not have a solution to this. We have a practice: publish the method so others can replace it, state the falsification conditions in public, and treat our own adoption as something to be audited rather than celebrated.

08What follows in practice

How this shows up in the work

• Everything is released as open source, under licences that permit replacement.
• We publish falsification conditions before results, not after.
• We do not produce country rankings.
• We state what we do not have as plainly as what we do.
• Where a question is empirical, we would rather be shown wrong early than be agreed with late.