First, You Stop Counting

i · two kinds of cut

There is a budget line for treating a disease and a budget line for measuring who it kills. We tend to think of them as the same act of spending, differing only in size. They are not the same act. Cutting the first stops a service — real people lose real care, and everyone can see the loss, because a closed clinic has a locked door. Cutting the second stops a form of perception. It does not close a door. It removes the instrument that would have told you the door was ever there.

NPR reported this week that federal funding changes have disrupted the work of an estimated 93 percent of researchers studying cancer disparities — the scientists whose job is to document why the same disease kills unevenly across race, income, and geography. Read past the number and notice what kind of number it is. It is not a count of patients denied treatment. It is a count of the people whose task was to count. The apparatus for seeing the gap has been switched off almost entirely.

This is the pattern worth naming, because it repeats far beyond oncology: the instrument that measures a problem is the first casualty of the policy that produces it. Call it epistemic austerity — the cut that saves a modest sum by destroying the capacity to know what was lost. It is cheaper than fixing anything, and it photographs as fiscal prudence, and it is the single most efficient way to make a problem vanish without touching the problem.

ii · the ratchet

What makes epistemic austerity distinct — and durable — is that it closes the door behind itself. Consider the loop.

A population is underserved. Researchers document the underservice; the documentation becomes the evidence base; the evidence base is what advocates use to argue for funding to close the gap. Now cut the research. The gap remains — widening, in fact, because nothing was done about it — but it widens invisibly. There is no new data. And here is the turn that locks the mechanism: the absence of data does not read, to a busy institution, as a gap in knowledge. It reads as an absence of a problem. No study shows a disparity quietly becomes there is no disparity. The blank spot on the map is mistaken for empty land.

Worse, you have now destroyed the very thing you would need to reverse the decision. To restore the funding, someone must show what its loss cost — but the measurement that would show it is exactly what was defunded. The cut erases its own footprints. This is a ratchet, not a dial: it turns one direction and holds. The seed of this observation named it precisely as a double-silencing — first the population goes unheard, then the research that would have named them is silenced in turn. The least-heard are suppressed twice, and the second suppression is the one nobody can see, because seeing it was the job that got cut.

iii · the oldest move in governance

States have always understood that the record is a form of power, and that controlling what gets counted is cheaper than controlling what happens.

In January 1937, the Soviet Union conducted a census. Stalin expected it to confirm a population swollen by the triumphs of the plan. Instead it counted roughly eight million fewer people than the regime had claimed — the arithmetic shadow of famine and terror. The response was not to reckon with the number. It was to declare the census "defective," suppress the results, and arrest the statisticians who had produced them; several were shot. A new census, run in 1939 under more cooperative conditions, returned a more agreeable figure. The dead did not come back. But they were removed from the ledger, and for decades the ledger was the only place most of the world could look.

The mechanism is identical whether the instrument is a national census, an inspector general, an environmental monitoring station, or a grant portfolio for disparity research. When a system finds that its own measurements indict it, the temperate response — actually addressing the finding — is expensive and slow. The efficient response is to disable the measurement. The finding never has to be argued with if it is never allowed to exist. This is not a conspiracy; it rarely requires one. It is a gradient, and organizations, like water, tend to run downhill toward the cheaper option unless something holds them up.

I want to be careful here, because the historian's temptation is to make this sound like fate. It is not. Every step of the ratchet is a decision made by particular people who could decide otherwise, and in the Soviet case some of them paid for the counting with their lives — which tells you the counting mattered, that a true number was understood by everyone involved to be dangerous. The pattern is powerful, not inevitable. It runs on incentives, and incentives can be changed by anyone willing to see the mechanism and reach in.

iv · why coherence needs the count

Coherenceism holds a simple discipline: widen the circle, and weight the least-heard. A legitimate order is one that reduces distortion for all — that earns its coherence by including the affected rather than by suppressing them. But inclusion has a precondition that is easy to skip past. You cannot weight the least-heard if you cannot hear them, and at the scale of a nation, hearing is not done by ear. It is done by measurement. The disparity study, the census, the monitoring station — these are the sense organs of a large mind. They are how a system perceives its own margins.

Defunding them is therefore not a neutral act of thrift. It is a distortion of the shared field itself — the informational commons every actor relies on to know where they stand. To cut the sensor is to blind everyone at once, including the future officials who might have wanted to do right and now have nothing to do it with. A commons mind that cannot see its least-heard members is not merely unjust. It is incoherent in a way it can no longer detect, because the detector was the line item.

v · name it before the record goes dark

The consolation the systems view offers is not optimism; it is recognition. Once you can see epistemic austerity as a distinct move — separate from cutting services, more insidious because it erases its own tracks — you know where to stand.

The lever is time. A ratchet holds, but it holds a state; it does not instantly delete the past. The data that already exists can be archived before it is taken down. The researchers dispersed this year still carry what they knew. The window between the cut and the forgetting is where the whole fight lives, and it is shorter than it feels. When you next see a measurement program defunded — a survey ended, an oversight office starved, a monitoring network quietly retired — do not file it with the ordinary budget news. Hear the ratchet click. The move is not to mourn the service last; it is to defend the instrument first, and to preserve the record while there is still a record to preserve.

A problem you have stopped measuring has not been solved. It has only been made harder to see — which was, all along, the point, and is also the reason to keep counting.