A government announces it will spend five per cent of national income on security and defence by the middle of the next decade. The figure is debated for a week. Is it enough, is it deliverable, how does it compare with the neighbours. What almost no one asks, because the conversation has no comfortable place to put the question, is the only one that matters: at the end of all that spending, will the country actually be more able to withstand the things that could break it?
The percentage is a proxy. It stands in for security because security itself cannot be counted directly. You can audit a budget line; you cannot audit resilience, at least not in time to put it in a press release. So the proxy gets managed and the thing it was meant to represent slips quietly out of view. This is not a failure of intelligence or good faith. It is what happens whenever a number that was invented to approximate something gradually becomes the something. The measure was a finger pointing at the moon, and the institution, asked to report on the moon, files quarterly figures on the finger.
The pattern has a name in the way organisations decay. A target chosen because it correlated with what you wanted becomes the object of effort in its own right, and the correlation it relied on quietly breaks. This is proxy-capture, and national security is now a textbook case of it. The two-per-cent and five-per-cent figures were never security. They were a rough signal that a country was taking the matter seriously. Treated as the goal, they reward whatever raises the number and ignore whatever does not, and a great deal of what makes a country hard to break does not show up in a defence budget at all.
You can watch the selection happening in the way risk itself is catalogued. The state keeps a register of the things that could go wrong, and the published version runs to dozens of discrete hazards: an attack, a flood, an outbreak, an accident. These are acute risks. Each is datable, attributable, and amenable to a visible response, and a minister can be photographed responding to one. But the same government recently had to commission a separate analysis of chronic risks, because the main apparatus does not natively see the slow, compounding kind: the steady erosion of a capability, the dependency that deepens for years before it snaps, the institutional knowledge that leaves and is not replaced. The split is the tell. The system sees clearly the risks shaped like the instruments it already owns, and badly the ones that are not.
This is what the-legibility-trade costs. To manage something at scale you must first make it legible, reduce it to what can be measured and compared, and the reduction always discards exactly the parts that resist measurement. In a living system, those discarded parts are often the source of its resilience: the redundancy that looks like waste, the local judgement that cannot be written into a procedure, the slack that an efficiency drive removes first. A security architecture optimised against legible proxies will, over time, dismantle the illegible capacities that actually let a country absorb a shock, and every quarterly report will show improvement while it does so. Call it false resilience: a system that satisfies every indicator while becoming less able to do the thing the indicators were standing in for. It is worth being precise about what-resilience-actually-means here, because the word has been worn smooth by use: not the capacity to predict and prevent named events, but the capacity to absorb and recover from the ones nobody named.
There is a deeper reason this persists, and it is structural rather than personal. The people accountable for security operate inside an incentive architecture that rewards the visible and the attributable. A disaster prevented leaves no trace and earns no credit; a disaster managed in public earns a great deal. So investment flows toward the things that can be seen and claimed, and away from the diffuse, long-horizon, unglamorous work whose only reward is that nothing happens. Naming this is not cynicism. It is the precondition for designing around it, because a problem misdescribed as a shortage of resources or competence will be met with more money and better people, and it will not move.
The correction is not another metric. It is a change in what we ask the existing ones to do. A number is a useful servant and a disastrous master. The honest question, put before any figure is agreed, is whether the thing being funded buys real capacity to absorb and recover, or merely raises a proxy that lets everyone declare the matter handled. The first is security. The second is the appearance of it, bought at the price of the real thing, and the gap between them is exactly where a country discovers, too late, what it had quietly traded away.
Further reading: why-a-country-feels-ungovernable follows the same loss one level up, asking what happens to a society when the integrating function that once held the whole picture together goes missing.
Garden notes
- proxy-capture is the mechanism named in the abstract; this essay is one specimen of it operating at the scale of the state.
- the-legibility-trade explains why the illegible capacities get discarded first, and why that discarding looks like progress on every chart.
- what-resilience-actually-means separates the capacity to absorb shock from the appearance of preparedness that proxies reward.