When the Map Changes the Territory

Consider what happened to British crime statistics in the 1970s. When police precincts began to be evaluated on clearance rates — the percentage of reported crimes that led to an arrest — the clearance rates went up. This should have been good news. Instead, it turned out that detectives had begun pressuring victims to withdraw their reports, reclassifying serious crimes as minor ones, and in some cases simply not recording crimes at all. The measure had become a target, and in becoming a target, had ceased to be a measure. The map had changed the territory.

This is sometimes called Goodhart’s Law, after the British economist Charles Goodhart, who observed it first in monetary policy: once a measure becomes a target, it ceases to be a good measure. The law is usually invoked to describe dysfunction — gaming, teaching to the test, optimizing the metric while hollowing out the thing the metric was supposed to represent. What interests me is something slightly different: not the perversion of a measure, but the way that sufficiently widespread models begin to reshape the systems they describe, even when everyone is acting in good faith.

The most vivid example from the last century is the efficient market hypothesis. When the idea that stock prices fully reflect all available information became the dominant theory in finance, it began to change how markets actually worked. Index funds proliferated because active management couldn’t reliably beat an efficient market. Algorithmic trading spread because any predictable pattern would be arbitraged away before it could be exploited. The theory that markets are efficient helped make them more efficient — not by revealing a pre-existing truth, but by changing the behavior of the people inside the system. This is a kind of feedback that has no analogue in physics: a theory about gravity does not cause objects to fall differently. A theory about how markets behave, if widely believed and acted upon, changes how markets behave.

The peculiar situation of theories about human systems is that they can be self-fulfilling or self-defeating in ways that theories about the physical world cannot. What economists call “performativity” — the capacity of economic models to enact the worlds they describe — is a specific case of a more general phenomenon. Any model of a social system, if it becomes influential enough, becomes part of the system it was built to understand. It is made of the same material as the territory it was meant to map.

This should change how we hold our models — not with less rigor, but with a different kind of attention. The relevant question isn’t only whether a model correctly describes the world as it currently exists. It’s also what world the model will help produce if enough people believe it and act accordingly. That is a harder question. It doesn’t submit neatly to hypothesis testing. But for models with any ambition to be useful — models that will actually inform policy, shape institutions, guide decisions — it is the question that matters most.