When a Measure Becomes a Target
In 1975 the British economist Charles Goodhart observed that statistical regularities tend to collapse once they are used for the purpose of control. The idea has since been popularised in a sharper form, usually called Goodhart's law: when a measure becomes a target, it ceases to be a good measure. It is one of those principles that sound almost too neat to be true, until one starts noticing it everywhere.
Consider a hospital that is judged on how quickly patients are seen after arriving in the emergency department. The intention is admirable; nobody wants the sick to wait. Yet once the waiting time becomes the figure on which managers are assessed, the incentive shifts subtly from treating patients promptly to recording that they were treated promptly. Patients may be "seen" briefly to stop the clock, then left waiting elsewhere. The number improves, while the experience it was meant to describe may not.
Schools offer an equally familiar example. Examination results are a reasonable indicator of what pupils have learned, provided nobody is trying too hard to raise them. Once a school's reputation, funding or a teacher's career depends on them, it becomes rational to teach precisely what the examination rewards and little else. Scores rise; whether understanding rises with them is another matter.
What makes the law so persistent is that nobody in these stories need be dishonest. Each individual responds sensibly to the incentives in front of them, and the distortion emerges from the system as a whole. That is also why simply adding more measures rarely solves the problem: every new indicator becomes a fresh target, with its own opportunities for gaming.
The more promising responses are less mechanical. Some organisations rotate the indicators they pay attention to, so that none can be optimised for long; others combine numbers with the judgement of people close enough to the work to see what the figures conceal. Neither approach is as tidy as a single, objective score. But tidiness, Goodhart's law suggests, may be precisely what we should be most suspicious of.