When the Measure Becomes the Target

Institutions measure what they care about for good reason. A hospital that cannot count its infections cannot reduce them, and a school that never tests its students cannot know whether they are learning. But most of what institutions care about is hard to observe directly: health, understanding, public safety. So they measure something nearby instead, an indicator that stands in for the quality itself. Waiting times stand in for the quality of care, test scores for learning, and recorded offenses for safety. The substitution is usually reasonable, and it is also where the trouble begins.

An indicator is useful only while it keeps moving together with the quality it represents, and it keeps doing so only because no one is trying to move it by itself. Once rewards or penalties are attached, that condition fails. A school judged on its test scores can raise them by drilling the tested material and neglecting what the test omits. A hospital judged on emergency-room waiting times can shorten the recorded wait by changing when the clock is said to start. A police department judged on recorded offenses can make crime fall by recording fewer of them. In each case the indicator improves while the quality it was meant to indicate does not, or gets worse.

It is tempting to explain such outcomes as cheating, and to expect honest management to avoid them. This diagnosis is too comfortable. Much of the distortion involves no dishonesty at all. A teacher who spends more of the year on tested topics is doing what the system plainly asks of her, and an administrator who rearranges a schedule to shorten waits may sincerely believe the patients are better served. The distortion is not a malfunction of the system but a result of its working as designed: people respond to what is measured, and the measure is not the thing.

A familiar remedy is to add more indicators, so that no single number can be pushed up at the expense of the rest. This helps, since raising many indicators together is harder than raising one. But each added indicator becomes a new target with its own distortions, and the burden of reporting grows until it consumes the time that the work itself requires. The opposite remedy, relying on the judgment of trained professionals, avoids the problem of targets but creates another: judgment is difficult to audit, and it can shelter incompetence as easily as it protects good practice.

No arrangement removes the difficulty entirely, but some handle it better than others. One principle is to keep the stakes attached to any indicator modest, so that the temptation to distort it remains small. Another is to treat unusual readings as occasions for inquiry rather than as verdicts: a school whose scores jump suddenly should be asked what changed, and not simply rewarded. Both principles have a price, for modest stakes mean weaker accountability. The choice is not between measuring and not measuring, but between using numbers to close questions and using them to open them.