# Metric Capture Goodhart's Law states that when a measure becomes a target, it ceases to be a good measure. Metric capture is the institutional form: the organized, often professionalized activity of managing the number instead of the thing the number was supposed to represent. The distinction from ordinary Goodhart drift matters. Goodhart describes a tendency. Metric capture describes **a stable equilibrium in which substantial resources are devoted to metric management, the practice is normalized, and the measurement system's original purpose has been abandoned without ever being formally repealed.** ## The mechanics **Definitional gaming.** Reclassify cases out of the measured category. Crime statistics are the durable example: downgrading offense classifications, discouraging report-taking, and reclassifying incidents all reduce measured crime with no change in events. Because classification requires judgment, this is nearly impossible to distinguish from legitimate discretion case by case. **Selection.** Improve the average by changing who is in the pool. Schools discouraging enrollment of low-scoring students; hospitals declining high-risk surgical candidates to protect mortality rates; agencies triaging easy cases to hit throughput targets. Every one of these improves the metric while degrading the mission, and often harms exactly the population the mission exists to serve. **Threshold effects.** Any bright-line target concentrates effort at the line. Schools graded on the share of students above a proficiency cutoff rationally concentrate resources on students near the cutoff, and rationally neglect both those comfortably above and those far below. **The metric does not merely fail to capture the students it ignores — it instructs the institution to ignore them.** **Timing manipulation.** Move activity across reporting-period boundaries. Same total, better-looking periods. ## Why it is politically load-bearing Metric capture is not merely an efficiency problem. It has three specific political functions. **It launders discretion as objectivity.** A decision made on a numerical criterion appears neutral. The judgment has been moved into the design of the criterion, where it is invisible and unappealable. This is the same relocation documented in [[Discount Rates as Policy Choices]] and [[Dynamic Scoring]]. **It supplies the case for dismantling.** Once a public institution is governed by a metric, the metric can be selected so the institution fails it, and the failure becomes evidence for defunding or privatization. Set an unreachable target, publish the shortfall, cite the shortfall. This pattern recurs across public education, the Postal Service, and the [[Administrative State]] generally, and it is a documented component of the [[Project 2025]] approach to agency restructuring — measurable "efficiency" as the stated justification for capacity reduction. **It converts governance into performance.** Audit culture generates enormous compliance labor whose output is documentation rather than service. The cost is real, falls on frontline staff, and is invisible in the metric, because the metric measures the documentation. ## The reason it is hard to fix The obvious remedy — more metrics, so no single one can be gamed — produces its own failure mode. Composite scorecards multiply gaming surfaces and add weighting decisions, which are themselves contestable and typically unexamined. The choice of weights across a dozen indicators is a policy judgment presented as an index. The other remedy — professional judgment instead of numerical targets — sacrifices exactly what quantification was adopted to provide: comparability, accountability to outsiders, and resistance to favoritism. Those were real goods, and pre-metric public administration was not a golden age. **There is no clean solution, and claiming one is a tell.** The defensible position is that metrics should be treated as evidence to be interpreted by accountable humans rather than as verdicts that execute automatically — which is precisely the arrangement that [[Criminal Risk Assessment Instruments]] and algorithmic decision systems remove. ## Related Goodhart's Law · [[The Politics of Official Statistics]] · [[GDP as a Constructed Measure]] · [[Criminal Risk Assessment Instruments]] · [[Fairness Impossibility Results]] · [[Regulatory Capture]] · [[Administrative State]] · [[Mathematics and Power Hub]]