Lucas Critique
A model built from how people behaved under the old rule tells you almost nothing about how they’ll behave once they know a new rule is in place.
What Is It?
The Lucas Critique, from economist Robert Lucas’s 1976 paper “Econometric Policy Evaluation: A Critique,” argues that it’s naive to predict the effects of a policy change by extrapolating from historical data collected before the policy existed. The relationships in that historical data, how much people spent, saved, or worked in response to past conditions, aren’t fixed laws of behavior. They’re the byproduct of decision rules people were using under the old policy regime. Change the regime, and the decision rules change with it, which breaks the very relationship the prediction was resting on.
Lucas’s original target was 1970s macroeconomic forecasting, where models built on decades of historical data kept failing to predict how the economy would actually respond to new government policy. The paper reshaped mainstream economics and contributed to Lucas’s 1995 Nobel Memorial Prize, largely for the rational-expectations framework that grew out of it.
Why Does It Matter?
The Lucas Critique helps explain why patterns like Goodhart’s Law and the Cobra Effect appear so reliably. Goodhart’s Law says a measure stops being a good measure once it becomes a target. The Cobra Effect describes an incentive that produces more of exactly the wrong thing. Both are describing outcomes, and each can arise through its own specific mechanism, statistical selection, proxy optimization, incentive design gone wrong, without necessarily requiring a historical predictive model at all. The Lucas Critique names one important reason those outcomes show up as often as they do: any model of behavior built from data collected before an incentive existed is a model of people who didn’t yet know that incentive was coming. Once they do know, they optimize against the new information, and the old model’s predictions stop applying, not because it was built badly, but because it was built for a world that the announcement itself just ended.
This shows up constantly whenever a new KPI, incentive structure, or policy gets introduced on the strength of a pilot or a historical trend. “Teams that did X performed better, so let’s require X of every team” treats X as a stable cause, when X may have only correlated with good performance because nobody was yet being evaluated on whether they did X.
What Changes Once You See It?
You stop treating a new incentive, KPI, or policy as something whose effects can be fully forecast from how people behaved before it existed. The forecast is a hypothesis about a world that hasn’t happened yet.
It is not a measurement of one that has.
You start watching the first real cycle under a new rule as fresh data, worth its own honest observation, rather than as confirmation of what the pre-launch model already predicted.
You start asking, before rolling out a new measure, not just “what did the data show under the old system” but “how will the people this affects change their behavior once they know this is how they’re being judged.”
Common Misunderstandings
- It is not a claim that historical data is useless. It’s a claim that historical data can’t be extrapolated across a genuine regime change without adjustment, since the regime is part of what produced the data in the first place.
- It is not the same claim as Goodhart’s Law, even though the two are closely related and often travel together. Goodhart’s Law is about what happens to a specific measure once it becomes a target. The Lucas Critique is the broader, underlying reason: any behavioral prediction built on old data is vulnerable once the people being predicted learn the rule has changed, whether or not a specific measure is even involved.
- It doesn’t mean prediction is hopeless. It means the honest version of prediction treats a new incentive as an experiment to observe, not a foregone conclusion already proven by the data that preceded it.
- It isn’t limited to economics, where it originated. It applies anywhere a leader tries to forecast the effect of a new rule from how people behaved under the old one.
Diagnostic Question
What evidence do we actually have about behavior under the new rule, rather than the old one?
Explore Further
Field Notes
None yet.
Related Field Guide
Origin
Robert E. Lucas Jr. published “Econometric Policy Evaluation: A Critique” in the Carnegie-Rochester Conference Series on Public Policy in 1976. The paper is regarded as one of the most influential in macroeconomics, and its underlying insight, rational expectations, contributed to Lucas’s 1995 Nobel Memorial Prize in Economic Sciences.