Goodhart’s Law & Policy Ineffectiveness
The moment a measure becomes a target, it stops measuring anything. Markets and regulators keep relearning this every cycle.
The diagram
The moment the measure becomes the target, it parts ways with the thing it used to measure.
March 3, 2026 7:20 AM EST
Economic Models Series
Goodhart’s Law & Policy Ineffectiveness
Goodhart’s Law — Target vs MeasureWhen a measure becomes a target, it ceases to be a good measureTarget SetBehavior ShiftsGamingDivergenceFailureDay 1MonthsYearsTIME AFTER POLICYMEASURE ≠ REALITY
When Targeting Metrics Breaks Their Predictive Power
Published February 2026
Reading time: 12 min
Goodhart’s Law — Target vs MeasureWhen a measure becomes a target, it ceases to be a good measureTarget SetBehavior ShiftsGamingDivergenceFailureDay 1MonthsYearsTIME AFTER POLICYMEASURE ≠ REALITY
Origin & History
Goodhart’s Law emerged from the work of Charles Goodhart, a renowned economist at the Bank of England, in 1975. Goodhart observed a persistent pattern in monetary policy: whenever central banks selected a particular monetary aggregate as a policy target, the relationship between that aggregate and inflation would break down. This observation evolved into his famous dictum: “Any observed statistical regularity will cease to exist once you put pressure on it for control purposes.”
The law has deep roots in earlier thinking. Donald T. Campbell, a psychologist, articulated a related principle in 1976, now known as Campbell’s Law: the more a social indicator is used for decision-making, the more it becomes subject to corruption and loses its social validity. The two principles are complementary, describing how performance metrics decay when optimized directly.
The phenomenon also connects to the Lucas Critique (Robert Lucas, 1976), which argues that policy parameters cannot be treated as constants when agents alter their behavior in response to policy changes. Together, these insights form a trilogy of fundamental challenges to mechanical policy implementation.
Key Proponents
- Charles Goodhart – Bank of England economist; original observer of monetary aggregate instability under targeting
- Donald T. Campbell – Psychologist; articulated general principle of metric corruption when used for control
- Robert Lucas – Economic theorist; provided microfounded explanation via rational expectations and policy regime shifts
- Marianne Bertrand & Sendhil Mullainathan – Behavioral economists; demonstrated metric gaming in corporate and public sector settings
- Jerry Z. Muller – Intellectual historian; comprehensive analysis of measurement culture and its pathologies
Core Mechanism
The mechanism operates through a fundamental asymmetry: indicators work because agents ignore them. A reliable statistical relationship exists precisely because it reflects autonomous economic behavior, uncontaminated by strategic responses. The moment policy begins targeting the indicator, agents perceive profit opportunities (or face pressure) in gaming the metric rather than changing underlying behavior.
This creates what economists call the cobra effect, named after a colonial India anecdote. The British authorities, concerned about cobra populations, offered bounties for cobra skins. Entrepreneurs simply bred snakes for profit, then released them when the bounty ended, worsening the original problem. Similarly, targeting unemployment statistics may incentivize labor force exodus rather than job creation; targeting hospital readmission rates may discourage appropriate readmissions; targeting student test scores may encourage teaching to the test rather than genuine learning.
The core insight is that what is measured and targeted diverges from what is desired. The indicator becomes decoupled from the underlying economic reality it once reflected. When enough agents behave strategically around the metric, the statistical relationship collapses, and policymakers must search for new reliable indicators—starting the cycle anew.
Mathematical Framework
Consider a simple formulation. Let y represent the true economic outcome of interest (actual inflation, actual unemployment, actual quality of care), and let x represent an observed indicator historically correlated with y:
y = α + βx + ε
where ε is random noise and β is the stable coefficient when no policy targets x. However, once policy targets x, agents respond strategically. The true outcome becomes:
y = α + β’x + γ·policy_signal + ε’
where β’ < β (the original relationship weakens), γ captures strategic responses that decouple x from y, and ε’ is larger (greater variance as agents game the system). The predictive power of x deteriorates, sometimes to statistical insignificance.
The mechanism can be modeled as a principal-agent problem with information asymmetry. When agents possess unobservable effort or substitutable actions, and their compensation or evaluation depends on a measured metric, they allocate effort to gaming that metric rather than producing the desired outcome.
Empirical Evidence
Monetary Aggregates: Goodhart’s original observation showed that M1 and M3 relationships to inflation broke down across countries in the 1980s and 1990s when central banks targeted them. The Bank of England abandoned monetary targeting in the 1980s after such instability became apparent.
Healthcare: Studies of hospital readmission penalties under the Affordable Care Act found that some hospitals reduced readmissions by up to 50%, but this coincided with increases in observation-unit admissions (a different metric) and longer initial stays. The underlying quality outcome remained ambiguous.
Education: No Child Left Behind’s testing regime in the US demonstrated systematic teaching-to-the-test, with some schools explicitly coaching narrow test skills while reading proficiency (the ultimate goal) stagnated. Average test scores rose while standardized SAT scores fell—the metric decoupled from actual achievement.
Corporate Metrics: Companies targeting quarterly earnings per share have been shown to underinvest in R&D and customer satisfaction, reducing long-term shareholder value. The metric optimized (earnings per share) diverged from the outcome desired (sustainable growth).
Police Performance: When departments targeted arrest numbers, arrests increased but crime remained unchanged. When departments targeted response times, response times improved but officer training and investigation quality suffered.
Criticisms & Limitations
Not Inevitable: Critics argue Goodhart’s Law describes a risk, not a law of nature. Carefully designed indicators that align incentives with true outcomes can remain stable indefinitely. Success depends on indicator design and the sophistication of agents gaming it.
Equilibration vs. Collapse: Some economists contend that agent responses to policy targeting eventually equilibrate to a new stable relationship, rather than causing complete collapse. The relationship changes but doesn’t necessarily become useless.
Measurement Improvements: Advances in data collection, real-time monitoring, and multidimensional metrics may reduce the gaming opportunities that Goodhart’s Law exploits. Modern big data environments could make metric corruption more difficult.
Context Dependency: The degree to which Goodhart’s Law applies varies dramatically by domain. In physical systems (temperature, chemical concentration), the law is irrelevant because agents cannot game physics. In behavioral systems (economic metrics, performance measurement), its force varies with agent sophistication and incentive intensity.
Competing Models
Rational Expectations & Lucas Critique: Rather than focusing on metric decay specifically, Lucas emphasized that any policy rule becomes ineffective once agents learn and adapt to it. Goodhart’s Law is a specific manifestation of this broader principle.
Efficient Markets Hypothesis: EMH proponents argue that markets cannot be systematically “gamed” because arbitrageurs exploit any divergence between indicators and fundamentals. Goodhart’s Law assumes persistent agent irrationality; EMH assumes it is competed away.
Indicator Replacement Models: Rather than a law describing permanent failure, some economists view Goodhart’s as a statement that indicator selection requires continuous refinement. Good policy uses multiple cross-checking metrics to prevent single-metric gaming.
5-Phase Goodhart Cycle
Phase 0: Identification
Policymakers observe that a statistical metric (monetary aggregate, unemployment rate, hospital readmission rate) reliably correlates with their policy objective. The relationship appears causal or at least highly predictive. Confidence in the indicator is high based on extensive historical evidence.
Phase 1: Policy Implementation
Authorities explicitly adopt the indicator as a policy target. Resources are allocated, regulations written, and incentives aligned to move the metric in the desired direction. Early results appear successful—the metric improves as predicted, seemingly validating the original correlation.
Phase 2: Agent Gaming
As the policy persists and agents recognize the targeting regime, they discover methods to improve the metric without addressing the underlying objective. Substitution effects emerge: agents shift behavior to the measured dimension while worsening unmeasured dimensions. The metric and the true outcome begin to diverge.
Phase 3: Policy Failure
The original relationship breaks down statistically. The metric continues to improve (or deteriorates unpredictably) while the true objective stalls or worsens. Policy effectiveness declines sharply. Evidence of gaming becomes undeniable—the law has manifested in its full form.
Phase 4: Metric Replacement
Authorities recognize the original indicator has lost validity and search for alternative metrics. This cycle can repeat—the new metric initially works before agents adapt to it as well. Sophisticated policymakers employ multidimensional frameworks to reduce gaming opportunities.
Current Status (February 2026)
Goodhart’s Law remains acutely relevant to contemporary policymaking, with several critical applications in focus:
CPI & Inflation Targeting Debates: Central banks have targeted inflation for decades, yet the relationship between policy tools and CPI outcomes has become increasingly unstable. The 2021-2023 inflation surge challenged the Phillips Curve framework (inflation vs. unemployment relationship), suggesting Goodhart’s decay. Policymakers debate whether core inflation, headline inflation, or alternative measures better reflect true price pressures. Some economists now advocate multi-metric inflation frameworks that resist single-metric gaming.
Unemployment Rate Structural Changes: Labor force participation has shifted dramatically post-pandemic, and the unemployment rate’s relationship to slack in the economy has shifted. Agents have altered labor market behavior (early retirement, gig economy participation), causing the traditional unemployment-inflation relationship to degrade. Policymakers increasingly supplement unemployment with underemployment, prime-age participation, and wage growth metrics.
Yield Curve Inversion Signal: The inverted yield curve has historically predicted recessions, but its predictive power has degraded as central bank intervention in bond markets increased. The metric was “gamed” by massive QE purchases and forward guidance, not by profit-seeking agents, but the result is identical: the signal’s reliability declined precisely when policymakers most wanted to rely on it.
ESG Metrics: Corporate environmental, social, and governance metrics are increasingly targeted by investors and regulators, but evidence of gaming is mounting. Companies “greenwash” or achieve ESG scores while worsening actual environmental outcomes. The metric-reality divergence is accelerating.
What to Watch
Key Developments & Implications
- Monetary Policy Regime Shift: Watch whether central banks explicitly move toward multi-dimensional frameworks that resist single-metric targeting. The Federal Reserve’s adoption of flexible average inflation targeting (2020) and average employment targeting partly reflects Goodhart concerns.
- Alternative Inflation Measures: Monitor proliferation of trimmed-mean, weighted median, and market-based inflation expectations as supplements to headline/core CPI. These represent explicit hedging against Goodhart’s Law applied to traditional inflation metrics.
- Real-Time High-Frequency Data: The shift to real-time, high-frequency economic data (job postings, credit card transactions, shipping data) may slow Goodhart’s decay by making metric gaming more difficult. Agents cannot instantly adapt to multidimensional real-time measurement as easily as they respond to lagged quarterly metrics.
- Labor Market Redefinition: The Fed’s recognition that “maximum employment” requires redefining labor market metrics could represent institutional learning from Goodhart’s Law. The expansion from unemployment alone to a dashboard of labor measures suggests policy sophistication increasing.
- ESG Metric Governance: Corporate disclosure regulation and third-party ESG auditing standards will determine whether ESG scores can resist gaming or become another Goodhart-Law casualty. This is a real-time test of whether better indicator design can prevent metric decay.
- AI-Era Policy Challenges: As AI systems optimize for measured objectives, the incentive to game metrics will intensify. Autonomous optimization compounds agent-based gaming. Watch whether policymakers develop truly game-resistant metrics or resort to regulation preventing AI-based metric optimization.
Implications for Economic Observers
Goodhart’s Law is not merely an academic curiosity—it fundamentally challenges how policymakers can be effective. It suggests that transparency itself is a vulnerability. The moment agents learn what policymakers are optimizing for, they have incentive to distort it. This creates a counterintuitive policy conclusion: sometimes, opacity or metric secrecy improves policy effectiveness by preventing metric gaming.
For investors and economic analysts, Goodhart’s Law implies that the most reliable economic indicators are often those not yet subject to explicit policy targeting. The moment a central bank or government explicitly targets an indicator, discount its predictive power for at least several years during the metric-decay process. The most valuable information comes from cross-checking multiple metrics to identify which relationships are degrading and which remain stable.
The law also suggests that policymakers should be humble about their ability to fine-tune economies. If the metrics upon which policy is based degrade when targeted, then the very act of policy implementation undermines its effectiveness. This argues for policy frameworks that are adaptive, multidimensional, and resistant to metric gaming—frameworks that accept uncertainty as a permanent feature rather than a problem to be solved through metric refinement.
BuildersLens.com — Economic Models Series
Related Signals in the 65-Signal Framework These signals directly connect to this economic theory.
Yield Curve (10Y-2Y)Central banks’ focus on yield curve targets triggers Goodhart’s law invalidation
← Return to 65-Signal Dashboard
Related Signals in the 65-Signal Framework These signals directly connect to this economic theory.
Yield Curve (10Y-2Y)Central banks’ focus on yield curve targets triggers Goodhart’s law invalidation
← Return to 65-Signal Dashboard
Related Signals in the 65-Signal Framework These signals directly connect to this economic theory.
Yield Curve (10Y-2Y)Central banks’ focus on yield curve targets triggers Goodhart’s law invalidation
← Return to 65-Signal Dashboard
Educational content describing an economic theory; inclusion is not endorsement. Not investment advice.