Academic and research sources
Potential GNP: Its Measurement and Significance
Okun's original output-gap and unemployment-gap link.
Okun's 1962 Cowles Foundation paper documented the empirical regularity that each percentage-point rise in unemployment above its trend level was associated with a roughly two-to-three percent shortfall of output below potential, using U.S. postwar data. Accessed 2026-11-14.
Cowles Foundation Paper 190 · 1962
Okun's Law: Fit at Fifty?
Ball, Leigh, and Loungani on cross-country stability and variation.
Ball, Leigh, and Loungani 2013 found that Okun's Law remained a robust and stable empirical relationship in the United States since 1948 and across twenty advanced economies since 1980, with the Okun coefficient varying across countries but averaging near two. Accessed 2026-05-23.
NBER · 2013
Research footing
Evidence and data
Use output and potential output to form a percent output gap, and unemployment and its natural-rate benchmark to form a percentage-point unemployment gap. Revisions to either benchmark can materially change the fitted relation.
Calibration or measurement
Beta is measured in output-gap percentage points per one unemployment percentage point. In u - u* = -output gap/beta, a larger beta means a smaller unemployment response to a fixed output-gap move.
Boundaries
- The linear relation is a local reduced-form approximation, not a causal theory of recessions.
- Participation, hours, productivity, labor hoarding, and sector mix can move beta.
- Extrapolation can imply impossible unemployment rates, and real-time potential estimates can be heavily revised.
Use Guidance
When sufficient
Use the page as a local cyclical consistency check near potential when output-gap percentages and unemployment-gap percentage points are measured on the same horizon and beta is appropriate to the country and period.
When sketch only
Do not use where the labor market is undergoing structural change. Labor-force participation shifts, sectoral reallocation, and hysteresis after a deep recession can all move the coefficient. The reduced-form relationship does not identify the shock that moved either output or unemployment.
When to switch
Switch to a search-and-matching model (Mortensen-Pissarides) when job-creation and job-destruction margins are the object of interest. Switch to a hysteresis-augmented Phillips curve or a model with state-dependent natural-rate estimation when a long slump has plausibly shifted the structural unemployment rate.
Falsification signal
Persistent, signed out-of-sample residuals near potential or a stable re-estimate of beta that differs materially from the supplied value rejects that calibration's local mapping. Such evidence signals coefficient drift or omitted margins; it does not falsify an identity because Okun's law is an empirical approximation.