Okun's Law

Origin and the original paper

In 1962 Arthur Okun, then on the staff of John F. Kennedy's Council of Economic Advisers, was asked a practical question: how much GDP would the United States gain if unemployment fell from its then-prevailing 6.7% to a target 4%? Okun pulled quarterly data from the postwar period and ran a simple regression. The answer, in his celebrated paper "Potential GNP: Its Measurement and Significance," was that closing 2.7 points of unemployment would raise output by roughly 8 points — a ratio of about three to one.

Okun called the result a rule of thumb, not a structural relationship, and he wasn't shy about its empirical roughness. But the regression has held up remarkably well across six decades, dozens of countries, and several recessions, which is why every undergraduate macro textbook and every central-bank briefing book contains some version of it.

The two versions

Okun's law exists in two algebraic forms. They are equivalent in steady state but differ in what they require you to measure.

Gap version. The output gap is proportional to the unemployment gap:

(Y − Y*) / Y* = −β · (u − u*)

where Y is real GDP, Y* is potential GDP, u is unemployment, u* is the natural rate, and β ≈ 2 in modern U.S. data. The trouble is that Y* and u* are unobservable. Estimates from the CBO, the Fed, and the IMF often disagree by a percentage point or more — enough to flip a recession verdict.

Growth version. The change in unemployment is related to GDP growth:

Δu = −β · (g − g*)

where g is real GDP growth and g* is trend (potential) growth, often around 2% in advanced economies. This version uses only observed quantities and is the form most commonly used in real-time forecasting.

Worked example — translating an output gap into jobs

Suppose the U.S. economy in some year has potential GDP of $25 trillion and a natural unemployment rate of 4%. The CBO publishes a real GDP estimate of $24 trillion. The output gap is:

gap = (24 − 25) / 25 = −4%

Plugging into Okun's law with β = 2:

−0.04 = −2 · (u − 0.04)
u − 0.04 = 0.02
u = 6%

The law predicts unemployment should be near 6%. If the actual print is 5.5%, the labor market is tighter than the output gap implies — perhaps a productivity slowdown is in the data, or potential GDP has been overestimated. If unemployment is 7%, the labor market is looser than expected, possibly because employers have made unusually deep layoffs. Either deviation is itself a signal worth investigating.

Reverse the calculation for policy: if a fiscal stimulus is projected to raise output by 2 percentage points, Okun's law predicts unemployment will fall by roughly 1 point. The CBO uses exactly this kind of arithmetic to score the labor-market effects of legislation.

Why the ratio isn't 1:1

Naively you'd think a 1% drop in employment should produce a 1% drop in output. The empirical 2:1 (or 3:1) coefficient reflects three margins beyond the headcount.

• Hours per worker. Recessions cut hours before they cut bodies. A firm that needs to reduce labor input by 5% might first move full-time staff to four-day weeks rather than fire 5% of them. Output falls; the unemployment rate doesn't move.
• Labor force participation. Discouraged workers leave the labor force entirely. They're no longer "unemployed" because they're no longer "actively seeking work" — but their lost output is still missing from GDP. This effect is large in deep U.S. recessions and explains a meaningful part of the 2:1 multiplier.
• Productivity. In slumps, firms hold on to their best workers (labor hoarding). The marginal worker who keeps her job produces less because demand has collapsed. Productivity per hour falls, magnifying the output decline.

Okun's law vs other macro relationships

	Okun's law	Phillips curve	Taylor rule	Solow growth model
Type	Empirical regularity	Empirical regularity	Policy rule	Theoretical model
Variables linked	Unemployment ↔ output gap	Unemployment ↔ inflation	Inflation, output gap → policy rate	Capital, labor → output
Time horizon	Quarter to year	Quarter to several years	Real-time policy	Decades
Policy use	Translating GDP to jobs	Inflation-targeting trade-off	Setting interest rates	Long-run growth analysis
Key parameter	Okun coefficient ≈ 2	Slope of curve	Reaction coefficients	Capital share α
Status	Stable, country-specific	Unstable post-1970	Followed approximately	Workhorse, contested
Originator	Okun (1962)	Phillips (1958)	Taylor (1993)	Solow (1956)

The coefficient varies across countries

Okun's β is not a universal constant. It reflects how flexible labor markets are and how firms adjust hours, productivity, and headcount during downturns.

• United States. β ≈ 2. Quick layoffs, modest hour adjustments. Postwar value was nearer 3 because productivity behavior was different.
• Spain. β closer to 1. Rigid employment law historically pushed adjustment onto the headcount rather than hours, so unemployment swings track output more tightly.
• Japan. β around 0.3 to 0.5. Lifetime-employment norms keep layoffs rare; output falls without unemployment moving much.
• Germany. β fell sharply after Kurzarbeit short-time work programs were institutionalized; in 2008–2009 output collapsed but unemployment barely moved.

Counterarguments and qualifications

The "jobless recovery" challenge. After the 2001 and 2008 recessions, U.S. unemployment fell more slowly than Okun's law predicted given the recovery in GDP. Some economists, notably Mary Daly and Bart Hobijn at the San Francisco Fed, argued the relationship had broken; others, like Laurence Ball, showed it tightened only when potential GDP itself was revised down. Once you fix the potential-GDP estimate, the law mostly held.

Asymmetry. Several studies (Cuaresma 2003, Lee 2000) find that β is larger in recessions than in expansions: unemployment rises faster when output drops than it falls when output rebounds. The implication is that the simple linear regression is hiding a piecewise-linear truth.

Time-varying coefficients. Knotek (2007) and others show β has drifted over decades — closer to 3 in the 1950s, near 2 today, possibly trending lower as service sectors with less labor hoarding grow. Forecasting models that use a fixed β can mislead.

The structural-change critique. Some macroeconomists argue Okun's law confounds three different phenomena: the cyclical relationship, the productivity-cycle relationship, and the participation-rate response. A reduced-form regression hides which is doing the work.

Common pitfalls

• Treating the coefficient as universal. β is a property of an economy, not of nature. Importing the U.S. value to Japan will mislead.
• Ignoring potential-GDP revisions. The output gap is the difference between actual and potential. When the CBO revises potential, the historical Okun ratio changes even if no data point moved.
• Using the gap version for forecasting. The gap version requires knowing u* and Y*, both unobservable. The growth version, using only changes, is more honest about uncertainty.
• Reading short-run noise as structural shift. Quarterly data has measurement error; one or two off-pattern points often revise away when the next vintage of data arrives.
• Forgetting hours. A flat unemployment rate with falling output usually means hours are absorbing the shock, not that Okun's law has broken.