Okun’s Law: GDP and Unemployment

By Ryan O'Connell, CFA, FRM Published November 5, 2026

An economy can produce more goods and services while its unemployment rate rises. It can also recover output before employers begin hiring again. Okun’s law helps explain the usual connection between real GDP and unemployment, while leaving room for those outcomes. Its usefulness depends on knowing which version of the relationship you are using, how its coefficients were estimated, and whether the units and time periods match.

This guide connects gross domestic product with the unemployment rate and labor force. It explains both the growth/change version and the output-gap version, using hypothetical numbers throughout the worked examples. Those examples illustrate the mechanics; they are not estimates of today’s economy or unemployment forecasts.

What Is Okun’s Law?

Okun’s law is an empirical relationship between economic output and unemployment, named after economist Arthur Okun’s work in 1962. Stronger real output growth generally accompanies a larger decline, or smaller increase, in the unemployment rate. Output below estimated potential is generally associated with unemployment above its estimated normal level.^[1]

The word “law” describes a statistical regularity. It does not establish an accounting identity, a guaranteed result for each year, or the causal effect of a particular policy.

Percentages and Percentage Points

The unemployment rate is a percentage of the labor force. In U.S. statistics, that labor force consists of employed and unemployed people; unemployment generally requires availability and active job search, with an exception for qualifying temporary layoffs awaiting recall.^[4] Once a rate is calculated, subtracting one rate from another produces a percentage-point difference.

A move in unemployment from 6% to 5% is a decline of one percentage point. Also distinguish GDP growth from a change in its growth rate: moving from 2% growth to 3% growth is a one-percentage-point increase in growth. The formulas below enter a 3% rate as the number 3, rather than 0.03.

The Growth/Change Version of Okun’s Law

A simple difference version relates the unemployment-rate change to real GDP growth over a compatible interval. This article defines the slope magnitude, b, as positive and writes the minus sign explicitly. Some sources instead put a negative coefficient directly before growth.^[1]

• Δu_t = u_t – u_t-1: the unemployment-rate change, in percentage points.
• u_t and u_t-1: unemployment rates for the current and preceding periods, in percent-number units.
• g_t: real GDP growth for that interval, in percent.
• a: the fitted unemployment-rate change at zero growth, in percentage points.
• b: the positive slope magnitude, in unemployment percentage points per one percentage point of real GDP growth.
• e_t: the residual, in unemployment percentage points: actual change minus the change explained by the fitted line.

For the annual examples here, t is a calendar year, real GDP is that year’s annual output, and u is the annual-average unemployment rate. Thus the GDP comparison and unemployment comparison both run from one calendar year to the next. A model estimated with another convention needs its own compatible inputs.

Use real rather than nominal GDP: a rise in dollar spending caused solely by higher prices does not represent more real production.

Growth Associated with Stable Unemployment

Set the fitted unemployment change to zero, leaving the residual at zero. Solving 0 = a – b × g gives a growth threshold:

The same fitted equation can be written as Δu_t = -b × (g_t – g_stable) + e_t. This makes an important point visible: positive GDP growth can coexist with rising unemployment if growth falls short of the fitted threshold.

That threshold is calculated from a particular regression’s intercept and slope. It is not an automatically observed, permanent potential-growth rate. Its interpretation depends on the sample, labor-force trends, productivity and the model’s assumptions.

Okun’s Law Example: Annual GDP Growth

The decline from 6.0% to 5.2% is 0.8 percentage point. Measured relative to the starting rate, it is a 100 × [(5.2 / 6.0) – 1] = approximately 13.33% decline. Okun’s equation uses the percentage-point change.

Now hold those hypothetical annual coefficients fixed and compare alternative annual growth rates. The fitted stable-unemployment threshold is 0.80 / 0.40 = 2%. The chart shows the model’s schedule; its points are possible growth scenarios, not observations from successive years.

Above 2% growth, the fitted unemployment change is negative. Below 2%, it is positive. For instance, even 1% positive growth gives 0.80 – (0.40 × 1) = a 0.40 percentage-point increase in this model.

Actual results need not lie on that line. If Year 2’s observed annual-average unemployment were instead 5.5%, the observed change would be 5.5 – 6.0 = -0.5 percentage point. The residual would be -0.5 – (-0.8) = +0.3 percentage point: unemployment fell less than the fitted relationship suggested.

The Output-Gap Version of Okun’s Law

The gap version compares actual conditions with estimated sustainable benchmarks. Potential output is the economy’s estimated sustainable production level. It is not the most output physically imaginable if everyone worked continuously. The natural or noncyclical unemployment rate is an estimated benchmark for unemployment apart from cyclical weakness; it is generally above zero.^[5][9] See types of unemployment for the distinction between cyclical unemployment and other sources of joblessness.

• Y_t: actual real output in period t.
• Y*_t: estimated potential real output for the same period, on the same price basis.
• q_t: output gap, in percent of potential output; negative means below potential.
• u_t and u*_t: actual and estimated natural unemployment rates, in percent-number units.
• u_t – u*_t: unemployment gap, in percentage points; positive means above the benchmark.
• c: positive slope magnitude, in unemployment percentage points per one percentage point of the output gap.
• v_t: residual, in unemployment percentage points.

With this actual-minus-potential convention, a negative output gap pairs with a positive fitted unemployment gap. Some presentations define a positive output shortfall instead; their displayed sign changes accordingly. Always read the definition before interpreting the coefficient.

A Separate Hypothetical Gap Example

To see the signs on either side of potential, keep this separate scenario’s coefficient and estimated benchmarks fixed, and compare alternative output gaps. The vertical quantity is the unemployment gap, not the unemployment rate or its annual change.

Output below potential is associated with unemployment above its benchmark. At a zero output gap, fitted unemployment equals the natural rate, which does not mean zero unemployment. The picture also permits output temporarily above estimated sustainable potential and unemployment below its estimated benchmark.

Growth Version vs. Gap Version

Growth is a change; a gap is a position. Output can grow rapidly while remaining below potential after a deep downturn. Conversely, output can grow slowly while its level remains above estimated potential. A statement about one quantity does not determine the other without more information.

Do not mechanically equate b and c, or treat them as reciprocals. The examples deliberately use different hypothetical coefficients. Deriving a growth equation from a gap equation requires assumptions about changes in potential output and natural unemployment, parameter stability, residuals and the growth approximation. Separately estimated regressions need not satisfy those assumptions.

Likewise, algebraically rearranging a fitted equation to put output on the left does not produce a separately estimated regression of output on unemployment. The direction of estimation and the treatment of unexplained variation matter.

Why Output and Unemployment Do Not Move One for One

Producing more output does not always require a proportional increase in the number of employed people. Employers can change hours and utilization before changing staffing, and productivity measures output per hour rather than the number of people employed.^[6][8]

These are possible adjustment channels, not a formula assigning each one a fixed contribution. In particular, a falling unemployment rate can partly reflect people leaving the labor force rather than finding work.^[4] Check participation, employment and hours alongside the headline rate. For the longer-run forces behind productive capacity, see economic growth and productivity determinants.

Why Okun’s Coefficients Differ

An Okun coefficient belongs to a country, sample and specification. Ball, Leigh and Loungani’s research across advanced economies finds substantial differences in coefficients across countries, partly reflecting features of national labor markets.^[2] A coefficient estimated for one economy should not automatically be transferred to another.

Within a country, changing the estimation period, quarterly versus annual frequency, output measure, lag structure or benchmark-estimation method can change the result. Including recessions rather than only expansion years can also change a fitted relationship. A contemporary response to growth and a response spread over several periods are different objects.

Researchers differ on how stable the relationship is. Cleveland Fed research emphasizes instability in simple rolling estimates, while Ball and coauthors find broad stability in most countries in their study. St. Louis Fed researchers discuss how differing frequencies and methods help explain such findings.^[7][2][9] The practical lesson is to inspect the model and evidence behind a quoted rule rather than assume either perfect stability or complete uselessness.

How to Calculate and Use an Okun’s Law Scenario

1. Choose the question. Use the growth version for a change in unemployment; use the gap version for unemployment relative to an estimated benchmark.
2. Identify the model. Record its country, sample period, variables, sign convention, coefficients and lags. Keep an illustrative classroom equation labeled hypothetical.
3. Match the time periods. For an annual model like the example, compare annual real GDP levels and annual-average unemployment rates from the same two years. Do not substitute a single month’s rate change.
4. Check units. Enter growth in the units used in estimation, distinguish percentages from percentage points, and use consistent actual/potential output measures.
5. Calculate and interpret. First compute the fitted unemployment change or gap. Then add it to the earlier rate or natural-rate benchmark, as appropriate.
6. Compare with other evidence. Explain residuals and uncertainty using labor-force participation, employment, hours, productivity and data revisions.

A Quarterly Annualized Rate Is a Different Input

BEA annualizes quarterly GDP growth by compounding the quarter’s growth rate over four quarters.^[3] For a hypothetical 1% quarter-to-quarter increase, the annualized rate is [(1.01)⁴ – 1] × 100 = 4.060401%, or about 4.06%. Only one quarter has been observed; this calculation does not establish the following three quarters’ growth.

Neither 1% quarterly growth nor its 4.06% annualized equivalent can simply be combined with a one-quarter unemployment change using the earlier example’s annual coefficients. Researchers can estimate quarterly models with annualized GDP growth, but the coefficients must be fitted for that convention. Converting the GDP number alone does not convert the model.

The existing GDP Calculator helps with expenditure-based output accounting, while the Unemployment Rate Calculator helps with labor-market rates. They support understanding the underlying measures; neither estimates an Okun regression or supplies potential output.

Common Mistakes with Okun’s Law

1. Treating a two-to-one rule as universal. A quoted ratio needs a specified equation, unit convention, country, period and estimation method. It is not a permanent conversion between output and unemployment.

2. Calling a percentage-point change a percent change. The worked decline from 6.0% to 5.2% is 0.8 percentage point, not a 0.8% relative decline.

3. Assuming GDP must contract for unemployment to rise. In a growth/change model, positive growth below its fitted stable-unemployment threshold can accompany an unemployment increase.

4. Mixing levels, growth and gaps. A -4% output gap means output is 4% below estimated potential. It does not mean output fell 4% during that year.

5. Ignoring frequency and lags. Annual, four-quarter and quarterly annualized GDP growth are distinct measurements. Pair each with the unemployment measure and lag structure used to estimate the equation.

6. Treating estimated benchmarks as observed facts. Potential output and natural unemployment cannot be read directly from a survey. Their estimation uncertainty carries into the gap calculation.

Limitations in Recessions, Recoveries and Forecasts

Timing can differ across the cycle. Output and staffing do not adjust simultaneously. Hours reductions or worker retention can soften the immediate unemployment response during a downturn; using existing workers more intensively can delay hiring during recovery. A same-period equation can miss that path.

Data revisions can change the apparent story. San Francisco Fed research found that later GDP revisions explained part of the apparent departure from the relationship during the Great Recession. Its analysis also highlights temporary cyclical deviations and adjustment lags.^[6] One unusual observation does not by itself prove a permanent structural break.

Benchmarks and economic conditions can change. A gap calculation inherits uncertainty about potential output and natural unemployment. Changes in productive capacity, labor supply or the way businesses adjust labor can make a historical equation a poor guide to a new situation.

Aggregate evidence has a limited reach. A national relationship does not determine which industries or households bear job losses, whether new jobs offer the same hours and wages, or how stock prices will respond. Use it as one check on a broader economic analysis.

Sources and Further Reading

1. Edward S. Knotek II, Federal Reserve Bank of Kansas City, How Useful Is Okun’s Law?, fourth quarter 2007. Difference, gap and dynamic formulations.
2. Laurence Ball, Daniel Leigh and Prakash Loungani, Okun’s Law: Fit at 50?, IMF Working Paper 2013/010, January 14, 2013. Official abstract; findings are the authors’ views.
3. Bureau of Economic Analysis, Why does BEA publish percent changes in quarterly series at annual rates? Quarterly growth and compounding conventions.
4. Bureau of Labor Statistics, Concepts and Definitions (CPS). Labor force, unemployment and participation.
5. Doreen Fagan, Federal Reserve Bank of St. Louis, Understanding Potential GDP and the Output Gap, August 4, 2021.
6. Federal Reserve Bank of San Francisco, Interpreting Deviations from Okun’s Law, April 21, 2014. Lags and GDP revisions.
7. Brent Meyer and Murat Tasci, Federal Reserve Bank of Cleveland, An Unstable Okun’s Law, Not the Best Rule of Thumb, June 7, 2012.
8. Bureau of Labor Statistics, Handbook of Methods: Office of Productivity and Technology—Concepts. Output per hour.
9. Michael T. Owyang, E. Katarina Vermann and Tatevik Sekhposyan, Federal Reserve Bank of St. Louis, Output and Unemployment: How Do They Relate Today?, October 1, 2013. Benchmarks and research comparisons.

Sources accessed November 5, 2026. The numerical examples and chart schedules are original hypothetical calculations, not empirical estimates from these publications.