Okun's law

How tightly are output growth and unemployment linked — and how fast does the economy have to grow just to keep the unemployment rate from rising?

Production requires labor, so falling output should mean fewer jobs. But the mapping is not one-for-one: firms hoard trained workers through downturns, hours and productivity absorb part of every shock, and the labor force itself grows and shrinks with opportunity. Okun’s law summarizes the net result as a linear rule: ΔU = a + b·g, where g is real GDP growth and b — the Okun coefficient — is expected to be negative and well below one in magnitude. The intercept matters too: because productivity and the labor force trend upward, zero growth is not enough to hold unemployment steady; the break-even growth rate is −a/b.

Real GDP growth (quarterly, % at a seasonally adjusted annual rate, as published) and the civilian unemployment rate (monthly, averaged within each quarter and then first-differenced). The sample runs from 1960 through Q1 2026, 265 quarters. The quarter containing the missing October 2025 unemployment release is averaged over its two available months — a documented approximation.

Ordinary least squares of the quarterly change in the unemployment rate on published real GDP growth. The scatter below colors quarters by subperiod; the fitted line is the full-sample estimate. Because growth is quoted at an annual rate while the unemployment change is per quarter, the slope reads as “percentage points of unemployment per quarter, per 1pp of annualized growth.”

The answer to the headline question is no — far from one-for-one. A quarter of growth 1pp (annualized) below the break-even pace has historically come with only about 0.12pp of extra unemployment; sustained for a year, roughly 0.5pp. Labor hoarding, hours cuts, and participation flows absorb most of an output shock before the unemployment rate moves. The break-even growth rate of about 3.0% also means “positive growth” and “healthy labor market” are different claims: an economy growing at 2% has, for most of this sample, been one where unemployment drifts up.

The full-sample fit leaves nearly half the quarter-to-quarter variation unexplained (R² 0.55), so the line is a tendency, not a rule for any single quarter. The 2020 pandemic quarters are extreme outliers that pull on any least-squares line — growth near −30% and +35% SAAR with double-digit unemployment swings — and the coefficient shifts visibly at the tenths level depending on whether such episodes are in the window. The subperiod estimates (-0.06 before 1990, -0.17 after) bound how much the “true” coefficient has moved.

• The coefficient is not structural. It drifts across decades — with demographics, labor-market institutions, and the industry mix — so extrapolating any single estimate far out of sample is unreliable by construction.
• Demand vs. supply shocks. Okun’s law describes demand-driven cycles. Supply shocks (an oil shock, a pandemic reopening) move output and unemployment in patterns the linear rule does not capture, and the relationship says nothing about which shock is under way.
• Association, not mechanism. Output and employment are determined together; the regression describes their historical co-movement and cannot say that a policy that raises measured GDP growth would lower unemployment by the slope.

The calculation lives in apps/web/src/app/research/okuns-law/page.tsx (estimation in okun.ts) and uses lib/econ; every number on this page is recomputed from the committed snapshots on each build. Data series and vintages are listed on the Data page.