The American labor market is, at any moment, the product of forces both deliberate and accidental — central bank decisions made years prior, technological displacements working their way through industries, and the diffuse pressures of demography and global trade. Below are its current vital signs.
Compare any two macroeconomic series across six decades. The interplay between unemployment and its companion variables — interest rates, inflation, participation — reveals the textbook relationships in their natural habitat.
The two series are normalized to share a vertical axis. The shaded bands denote NBER recession periods. The correlation coefficient in the upper-right is computed across the visible window — it measures the linear association, not causation.
Unemployment & GDP growth illustrates Okun's law. Unemployment & CPI traces the Phillips curve, including its 1970s breakdown. Fed Funds & Unemployment shows monetary transmission with a characteristic 12-to-18-month lag.
Each slider below represents a shock to a single variable, holding others at their long-run averages. The calibrated model — built from documented elasticities in the empirical literature — translates these shocks into a steady-state prediction for the unemployment rate, decomposed into its constituent channels.
Apply a shock to a chosen variable and observe its propagation through the next twelve quarters. The forecast combines an autoregressive baseline with the calibrated impulse-response from the structural model. Eighty-percent confidence bands reflect the historical variance of the residuals.
Macroeconomics' two most-cited stylized facts — Okun's law and the Phillips curve — are not laws in the physical sense. They are tendencies, visible in the right slice of data, fragile in others. The charts below let you choose your slice.
A negative slope confirms Okun's intuition: faster growth lowers unemployment. The slope coefficient (β) is roughly 0.4–0.5 in U.S. post-war data.
The relationship is unstable across decades. Use the era selector to see how the curve shifted in the 1970s, flattened in the 1990s, and re-emerged in the 2020s.
A working economist's model is, at its best, a compression of decades of accumulated empirical work. What follows describes the equations powering this instrument — and where they break down.
The simulator decomposes the predicted unemployment rate u into a baseline natural rate (u*) plus contributions from cyclical, structural, and frictional channels. The natural rate, here taken at 4.5 percent, reflects the long-run equilibrium of labor demand and supply in the absence of business-cycle shocks. The Congressional Budget Office estimates it at 4.4 percent for the current decade.
Arthur Okun's 1962 observation, that each percentage point of GDP growth above trend reduces unemployment by roughly half a point, anchors the cyclical block. The implementation uses:
where g* is trend growth (2.0%) and the second term captures monetary transmission with a four-quarter lag. The interest rate elasticity (0.15) is drawn from Romer & Romer (2004) narrative shocks.
Government spending and confidence both feed the cyclical channel via the multiplier:
The fiscal multiplier of roughly 0.5 on output translates to about 0.2 on unemployment at the one-year horizon (Blanchard & Leigh, 2013). Confidence is treated as a leading indicator with weaker pass-through.
Structural unemployment captures persistent mismatches that survive even a hot economy. The model aggregates four sub-channels:
Skills mismatch carries the largest weight, reflecting the Beveridge-curve shifts documented since 2008. Minimum-wage effects are modest in line with the Cengiz et al. (2019) bunching estimator; trade-shock elasticities follow Autor, Dorn, & Hanson (2013).
Frictional unemployment — the unavoidable churn of job-search — responds to benefit generosity, search costs, and geographic mobility. Each is normalized to a baseline of zero, with elasticities sourced from Card, Chetty, & Weber (2007) for benefits and Molloy, Smith, & Wozniak (2011) for mobility.
The inflation-expectations channel works in reverse: rising expectations lower measured unemployment in the short run by reducing real wages, before re-anchoring upward in the long run. The model captures only the short-run effect, consistent with the standard accelerationist formulation.
Baseline forecasts use one of four univariate methods — AR(1), naive persistence, linear trend, or exponential smoothing (Holt). The structural impulse is then added quarter-by-quarter, with a damped propagation profile peaking around the fourth quarter for monetary shocks and the second quarter for fiscal shocks. Confidence bands are constructed from the empirical variance of the historical residuals, inflated by √h for forecast horizon h.
This is a teaching instrument, not a forecasting service. The model is linear and time-invariant where the real economy is neither. It does not handle regime changes (the zero lower bound, structural breaks), nonlinearities (financial crises, hysteresis), or simultaneity (variables that move together for common reasons). The 1970s and the early 2020s both produced episodes where standard relationships broke down conspicuously. Read its outputs the way a navigator reads dead-reckoning estimates: useful for orientation, not for landfall.
The embedded historical series — unemployment, GDP growth, federal funds rate, CPI inflation, and labor-force participation — are annual averages drawn from the Federal Reserve Economic Database (FRED), maintained by the St. Louis Fed. Series identifiers: UNRATE, A191RL1A225NBEA, FEDFUNDS, CPIAUCSL, CIVPART. To refresh with the latest values, supply a free FRED API key in the field above; otherwise, the values current to 2024 are used.
Click a posting to see which majors feed into it, and how the wage trajectory unfolds in the Journey section below. Growth percentages and openings are BLS Employment Projections through 2033; wages are 2023 OEWS medians.
U.S. bachelor's degree production by broad field, 2010 to 2022. The boom in computer science, the secular decline in education, and the steady rise of the health professions are the period's three loudest stories. Source: NCES Digest of Education Statistics, Table 322.10.
A choropleth tile cartogram of the United States, sized to give each state equal visual weight regardless of geographic area — useful for labor-market metrics that don't track with acreage. Click a state for details. Sources: BLS LAUS, BLS OEWS, NCES, Census ACS.
Pick a major. We'll show you where its graduates actually end up working (ACS Field-of-Degree data), how their wages evolve over twenty years (College Scorecard + ACS), and how often they stay in the field they trained for. Toggle compare-mode to overlay a second major.
Shares of bachelor's-degree holders by field who report this occupation as primary employment, ACS 5-year 2018–2022.
Median earnings at 1 / 5 / 10 / 20 years post-degree. Sources: College Scorecard cohort earnings; ACS earnings by field of degree.
A log-log scatter of annual projected job openings against annual U.S. bachelor's-degree production for each broad field. Points above the diagonal: more graduates than jobs (oversupply). Points below: more jobs than graduates (shortage). Hover any point for details.
Refreshes state-level unemployment and labor-force metrics from the public DataUSA API. Works directly from the browser. Toggle on, then click Refresh.
Free registration at api.data.gov/signup. Used to refresh wage-by-major data. Stored only in your browser's local storage.
Free at bls.gov/developers. Some series support browser CORS; many do not. Refreshes fail gracefully back to embedded snapshots.
FRED endpoints don't permit browser CORS, so state-level FRED-sourced unemployment trends remain on their embedded vintage. To refresh, the file would need a backend proxy — out of scope for a single static page.