Beveridge Curve: Vacancies and Jobs
Table of Contents
The Beveridge curve helps explain why employers can struggle to hire while people struggle to find work. It places job vacancies beside unemployment, allowing an analyst to ask whether hiring demand is cooling, whether matching has changed, or whether several forces are moving together. Those distinctions matter when interpreting labor-market news and claims about an economic soft landing.
This guide uses U.S. Bureau of Labor Statistics definitions, separates movements along a curve from shifts, and checks the interpretation against historical examples. The curve organizes evidence; it does not by itself identify a cause, predict inflation or determine an interest-rate decision.
What Is the Beveridge Curve?
A Beveridge curve plots the unemployment rate on the horizontal axis and a vacancy rate on the vertical axis. Each monthly observation pairs the two measures. The usual downward pattern means that stronger hiring demand tends to accompany more openings and less unemployment, while weaker conditions tend to accompany fewer openings and more unemployment. BLS publishes a seasonally adjusted plot using its job openings rate and the civilian unemployment rate. [7]
A movement along a given curve changes the vacancy–unemployment combination while its underlying relationship is held fixed. A shift changes the combinations associated with a given unemployment rate. Observed monthly points can reflect both changes, plus adjustment delays.
Vacancies and unemployment can coexist because a job match takes time. An opening may require a particular skill, location, schedule or compensation agreement. The types of unemployment framework explains these search and mismatch issues; the Beveridge curve examines their aggregate relationship with available jobs.
The Two Rates Use Different Denominators
The headline unemployment rate comes from the Current Population Survey (CPS), a household survey. The job openings rate comes from the Job Openings and Labor Turnover Survey (JOLTS), an establishment survey. Both are percentages, but they are percentages of different things.
The CPS covers the civilian noninstitutional population age 16 and older in the 50 states and District of Columbia. Unemployed people generally have no employment, are available for work and actively searched during the previous four weeks; qualifying temporary layoffs awaiting recall are the search exception. Each employed person is counted once, including a person holding several jobs. [1]
Here, V is job openings and Epayroll is establishment payroll employment. JOLTS covers nonfarm private and civilian government establishments, with exclusions such as farm workers, the unincorporated self-employed and private-household workers. Its jobs measure is different from CPS employed people. [3]
A JOLTS opening must be a specific position with work available, could start within 30 days, and involve active recruiting from outside the establishment. A position reserved only for an internal transfer or recall from layoff does not qualify. An online advertisement is therefore not automatically a separate JOLTS vacancy. [2]
| Measure | Reference period | What is counted? |
|---|---|---|
| CPS unemployment | Survey reference week, usually including the 12th; some holiday exceptions. | People classified as unemployed within the civilian labor force. |
| JOLTS openings | Last business day of the month. | Qualifying unfilled positions. |
| JOLTS employment denominator | Pay period including the 12th. | Covered payroll jobs. |
| JOLTS hires and separations | Entire month. | Payroll additions and departures: flows, rather than end-of-month stocks. |
The differing reference periods are part of the official measures. Pairing observations from the same month does not make them simultaneous snapshots of an identical population. The unemployment rate and labor force guide develops the household-survey boundaries in more detail. [1] [2] [4]
A Worked Example: Rates vs the V/U Ratio
Consider hypothetical survey results for one month. The household survey estimates 152 million employed and 8 million unemployed people. A separate establishment survey estimates 144 million filled payroll jobs and 6 million openings. These inputs describe different survey populations.
Unemployment rate: 8 / (152 + 8) × 100 = 5.0%.
Job openings rate: 6 / (144 + 6) × 100 = 4.0%.
Vacancy-to-unemployment ratio: V/U = 6 / 8 = 0.75 openings per unemployed person.
Dividing the published percentages gives 4.0 / 5.0 = 0.80, which is different from 0.75. Calculate V/U from the levels. Some research instead normalizes vacancies by the labor force; its vacancy percentage must be distinguished from the official JOLTS rate.
Even a V/U ratio above one does not mean each unemployed person has a suitable offer. Employed people also compete for openings, and the aggregate counts contain no assignment of particular workers to particular jobs.
Movements Along the Curve vs Shifts
In a simplified analysis with matching conditions and other relevant factors unchanged, stronger demand encourages vacancy creation and helps unemployed people find work. The economy moves up and left. Weaker demand produces a movement down and right. This is the usual connection to the business cycle, rather than a claim that every month follows the same path. [5]
To isolate the geometry, compare two hypothetical schedules. The baseline is v = 18/u and the outward schedule is v = 24/u, with both rates entered as percentage numbers. These are chosen teaching relationships, not estimates of the U.S. economy. Moving between points on the baseline holds the schedule fixed; comparing the schedules at the same unemployment rate isolates a shift.
| Unemployment rate (%) | Baseline (%) | Outward shift (%) |
|---|---|---|
| 3 | 6 | 8 |
| 4 | 4.5 | 6 |
| 6 | 3 | 4 |
| 8 | 2.25 | 3 |
Hypothetical schedules v = 18/u and v = 24/u; both rates entered as percentage numbers. Sampled points illustrate the relationships, not empirical estimates or forecasts.
On the baseline, moving from 4% unemployment and 4.5% openings to 6% unemployment and 3% openings is a movement down and right. At 4% unemployment, moving from 4.5% to 6% openings instead compares two schedules: the outward shift adds 1.5 percentage points of vacancies at that unemployment rate.
An inward shift reverses that comparison: fewer vacancies coexist with the same unemployment rate, or less unemployment with the same vacancy rate. It can be consistent with better matching, but the graph alone does not establish that explanation. In real data, changing separations and adjustment away from steady state can also move the observed relationship. [6]
How Matching Efficiency and Flows Affect the Curve
Matching efficiency describes how effectively available workers and vacancies generate matches, after accounting for the model’s measured inputs. If fewer matches form from the same inputs, unemployment can remain higher and openings can remain unfilled longer. Skills and geography can matter, but so can worker search effort, employer recruiting intensity and the mix of workers or vacancies. A model’s efficiency estimate can absorb several of these omitted differences. [5] [9]
For example, two employers may each report an actively recruited opening while devoting different amounts of time to interviews or offering different compensation. A vacancy count alone does not measure those efforts. Similarly, an aggregate relationship can change when employment demand moves between sectors whose jobs take different amounts of time to fill.
Unemployment Depends on Inflows as Well as Hiring
To see why matching is only part of the explanation, use a separate, stripped-down model. Hold the labor force fixed and allow only employment and unemployment. Let s be the monthly probability that an employed person enters unemployment, and f the monthly probability that an unemployed person finds employment. In steady state, inflows and outflows balance:
u = s / (s + f)
With hypothetical monthly probabilities s = 0.01 and f = 0.19, unemployment is 0.01 / (0.01 + 0.19) = 0.05, or 5%. Keeping s fixed while f rises to 0.24 gives 0.01 / (0.01 + 0.24) = 0.04, or 4%. Reducing the inflow probability s could also lower unemployment without an improvement in job finding. [5]
These probabilities are not interchangeable with JOLTS totals. Total separations include people moving directly to another job or leaving the labor force. Total hires include workers arriving from other jobs or from outside the labor force. Dividing total JOLTS hires by unemployed people therefore does not give this model’s unemployment-to-employment probability.
During rapid adjustment, unemployment is also not necessarily at its steady-state value. A scatterplot can trace a loop because vacancies and unemployment respond at different speeds. Federal Reserve research identifies matching efficiency, separations and these dynamics as distinct contributors to observed shifts. [6]
U.S. Examples: Recession, Pandemic and Cooling
The following observations come from the BLS seasonally adjusted Beveridge-curve table accessed November 15, 2026. They use the official job openings rate. They are selected historical endpoints from one retrieval vintage, rather than the numbers first published in each original release. [7]
| Month | Unemployment rate | Job openings rate |
|---|---|---|
| December 2007 | 5.0% | 3.2% |
| June 2009 | 9.5% | 1.9% |
| February 2020 | 3.5% | 4.4% |
| April 2020 | 14.8% | 3.4% |
| July 2022 | 3.5% | 7.0% |
| June 2024 | 4.1% | 4.5% |
The Great Recession
Between December 2007 and June 2009, unemployment rose 9.5% − 5.0% = 4.5 percentage points, while the openings rate fell 1.9% − 3.2% = −1.3 percentage points. This down-and-right direction fits weakening labor demand. It does not establish that every underlying determinant remained fixed.
The subsequent recovery prompted debate about whether vacancies were unusually high relative to unemployment. Ahn and Crane’s model-based decomposition found that reduced matching efficiency and adjustment dynamics pushed the post-2010 relationship outward, while changes in separations partly offset them. Calling the entire shift “skills mismatch” would discard that distinction. [6]
The Pandemic and Reopening
The April 2020 combination reflected an extraordinary disruption. Temporary layoffs and later recalls, changes in labor-force participation, and demand moving between in-person services and other sectors complicated ordinary matching. The same workers did not always need to find entirely new employers, and the composition of available jobs changed. These conditions make a simple permanent-efficiency interpretation especially fragile. [8]
By July 2022, unemployment matched February 2020’s 3.5%, but the openings rate was 7.0% − 4.4% = 2.6 percentage points higher. That is a clear change in the observed vacancy–unemployment combination. Identifying its causes requires additional evidence about search, recruiting, separations and composition.
Vacancies also serve different hiring purposes. Research discussed by the St. Louis Fed distinguishes recruiting unemployed workers from recruiting people already employed elsewhere. A greater share of the latter can raise vacancies without a proportional increase in exits from unemployment. This is one model-based explanation to investigate, rather than proof that official vacancies are merely empty advertisements. [10]
Cooling Does Not Have One Predetermined Cost
From July 2022 to June 2024, the openings rate fell 4.5% − 7.0% = −2.5 percentage points, while unemployment rose 4.1% − 3.5% = 0.6 percentage points. Those endpoints show that a large decline in vacancies can coexist with a much smaller unemployment increase. They do not show how much any one policy caused.
Contemporaneous researchers disagreed about the likely path because matching-function shape, matching efficiency and future layoffs were uncertain. A steep portion of a curve can permit a large vertical change with a small horizontal change; an inward shift can also alter the outcome. Neither feature guarantees the next episode will follow the same course. [8] [9]
Beveridge Curve vs Phillips Curve
| Framework | Variables | Main use |
|---|---|---|
| Beveridge curve | Vacancies and unemployment. | Examine hiring demand, matching and labor-market adjustment. |
| Phillips curve | Inflation or wage growth and a measure of economic slack, with expectations and other factors. | Analyze inflation pressure under a specified model. |
| V/U ratio | Openings divided by unemployed people. | Summarize one aspect of labor-market tightness; it is not the full two-variable curve. |
The connection is competition for labor. When vacancies are plentiful relative to available workers, employers may face greater pressure to improve pay. The Phillips curve supplies the separate inflation framework; inflation expectations, productivity and supply conditions affect how labor tightness translates into wage and price changes.
The Beveridge curve contains no price variable and supplies no inflation forecast by itself. Check observed compensation measures such as the Employment Cost Index, rather than assuming a vacancy movement establishes a particular wage outcome. The curve also does not identify the unemployment rate consistent with stable inflation without additional assumptions.
How to Analyze a Beveridge-Curve Update
- Read both axis definitions. Identify official job openings/(employment + openings), vacancies/labor force, or another measure. Do not combine their percentages as if they were the same series.
- Match the data period and seasonal treatment. Pair the same month and consistent adjustment. Record the release date and the vintage actually available when studying a historical decision.
- Describe the observed change first. State whether openings and unemployment rose or fell before labeling a demand movement or structural shift.
- Check flows and composition. Examine layoffs, quits, hires, participation and sector differences. Total hires are broader than job finding from unemployment.
- Compare explanations. Consider matching/search effort, employer recruiting, separations and adjustment delays; avoid assigning a unique cause from two coordinates.
- Keep any projection conditional. Specify the curve, range and held-fixed assumptions. Cross-check compensation and inflation evidence before drawing a policy conclusion.
Common Mistakes
- Using the labor force as the JOLTS denominator. The published openings rate uses covered payroll employment plus openings.
- Dividing the two percentages to calculate V/U. Their denominators differ; use the levels.
- Calling every outward change a loss of matching efficiency. Separations, composition and adjustment dynamics can change the observed relationship too.
- Equating all hires with unemployed people finding jobs. Some hires come from other employers or outside the labor force.
- Ignoring the data vintage. Original releases and revised histories can give different values for the same month.
- Reading a permanent policy menu from a fitted line. The relationship and the economy’s position can change together.
Limitations of the Beveridge Curve
Two vacancy–unemployment coordinates cannot distinguish all changes in matching, labor demand, separations, participation and recruiting behavior. A fitted curve’s interpretation depends on its model and sample.
Aggregation can hide a shortage of workers in one occupation alongside a shortage of suitable jobs elsewhere. National JOLTS does not collect occupational vacancy information, so its aggregate curve cannot directly diagnose a particular skills shortage. Nor does the unemployment axis include every potential job seeker: employed workers and people outside the labor force can enter hiring. [4]
Both axes are survey estimates. Sampling and reporting uncertainty affect the observed points, and JOLTS histories change as additional responses and updated seasonal and employment information arrive. Short-term wiggles deserve less confidence than a sustained pattern supported by other evidence. [3] [4]
Finally, an estimated curve describes the sample used to fit it. Extrapolating beyond that range, or holding layoffs and matching fixed through a major shock, can produce misleadingly precise predictions. Use the plot to organize a labor-market diagnosis and make assumptions explicit.
Frequently Asked Questions
Sources
- BLS: Concepts and Definitions (CPS). Labor-force coverage, unemployment, person counts and reference weeks.
- BLS: JOLTS Data Definitions. Qualifying openings, employment and payroll flows.
- BLS: JOLTS Field Definitions, Rate/Level; Handbook of Methods, Concepts and Design. Rate denominator, current coverage and sample design.
- BLS: JOLTS Frequently Asked Questions. Stocks versus flows, revisions and absence of occupational vacancy data.
- Figura and Waller: What does the Beveridge curve tell us about the likelihood of a soft landing? Federal Reserve FEDS Notes, July 29, 2022. Flow-balance model, matching and separations.
- Ahn and Crane: Dynamic Beveridge Curve Accounting. Federal Reserve, March 2020, public abstract. Model-based decomposition of shifts and Great Recession findings.
- BLS: The Beveridge Curve, seasonally adjusted. Historical data table accessed November 15, 2026; the article uses that retrieval vintage.
- Bok, Petrosky-Nadeau, Valletta and Yilma: Finding a Soft Landing along the Beveridge Curve. San Francisco Fed Economic Letter, August 29, 2022. Model dependence and pandemic reallocation.
- Richter and Throckmorton: Job vacancy, unemployment relationship clouds ‘soft landing’ prospects. Dallas Fed, February 7, 2023. Recruiting/search intensity, parameter uncertainty and conditional projections.
- Restrepo-Echavarría and Bass: Why Job Vacancy Types Matter for Monetary Policy. St. Louis Fed, November 18, 2025. Model-based distinction between hiring from unemployment and recruiting employed workers.
Sources accessed November 15, 2026 UTC. Hypothetical counts, schedules and flow probabilities are teaching examples, not measured U.S. parameters. Historical estimates can be revised.
Disclaimer
This article is for educational and informational purposes only and does not constitute investment, employment or policy advice. Economic estimates and relationships can change. Consult the relevant official releases and qualified professionals when making decisions.
