Real-time research-based analyses of the state of the U.S. labor market

This page contains exhibits with up-to-date analyses of labor markets, especially for the U.S., related to our, Ayşegül Şahin and Bart Hobijn, research. The exhibits are organized in a set of themes: Headline Numbers, Slack Indicators, Principal Components Analysis, Unemployment Insurance (UI), Labor Market Flows, Job Openings and Labor Turnover, Labor Force Participation, Wages and Compensation, Composition and Aggregate Wage Growth, Wage Rigidity Meter , Labor Share, Dual Labor Market, and Seasonality.

All exhibits are downloadable in PDF/PNG/SVG format. For those exhibits that contain calculated data we also provide the underlying data. For others we provide references to the data source. In addition, we provide a set of up-to-date slides for teaching as well as a paper with an overview of this site in the Teaching Materials section.

If you use any of the content from this website for your own analysis, please cite this website as well as the research on which the data that you use is based. The "Related to..." section for each item and the Biblio page provide the relevant references. This site is curated by ourselves. The views expressed here are ours and not necessarily those of institutions we are or have been affiliated with, including the Federal Reserve Bank of Chicago, Federal Reserve System, Princeton University, and NBER. See disclaimer for more details.

This section contains links to the metadata, data (if available), and charts that are available to download from the site. If the data is directly taken from another source, we provide a link to that source. (Metadata)

Main labor market indicators

Much of our research is focused on better understanding the cyclical fluctuations and long-run trends in the main indicators for the U.S. labor market. To set the tone for the rest of the analyses on this webpage, we include the main headline indictors, as well as some analyses directly related to them in this section.

This section contains links to the metadata, data (if available), and charts that are available to download from the site. If the data is directly taken from another source, we provide a link to that source. (Metadata)

Unemployment rate

Share of the labor force that is actively looking for work but is not employed.

Metadata, Data, Chart: (PDF/PNG/SVG)

Unemployment rate for civilian non-institutionalized population of age 16 and older. Monthly observations; seasonally adjusted; share of labor force

Monthly change in nonfarm payroll employment and its 3-month moving average

One of the two main indicators of health of labor market. It is timely, generally released on the first Friday of the month, and is based on a survey of a very large number of establishments.

Metadata, Data, Chart: (PDF/PNG/SVG)

Payroll employment in the nonfarm sector. Monthly observations; seasonally adjusted; month-over-month change; thousands of jobs

Labor force participation rate (LFPR)

Share of the civilian noninstitutional population 16 years and older that either works or is actively looking for a job, i.e. that actively participate in the labor market.

This measure is closely followed because it quantifies the total supply of labor, in terms of persons, measured as a share of the population.

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of the civilian noninstitutional population 16 years and older that either works or is actively looking for a job, i.e. that actively participate in the labor market. Monthly observations; seasonally adjusted; share of civilian noninstitutional population 16 years and older

Employment-Population Ratio (EPOP)

Share of the civilian noninstitutional population 16 years and older that works (either is on payroll, self employed, or works unpaid for family).

This is the broadest measure of employment, in terms of persons, relative to the size of the population.

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of the civilian noninstitutional population 16 years and older that either works or is actively looking for a job, i.e. that actively participate in the labor market. Monthly observations; seasonally adjusted; share of civilian noninstitutional population 16 years and older

Job openings rate

The job openings rate, a.k.a. vacancy rate, measures the strength of labor demand by capturing the share of the demand for labor, proxied by payroll jobs and job openings, that is unfilled.

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of labor demand (job openings plus existing jobs) that is unfilled. Monthly observations; seasonally adjusted; share of labor demand (jobs and job openings)

U.S. Beveridge curve

Graphical representation of the negative relationship between labor demand, i.e. vacancies (job openings), and the unemployment rate, first conjectured by Beveridge (1944). This plot includes data on the job openings rate from January 2000 through now from JOLTS.

The negative relationship between the vacancy rate and unemployment reflects that most of the fluctuations in the unemployment rate are driven by changes in labor demand rather than labor supply.

Metadata, Data, Chart: (PDF/PNG/SVG)

Beveridge curve for January 2000 until now, which is the sample period for which JOLTS data is available Monthly observations; seasonally adjusted

Historical U.S. Beveridge curve

Graphical representation of the negative relationship between labor demand, i.e. vacancies (job openings), and the unemployment rate, first conjectured by Beveridge (1944). Historical data on the unemployment rate and vacancies are from Petrosky-Nadeau and Zhang (2021)

The negative relationship between the vacancy rate and unemployment reflects that most of the fluctuations in the unemployment rate are driven by changes in labor demand rather than labor supply.

Metadata, Data, Chart: (PDF/PNG/SVG)

Beveridge curve for 1919 until now. Monthly observations; seasonally adjusted

Three employment concepts published by the Bureau of Labor Statistics

The headline employment concept is nonfarm payroll jobs, collected as part of the Current Employment Statistics (Establishment Survey). But the BLS also publishes an equivalent measure based on household responses to the Current Population Survey (Household Survey). That measure differs from the third measure, namely the number of persons employed, that is used to calculate the size of the labor force, in three ways. First, it counts persons, e.g. multiple job holders count only once. Second, it includes the farm sector. Finally, it includes the self-employed and unpaid family workers.

Metadata, Data, Chart: (PDF/PNG/SVG)

Unemployment rate. Monthly observations; seasonally adjusted; millions of jobs (CES) or persons (CPS)

Measures of Resource Slack and the Natural Rate of Unemployment

The extent to which labor is underutilized, i.e. there is resource slack in the labor market, is important for the assessment of macroeconomic stabilization policy options.

In this section we provide several measures of slack as well as the neutral level of the unemployment rate that reflects neither slack nor tightness.

This section contains links to the metadata, data (if available), and charts that are available to download from the site. If the data is directly taken from another source, we provide a link to that source. (Metadata)

Stylized narrative of fluctuations in the labor market

We find this slide useful to tie together the many labor market indicators that we study to analyze fluctuations in the labor market. Often, we focus a lot on the numbers but forget the broad narrative that almost all labor market cycles have in common. This slide is a good reminder of this broad narrative.

(Metadata).

Broad measures of unemployment

U3 is the headline unemployment rate, U4 through U6 expand those to increasingly include broader groups as un(der)employed.

Metadata, Data, Chart: (PDF/PNG/SVG)

Rates of underutilization of the labor supply. Monthly observations; seasonally adjusted; share of (extended) labor force

Involuntary part-time employment because of slack economic conditions

Share of the labor force that is working part-time because they are not able to find a job at which they can work more hours. It is a measure of underemployment because it captures the fraction of the labor force that works fewer hours than they would like to.

Note: There is a break in the series in January 1994 due to the redesign of the CPS survey instrument.

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of the labor force that is working part-time because they are not able to find a job at which they can work more hours. Monthly observations; seasonally adjusted; share of labor force

Hornstein-Kudlyak-Lange Non-Employment Index (NEI)

This index is a weighted average of all non-employed individuals, expressed as a share of the civilian non-institutionalized population aged 16 and over. The weights reflect persistent differences in each group’s likelihood of returning to work. By accounting for these variations and including a broader set of non-employed individuals, the NEI arguably offers a more accurate measure of labor market conditions than the standard unemployment rate.

Detailed explanation and source data available at the Federal Reserve Bank of Richmond. Chart: (PDF/PNG/SVG)

Non employment index . Monthly observations; seasonally adjusted; share of population

Noncyclical Rate of Unemployment and actual unemployment rate

The noncyclical rate of unemployment is the rate of unemployment arising from all sources except fluctuations in aggregate demand. It measures the level of the unemployment at which there is no resource slack or tightness due to aggregate-demand fluctuations.

Metadata, Data, Chart: (PDF/PNG/SVG)

Time series of unemployment rate and non-cyclical rate of unemployment and unemployment gap highlighted. Monthly observations; seasonally adjusted; 'natural' is CBO's noncyclical rate

Okun's Law in Gaps

The negative relationship between resource slack in output and in the labor market. This is called Okun's Law because it was first documented by Arthur Okun in Okun (1963). Knotek (2007) is a very useful review of Okun's Law.

Metadata, Data, Chart: (PDF/PNG/SVG)

Okun's Law that relates resource slack in output to resource slack in the labor market. Quarterly observations; 1948 - now

Okun's Law in Changes

The negative relationship between changes in the unemployment rate and GDP growth, that captures that improvements in labor market resource utilization are generally associated with faster growth in economic activity.

This is called Okun's Law because it was first documented by Arthur Okun in Okun (1963). Knotek (2007) is a very useful review of Okun's Law.

Metadata, Data, Chart: (PDF/PNG/SVG)

Okun's Law that relates changes in the unemployment rate to GDP growth. Quarterly observations; 1948 - now

Two measures of u^* based on unified approach

Estimates of the noncyclical rate of unemployment used to be either based on inflation and price data or on detailed demographic labor-market data. In Crump, Eusepi, Giannoni, and Şahin (2019) developed such measures and updated them in Crump, Eusepi, Giannoni, and Şahin (2024). We report these measures here.

ū is the secular trend of unemployment which is estimated using detailed unemployment flows by age and gender.

u^* is defined as the unemployment rate such that, controlling for supply shocks, inflation remains stable. It is estimated using ū, wage and price inflation, and inflation expectations with Bayesian methods.

Note: Because of the Bayesian estimation methods involved in constructing these measures, we update them relatively infrequently compared to other data on this site.

Metadata, Data, Chart: (PDF/PNG/SVG)

Two estimates of the non-cyclical rate of unemployment that combine the use of inflation and detailed demographic labor market data. Quarterly; observations; seasonally adjusted; share of labor force

Slack Dash(board): Visualization of wide range of labor-market indicators

Labor market “slack” is not a unidimensional concept. Though many labor-market indicators have a lot of their cyclical movements in common, it is useful to have a visualization of these comovements and deviations from them. The Slack Dash provides such a visualization at a particular point in time.

The Kineograph (PDF, 19.1 MB) allows for flipping through these snapshots for different months to see the evolution of labor-market “slack” over time from the large menu of indicators included. It also contains an explanation of how to read the Slack Dash.

Metadata, Chart: (PDF/PNG/SVG), Kineograph (PDF, 19.1 MB)

Data is from many different sources and not available for download.

Labor Market Slack Dashboard: 3-m MA for latest available $\leq$ Mar-2026 and 12-m lagged; prctile of historical procyclical time series

Principal Component Analysis (PCA): "Level" and "Momentum" of labor market slack

Labor market indicators tend to move together a lot. Principal Component Analysis (PCA) reveals that the first two components explain about two thirds of the movements in the indicators. These components are often referred to as "level" and "momentum", because they tend to capture the level of slack and the rate of change.

The exhibits in this part present results of a Principal Component Analysis of a dashboard of labor market indicators using data from 1990 onwards. The indicators used are those from the Slack Dash(board).

(Metadata).

First Two Principal Components from labor market indicators: "Level" and "Momentum"

Labor market indicators tend to move together a lot. Principal Component Analysis (PCA) reveals that the first two components explain about two thirds of the movements in the indicators. These components are often referred to as "level" and "momentum", because they tend to capture the level of slack and the rate of change. They are plotted here.

The "Level" and "Momentum" labels go back to at least Daly, Hobijn, and Bradshaw (2013) and have also been adopted in the KC Fed's Labor Market Conditions Indicators (Hakkio and Willis (2014)).

Metadata, Data, Chart: (PDF/PNG/SVG)

First two Principal Components (level/momentum) extracted from wide range of labor market indicators. Monthly observations; seasonally adjusted; First two Principal Components

First two components, "Level" and "Momentum", explain bulk of labor market fluctuations

This shows the share of the variation in the labor market indicators captures by the Level and Momentum components in the PCA. The labor market indicators used are those from the Slack Dash(board).

The "Level" and "Momentum" labels go back to at least Daly, Hobijn, and Bradshaw (2013) and have also been adopted in the KC Fed's Labor Market Conditions Indicators (Hakkio and Willis (2014)).

Metadata, Data, Chart: (PDF/PNG/SVG)

Description

Factor loadings of "Level" and "Momentum" for indicators

The factor loadings show which indicators provide information about the Level and/or Momentum components of labor market slack. The indicators used are those from the Slack Dash(board).

The "Level" and "Momentum" labels go back to at least Daly, Hobijn, and Bradshaw (2013) and have also been adopted in the KC Fed's Labor Market Conditions Indicators (Hakkio and Willis (2014)).

Metadata, Data, Chart: (PDF/PNG/SVG)

Description

Standard deviation of residuals from PCA fit

This shows the monthly cross-indicators standard deviation of the residuals of the regression of the indicators on the Level and Momentum components. The larger this standard deviation the more mixed the signals about labor market slack from the indicators are. The labor market indicators used are those from the Slack Dash(board).

Because not all indicators might be available in the last two months of the sample, the standard deviation can be artifically low in those months.

Metadata, Data, Chart: (PDF/PNG/SVG)

Description

Actual unemployment rate and fitted one implied by first two Principal Components

Labor market indicators tend to move together a lot. Principal Component Analysis (PCA) reveals that the first two components explain about two thirds of the movements in the indicators. These components are often referred to as "level" and "momentum", because they tend to capture the level of slack and the rate of change.

One way to see whether the unemployment rate signals a lot more or less slack than other indicators is to compare it to its level predicted based on these two factors. This is what this figure does. It shows the actual unemployment rate and the one predicted by the two components from the PCA, estimated based on data from 1990 onwards.

Metadata, Data, Chart: (PDF/PNG/SVG)

Actual and fitted unemployment rate based on first two components (level/momentum) of PCA. Monthly observations; seasonally adjusted; percent of labor force; fit based on first two Principal Components

Deviations from signals sent by "Level" and "Momentum" components

This shows the deviation of each indicator from the regression on the first two principal components. It provides an insight into what each indicator signals about labor market slack in addition to these two components. The labor market indicators used are those from the Slack Dash(board).

Z-score is measured in average historical standard deviations of residual of regression of indicator on "Level" and "Momentum" components.

Metadata, Data, Chart: (PDF/PNG/SVG)

Description

Unemployment insurance (UI) related measures

The use of unemployment insurance by workers is an important gauge of the state of the labor market and the underutilization of labor.

In this section we provide several measures related to the use of UI programs.

This section contains links to the metadata, data (if available), and charts that are available to download from the site. If the data is directly taken from another source, we provide a link to that source. (Metadata)

Initial claims for UI and the unemployment rate

Initial claims for unemployment insurance (UI) are often looked at because they are one of the few leading indicators that signal downturns in the labor market. They are published on a weekly basis and are affected by layoffs as well as how fast those laid off find jobs again. If it takes a while for those laid off to find a job then they are more likely to file for UI.

Metadata, Data, Chart: (PDF/PNG/SVG)

Initial claims for UI and the unemployment rate. Weekly and monthly observations; seasonally adjusted

Two layoff rates - JOLTS and UI

The UI layoff rate is the percent of the employed covered for UI insurance that files an initial claim in a month. The JOLTS layoff rate is layoffs and discharges as a percent of nonfarm payroll employment

Note: Layoff rates spiked during the pandemic in 2020. Because this period is not representative of the normal dynamics we reduced the scale on the y-axis to focus more on other recessions. The data for the pandemic are downloadable.

Metadata, Data, Chart: (PDF/PNG/SVG)

JOLTS layoff rate is layoffs divided to nonfarm payroll employment. UI layoff rate is share of the covered employed that filed for initial claims in a month. Monthly observations; seasonally adjusted; share of employed

Take-up rate for UI

The take-up rate is looked at as a leading indictor for labor market downturns. The reason is that when persons covered by UI get laid off and have a harder time finding a job (which is an early sign of a downturn), they tend to file for unemployment and their share of total unemployment increases. This can be considered as the inflows into the UI recipiency rate.

Note: The take-up rate spiked during the Covid pandemic due to the Pandemic Unemployment Assistance (PUA) program. Because this period is not representative of the normal dynamics we reduced the scale on the y-axis to focus more on other recessions. The data for the pandemic are downloadable.

Metadata, Data, Chart: (PDF/PNG/SVG)

Take up rate is estimate of fraction of those laid off and eligible for UI insurance that file for UI. Monthly observations; seasonally adjusted; share of UI eligible layoffs. Index 2020-Jan = 100

Share of unemployed that receive UI benefits

The recipiency rate is looked at as a leading indictor for labor market downturns. The reason is that when persons covered by UI get laid off and have a harder time finding a job (which is an early sign of a downturn), they tend to file for unemployment and their share of total unemployment increases.

Note: The recipiency rate spiked during the Covid pandemic due to the Pandemic Unemployment Assistance (PUA) program. Because this period is not representative of the normal dynamics we reduced the scale on the y-axis to focus more on other recessions. The data for the pandemic are downloadable.

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of the unemployed that are covered by UI and receive continuing claims. Monthly observations; seasonally adjusted; share of unemployed

Flow representation of the labor market

The exhibits in this part provide estimates of stocks and flows for both 2- and 3-state representations of the labor market. In addition, they include flow decompositions of the stocks.

This section contains links to the metadata, data (if available), and charts that are available to download from the site. If the data is directly taken from another source, we provide a link to that source. (Metadata)

Bathtub analogy for unemployment dynamics

Because the unemployment rate is a stock, its dynamics are determined by the size of the flows into and out of unemployment. Two-state approximations account for flows only between employment and unemployment. Three-state approximations also recognize that people move between unemployment and being out of the labor force.

(Metadata).

Unemployment outflow rate from 2-state flow representation

This is the estimated rate at which the employed separate from jobs and flow into unemployment (s), using a 2-state approximation of the labor market.

This a proxy for the job-finding rate. Though flows out of unemployment can also go to non-participation. A high outflow rate out of unemployment puts downward pressure on the unemployment rate.

Metadata, Data, Chart: (PDF/PNG/SVG)

Estimated hazard rate of flows out of unemployment based on unemployment duration distribution data. Monthly observations; seasonally adjusted; percent of unemployed previous month

Unemployment inflow rate from 2-state flow representation

This is the estimated rate at which the employed separate from jobs and flow into unemployment (s), using a 2-state approximation of the labor market.

This a proxy for layoffs. A high inflow rate into unemployment puts upward pressure on the unemployment rate. The inflow rate first increased and has declined since the early 1980s, largely due to the aging of the population and an increase in female labor force participation.

Note: The inflow rate spiked during the Covid pandemic and we rescaled the y-axis to focus on rest of the time series. Covid observations are available with the downloadable data.

Metadata, Data, Chart: (PDF/PNG/SVG)

Estimate of monthly hazard rate for two-state unemployment equation. Monthly observations; seasonally adjusted; percent of employed previous month

Flow steady-state unemployment rate for 2-state flow representation.

In a two state flow representation of the labor market, if the rates at which workers flow between unemployment and employment, i.e. s and f, remain constant, then the unemployment rate will settle at its flow-steady-state of s/(s+f).

We follow the flow steady state, because it captures where the unemployment rate would be heading if current labor market dynamics remain unchanged.

Metadata, Data, Chart: (PDF/PNG/SVG)

Actual and flow steady-state unemployment rate based on two-state decomposition of unemployment flows. Monthly observations; seasonally adjusted; percent of labor force

Inflow-outflow decomposition of 12-month changes in the unemployment rate

This decomposes the 12-month change into parts due to changes in the inflow rate s and outflow rate f as well as an approximation error.

Note: The decomposition works well for small changes in the inflow and outflow rates. But during the Covid-pandemic the changes were very big and the approximation did not work very well. We report the results anyway. We have rescaled the axis, however, to have the figure focus on the episodes where the approximation is most useful.

Metadata, Data, Chart: (PDF/PNG/SVG)

Percentage point contributions of changes in inflow and outflow rates to the 12-month change in the unemployment rate. Monthly observations; seasonally adjusted; percentage point change

Changes in unemployment rate by business cycle decomposed into parts due to inflows and outflows

Cumulative effect on unemployment rate from every trough in the unemployment rate (dashed vertical lines) into parts due to changes in the inflow rate s and outflow rate f as well as an approximation error.

Note: Our decomposition works well for relatively small changes in the flow rates. Because these changes were very large during the Covid pandemic our method is not reliable for those months. For that period we approximate their contribution by their relative impact on the flow steady-state unemployment rate.

Metadata, Data, Chart: (PDF/PNG/SVG)

Percentage point contributions of changes in inflow and outflow rates to trough-to-trough change in the unemployment rate. Monthly observations; seasonally adjusted; percentage point change

Labor force stocks and flows between employment, unemployment, and non-participation

Three-state (employment/unemployment/non-participation) stock-flow representation of the U.S. labor market with the most recent level of the stocks and of the monthly seasonally-adjusted flows.

This figure illustrates the large size of gross worker flows between employment, unemployment, and non-participation. Even when the number of persons in these labor force states change little there are still large offsetting flows that contribute to the dynamics of the U.S. labor market.

Metadata, Data, Chart: (PDF/PNG/SVG)

Stocks of and monthly flows between labor force states in millions of persons.Monthly observations; seasonally adjusted

Employment-to-Unemployment flows

Monthly transition rate from employment to unemployment, which represents the persent of those employed in the last month that are unemployed in the current month.

Note: These flows are margin adjusted using the method from Elsby et al. (2015) such that the estimated transition probabilities map into the evolution of the shares of the population employed (E), unemployed (U), and not participating in the labor market (N).

Metadata, Data, Chart: (PDF/PNG/SVG).

Transition probability from Employment to Unemployment. Monthly observations; seasonally adjusted; margin adjusted; percent of employed in previous month

Unemployment-to-Employment flows

Monthly transition rate from unemployment to employment, which represents the persent of those unemployed in the last month that are employed in the current month.

Note: These flows are margin adjusted using the method from Elsby et al. (2015) such that the estimated transition probabilities map into the evolution of the shares of the population employed (E), unemployed (U), and not participating in the labor market (N).

Metadata, Data, Chart: (PDF/PNG/SVG).

Transition probability from Unemployment to Employment. Monthly observations; seasonally adjusted; margin adjusted; percent of unemployed in previous month

Employment-to-Non-Participation flows

Monthly transition rate from employment to non-participation, which represents the persent of those employed in the last month that are non-participants in the current month.

Note: These flows are margin adjusted using the method from Elsby et al. (2015) such that the estimated transition probabilities map into the evolution of the shares of the population employed (E), unemployed (U), and not participating in the labor market (N).

Metadata, Data, Chart: (PDF/PNG/SVG).

Transition probability from Employment to Non-Participation. Monthly observations; seasonally adjusted; margin adjusted; percent of employed in previous month

Non-Participation-to-Employment flows

Monthly transition rate from non-participation to employment, which represents the persent of those non-participants in the last month that are employed in the current month.

Note: These flows are margin adjusted using the method from Elsby et al. (2015) such that the estimated transition probabilities map into the evolution of the shares of the population employed (E), unemployed (U), and not participating in the labor market (N).

Metadata, Data, Chart: (PDF/PNG/SVG).

Transition probability from Non-Participation to Employment. Monthly observations; seasonally adjusted; margin adjusted; percent of non-participants in previous month

Non-Participation-to-Unemployment flows

Monthly transition rate from non-participation to unemployment, which represents the persent of those non-participants in the last month that are unemployed in the current month.

Note: These flows are margin adjusted using the method from Elsby et al. (2015) such that the estimated transition probabilities map into the evolution of the shares of the population employed (E), unemployed (U), and not participating in the labor market (N).

Metadata, Data, Chart: (PDF/PNG/SVG).

Transition probability from Non-Participation to Unemployment. Monthly observations; seasonally adjusted; margin adjusted; percent of non-participants in previous month

Unemployment-to-Non-Participation flows

Monthly transition rate from unemployment to non-participation, which represents the persent of those unemployed in the last month that are non-participants in the current month.

Note: These flows are margin adjusted using the method from Elsby et al. (2015) such that the estimated transition probabilities map into the evolution of the shares of the population employed (E), unemployed (U), and not participating in the labor market (N).

Metadata, Data, Chart: (PDF/PNG/SVG).

Transition probability from Unemployment to Non-Participation. Monthly observations; seasonally adjusted; margin adjusted; percent of unemployed in previous month

Flow steady-state unemployment rate for 3-state flow representation.

In a three-state flow representation of the labor market, if the rates at which workers flow between employment (E), unemployment (U), and non-participation (N)remain constant, then the unemployment rate will settle at its flow-steady-state.

We follow the flow steady state, because it captures where the unemployment rate would be heading if current labor market dynamics remain unchanged.

Note: The flow steady state in this exhibit is based on the margin-adjusted three-state transition probabilities E → U, U → E, E → N, E → N, N → U, and U → N.

Metadata, Data, Chart: (PDF/PNG/SVG).

Actual unemployment rate and the one implied as a steady state from the 3-state flows approximation. Monthly observations; seasonally adjusted; percent of labor force

12-month change in unemployment rate decomposed into 3-state flow sources

Parts of the 12-month change in the unemployment rate attributable to different 3-state flows. Lines depict contributions of changes in transition probabilities U → E, E → U. , U → N , N → U as well as N → E and E → N together.

The decomposition works well for small changes in the inflow and outflow rates. But during the Covid-pandemic the changes were very big and the approximation did not work very well. We report the results anyway. We have rescaled the axis, however, to have the figure focus on the episodes where the approximation is most useful.

Metadata, Data, Chart: (PDF/PNG/SVG).

Time series chart of components of 12-month change in unemployment rate from 3-state decomposition. Monthly observations; seasonally adjusted; 12-month change; Total

Changes in unemployment rate by business cycle decomposed into flow sources

Cumulative effect on unemployment rate from every trough in the unemployment rate (dashed vertical lines). Lines depict contributions of changes in transition probabilities U → E, E → U. , U → N , N → U as well as N → E and E → N together.

Note: Our decompsition works for relatively small changes in the flow rates. Because these changes were very large during the Covid pandemic our method is not reliable for those months. For that period we approximate their contribution by their impact on the flow steady-state unemployment rate.

Metadata, Data, Chart: (PDF/PNG/SVG).

Spider chart of changes in unemployment rate by recession into parts from three-state flow rates. Monthly observations; seasonally adjusted; cumulative change since unemployment trough; Total

Participation margin accounts for a third of unemployment fluctuations

In Elsby, Hobijn, and Şahin (2015) we used a three-state flow decomposition of changes in the unemployment rate to make the point that the participation margin accounts for about a third of the fluctuations in the unemployment rate. This figure makes that point a bit more concisely than we did in that paper.

This is a differently aggregated version of the decomposition of trough-to-trough change of unemployment rate.

Metadata, Chart: (PDF/PNG/SVG).

Spider chart of changes in unemployment rate by recession into parts from job-loss, job-finding, and the participation margin. Monthly observations; seasonally adjusted; cumulative change since unemployment trough; Total

Flow steady-state EPOP for 3-state flow representation.

In a three-state flow representation of the labor market, if the rates at which workers flow between employment (E), unemployment (U), and non-participation (N) remain constant, then the EPOP ratio will settle at its flow-steady-state.

We follow the flow steady state, because it captures where the EPOP ratio would be heading if current labor market dynamics remain unchanged.

Note: The flow steady state in this exhibit is based on the margin-adjusted three-state transition probabilities E → U, U → E, E → N, E → N, N → U, and U → N. Y-axis is cut off at 40. During Covid the flow steady-state EPOP was below that. Values for the Covid period reported in the data download.

Metadata, Data, Chart: (PDF/PNG/SVG).

Actual EPOP ratio and the one implied as a steady state from the 3-state flows approximation. Monthly observations; seasonally adjusted; percent of labor force

12-month change in EPOP ratio into 3-state flow sources

Parts of the 12-month change in the EPOP ratio attributable to different 3-state flows. Lines depict contributions of changes in transition probabilities U → E, E → U. , U → N , N → U as well as N → E and E → N together.

The decomposition works well for small changes in the inflow and outflow rates. But during the Covid-pandemic the changes were very big and the approximation did not work very well. We report the results anyway. We have rescaled the axis, however, to have the figure focus on the episodes where the approximation is most useful.

Metadata, Data, Chart: (PDF/PNG/SVG).

Time series chart of components of 12-month change in EPOP ratio from 3-state decomposition. Monthly observations; seasonally adjusted; 12-month change; Total

Changes in EPOP by business cycle decomposed into flow sources

Cumulative effect on EPOP ratio from every trough in the unemployment rate (dashed vertical lines). Lines depict contributions of changes in transition probabilities U → E, E → U. , U → N , N → U as well as N → E and E → N together.

Note: Our decomposition works for relatively small changes in the flow rates. Because these changes were very large during the Covid pandemic our method is not reliable for those months. For that period we approximate their contribution by their impact on the flow steady-state EPOP rate.

Metadata, Data, Chart: (PDF/PNG/SVG).

Spider chart of changes in EPOP ratio by recession into parts from three-state flow rates. Monthly observations; seasonally adjusted; cumulative change since unemployment trough; Total

Employment-to-employment flows capture those who go from job to job

Monthly transition rate from one job to another in the Current Population Survey.

A large part of turnover in the labor market consists of workers going from one job to another. There are several estimates of the transition probability of doing so based on the Current Population Survey. This one includes the adjustment for a change in the survey design proposed by Fujita, Moscarini, and Postel-Vinay (2024).

Metadata, Chart: (PDF/PNG/SVG). The data come from the Federal Reserve Bank of Philadelphia.

J2J flows from the Current Population Survey. Monthly observations; seasonally adjusted; percent of employed in previous month

Unemployment in- and outflows by duration crucial for understanding duration distribution

Not everyone who becomes unemployed in a month reports a duration of unemployment shorter than 5 weeks. Many report that they have been looking for a job for several months. Moreover, those who report a longer duration of unemployment are less likely to find a job (duration dependence). To understand the duration distribution of unemployment, and its evolution, it is important to account for both inflows and outflows by duration. This is a point we made in Elsby, Hobijn, Şahin, and Valetta (2011). It was especially relevant for understanding the persistently elevated level of long-term unemployment during the recovery from the Great Recession.

(Metadata).

Duration distribution of unemployment

Unemployment rate split up between those that have been unemployed for: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longer

The duration of unemployment indicates how many people flow into it and how hard it is for those unemployed to find a job. For example, after the Great Recession, that started in 2008, the duration of unemployment remained persistently high, indicating a sluggish labor market during the recovery in which those unemployed had a hard time finding a job.

Metadata, Data, Chart: (PDF/PNG/SVG)

Unemployment rate split up between those that have been unemployed for: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longerMonthly observations; seasonally adjusted; share of the labor force

Unemployment to Employment flows by duration

Share of those unemployed in previous month that are employed in current month by unemployment duration.

U → E rates tends to be lower the longer someone is unemployed. This is known as "Duration dependence" (e.g. Kroft, Lange, Notowidigdo (2013)).

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of those unemployed in previous month employed in current month by duration: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longerMonthly observations; seasonally adjusted; 12-month moving average; share of those unemployed in duration group

Unemployment to Nonparticipation flows by duration

Share of those unemployed in previous month that are out of the labor force in the current month by unemployment duration.

U → N rates tend to be higher those unemployed for a year or longer.

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of those unemployed in previous month not in the labor force in the current month by duration: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longerMonthly observations; seasonally adjusted; 12-month moving average; share of those unemployed in duration group

Outflow rates by unemployment duration

Share of those unemployed in previous month that are not unemployed in current month by unemployment duration.

For each duration group, this is the sum of the outflow rate to employment and that to nonparticipation.

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of those unemployed in previous month not unemployed in current month by duration: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longerMonthly observations; seasonally adjusted; 12-month moving average; share of those unemployed in duration group

Outflow rates by duration and destination

Share of those unemployed in previous month that are not unemployed in current month by unemployment duration and their destination.

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of those unemployed in previous month either employed or not participating in current month by duration: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longer12-month average of seasonally adjusted rates; share of those unemployed in duration group

Duration distribution of unemployment inflows

Unemployment inflows as a percent of the labor force split up between those that have been unemployed for: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 to 51 weeks, 52 weeks and longer

Many of those that become unemployed in a month, either because they lost or quit their job or because they joined the labor force, report a duration of unemployment longer than 4 weeks. This is because the duration of unemployment is based on a question about how long a person has been looking for a job and many CPS respondents consider themselves looking for a job when they have a job, i.e. they are looking for a better job, or when they are out of the labor force but still had the intent to find a job.

Metadata, Data, Chart: (PDF/PNG/SVG)

Inflows into unemployment from employment and non-participation split up into the unemployment duration (months they have been looking for a job) the persons report when they become unemployed: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longerMonthly observations; seasonally adjusted; 12-month moving average; share of the labor force

Inflows into unemployment by duration and origin

Share of labor force that flowed into unemployment by origin for different durations of unemployment.

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of labor force that has flowed into unemployment over last month from either employment or nonparticipation by duration: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longer12-month average of seasonally adjusted rates; share of labor force

Unemployment rate by reason

Unemployment rate split up into why persons report to be unemployed: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.

Certain reasons are more prevalent at different stages of the labor market cycle and the composition of the unemployed by reason during downturns is indicative of the speed of the recovery of the labor market that follows (e.g. Groshen and Potter (2003)).

Note: There is a break in the series in January 1994 due to the redesign of the CPS survey instrument.

Metadata, Data, Chart: (PDF/PNG/SVG)

Unemployment rate split up into reason for unemployment. Monthly observations; seasonally adjusted; share of the labor force

Unemployment to Employment flows by reason for unemployment

Share of those unemployed in previous month that are employed in current month by unemployment reason: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.

Before the CPS redesign in 1994, permanent layoffs and completions of temporary jobs were in the same reason category

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of those unemployed in previous month employed in current month by reason: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.Monthly observations; seasonally adjusted; 12-month moving average; share of those unemployed by reason

Unemployment to Nonparticipation flows by reason

Share of those unemployed in previous month that are out of the labor force in the current month by unemployment reason: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.

Before the CPS redesign in 1994, permanent layoffs and completions of temporary jobs were in the same reason category

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of those unemployed in previous month not in the labor force in the current month by reason: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.Monthly observations; seasonally adjusted; 12-month moving average; share of those unemployed by reason

Outflow rates by unemployment reason

Share of those unemployed in previous month that are not unemployed in current month by reason for unemployment: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.

For each duration group, this is the sum of the outflow rate to employment and that to nonparticipation.

Before the CPS redesign in 1994, permanent layoffs and completions of temporary jobs were in the same reason category

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of those unemployed in previous month not unemployed in current month by duration: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longerMonthly observations; seasonally adjusted; 12-month moving average; share of those unemployed by reason

Outflow rates by reason and destination

Share of those unemployed in previous month that are not unemployed in current month by unemployment reason and their destination.

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of those unemployed in previous month either employed or not participating in current month by duration: Less than 5 weeks , 5 to 14 weeks , 15 to 26 weeks , 27 weeks and longer12-month average of seasonally adjusted rates; share of those unemployed by reason

Distribution of unemployment inflows by reason

Unemployment inflows as a percent of the labor force split up between the reported reason for unemployment: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.

Before the CPS redesign in 1994, permanent layoffs and completions of temporary jobs were in the same reason category

Metadata, Data, Chart: (PDF/PNG/SVG)

Inflows into unemployment from employment and non-participation split up by reason: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.Monthly observations; seasonally adjusted; 12-month moving average; share of the labor force

Inflows into unemployment by reason and origin

Share of labor force that flowed into unemployment by origin for different durations of unemployment.

Metadata, Data, Chart: (PDF/PNG/SVG)

Fraction of labor force that has flowed into unemployment over last month from either employment or nonparticipation by reason: Because they are on temporary layoff, got permanently laid off, had a temporary job that ended, left their job voluntarily, rejoined the labor force, or started looking for work for the first time in their life.12-month average of seasonally adjusted rates; share of labor force

JOLTS release provides measures of labor turnover

As an alternative to worker flows, the JOLTS data release contains information about turnover of jobs at businesses. This includes measures of hires, layoffs, and quits. It also, of course, includes the job openings rate.

(Metadata).

Quits rate

Number of quits, a voluntary separation of an employee from an establishment, in nonfarm sector as a share of payroll employment.

Metadata, Data, Chart: (PDF/PNG/SVG)

Quits rate. Monthly observations; seasonally adjusted; share of nonfarm payroll employment

Separations from payroll employment by reason

Share of payroll jobs that end in each month, either because a worker quits, gets laid off, or for other reasons. The reason for the separation is reported by the employer.

Metadata, Data, Chart: (PDF/PNG/SVG)

Separations rate split into involuntary (layoffs), voluntary (quits), and otherMonthly observations; seasonally adjusted; share of nonfarm payroll employment

Hires rate

Number of hires in nonfarm sector as a share of payroll employment.

Metadata, Data, Chart: (PDF/PNG/SVG)

HIres rate. Monthly observations; seasonally adjusted; share of nonfarm payroll employment

Job openings per unemployed (V/U-ratio)

This ratio is often considered as the main proxy for the tightness of the labor market.

Several studies have argued that this ratio being equal to one is a specifically useful benchmark (e.g. Benigno and Eggertsson (2023)) or even signals labor market efficiency (e.g. Michaillat and Saez (2024)). But one has to be careful because the number of job openings in JOLTS is not a measure of the level of unmet labor demand but rather a proxy that comoves with the level. So, the V/U-ratio measures fluctuations in labor market tightness but not its level.

Metadata, Data, Chart: (PDF/PNG/SVG)

V/U-ratio. Monthly observations; seasonally adjusted; job openings per unemployed person

U.S. matching function

Graphical representation of the relationship between how easy it is for firms to find workers, measured by the number of hires per vacancy on the y-axis, and how tight the labor market is, measured by the number of vacancies per unemployed (V-U ratio).

Shifts in this relationship are often interpreted as changes in match efficiency.

Metadata, Data, Chart: (PDF/PNG/SVG)

Matching function captures how easy it is for firms to hire workers as a function of the tightness of the labor market for January 2000 until now, which is the sample period for which JOLTS data is available Monthly observations; seasonally adjusted

Mismatch Between Unemployed and Vacancies

Index that measures the difference between the cross-industry distributions of unemployment and vacancies.

A large difference between these distributions indicates lower match efficiency in the labor market because of skill mismatch between the labor supplied and that demanded.

Metadata, Data, Chart: (PDF/PNG/SVG)

Mismatch index between industry composition of vacancies and unemployedMonthly observations; seasonally adjusted; 3m-MA; index (Dec-2007=100)

Counterfactual job openings rates

Vacancies are either posted to hire someone for a position that became available when another employee left or to expand payrolls. How many vacancies are posted depends on how hard it is to hire workers, i.e. the job filling rate. This figure illustrates that the main driver of fluctuations in the job openings (vacancy) rate is movements in the job filling rate, not fluctuations in worker turnover or net hiring. It does so by showing that the counterfactual job openings rate for which the job filling rate is held at its historical average moves very little compared to other counterfactuals.

Metadata, Data, Chart: (PDF/PNG/SVG).

Illustration of determinants of job openings rate Monthly observations; seasonally adjusted; share of available jobs

Most vacancies are posted to fill existing positions not to expand payrolls

This figure shows vacancies inflows, a measure of new vacancies added over the month, and separations, i.e. layoffs, quits, and retirements. These two time series move closely together, suggesting that most vacancies are posted to accommodate turnover in the labor market, not net hiring.

Vacancy inflows are measured using Davis et al. (2013) using a simplification proposed by Rob Shimer .

Metadata, Data, Chart: (PDF/PNG/SVG).

Vacancy inflows compared to separations Monthly observations; seasonally adjusted; millions of jobs/job openings

Workers' willingness to quit induces vacancy chains

The cyclical fluctuations in vacancy inflows reflect those of quits. The causal relationship between these two series is ambiguous. On the one hand, a decline in labor demand,reflected by declining vacancies, reduces job opportunities for workers and induces quits to go down. On the other, a decline in workers’s willingness to quit reduced the need for replacement hiring. The distinction between these two explanations is important because they have very different implications for the impact of a decline in vacancies on the path of the unemployment rate.

Vacancy inflows are measured using Davis et al. (2013) using a simplification proposed by Rob Shimer .

Metadata, Data, Chart: (PDF/PNG/SVG).

Vacancy inflows compared to quits Monthly observations; seasonally adjusted; millions of jobs/job openings

Historical labor turnover data available for 1930-1981

The JOLTS release has been published since December 2000. So, there is no data available on job openings, hires, quits, layoffs, and other separations before that date. However, the BLS published monthly data on labor turnover, including hires, separations, quits, and layoffs, as well as accessions from 1930 through 1981 for a sample of goods producing sectors. The above shows the time series for the historical quits rate in the manufacturing sector.

Metadata, Data, Chart: (PDF/PNG/SVG)

Quits rate . Monthly observations; seasonally adjusted; share of payroll employment

Cyclicality of The Labor Force Participation Rate

Exhibits related to the cyclicality of the labor force participation rate and its importance for movements in the unemployment rate.

This section contains links to the metadata, data (if available), and charts that are available to download from the site. If the data is directly taken from another source, we provide a link to that source. (Metadata)

Estimates of trend labor force participation vary a lot

The actual LFPR is the blue line. The trend estimates by source: ● from various CBO trend estimates, ■ from various Tealbook estimates (backward-looking), ★ from Aaronson et al. (2014), and ▲ from Aaronson et al. (2006), Aaronson, Davis, and Hu (2012), Van Zandweghe (2012), Aaronson et al. (2014), and Hornstein, Kudlyak, and Schweinert (2018).

We made this chart to illustrate that the common method of estimating the cyclical component of labor force participation as the deviation from the trend participation rate is subject to a lot of uncertainty. This is because the estimated trend participation rates vary a lot across sources and over time.

Metadata, Data, Chart: (PDF/PNG/SVG)

Labor force participation rate with projections from several academic studies, Tealbook, and CBOMonthly observations; seasonally adjusted; percent of the population

Flow origins of the Participation Cycle and the Attachment Wedge.

In Hobijn and Şahin (2022) we made the point that a large part of the cyclicality of the participation rate is driven by flows between unemployment and employment (even though they do not cross the participation margin). This is because those who become unemployed are more likely to subsequently drop out of the labor force. The difference between the likelihood of an unemployed person leaving the labor force and that of and employed person is what we call the Attachment Wedge. So, if the unemployment rate goes up then this puts downward pressure on the participation rate with a delay.

Note: The transition probabilities, in bold in this figure are the margin-adjusted three-state transition probabilities E → U, U → E, E → N, E → N, N → U, and U → N.

Metadata, Data, Chart: (PDF/PNG/SVG).

Stocks and flows (transition probabilities) organized to illustrate that those unemployed are more likely to leave the labor force. Monthly observations; seasonally adjusted; percent of population (stock) and of origin (flow)

Contribution of Participation Cycle to 12-month change in LFPR

Percentage point contribution of the Participation Cycle to the 12-month change in the participation rate and the actual 12-month change.

It is hard to quantify the cyclical component of movements in the participation rate. One way to do this is to isolate the Participation Cycle, i.e. the part of the fluctuations in the LFPR that is due to changes in the E → U and U → E flow rates. These are the rates within the labor force that drive the bulk of fluctuations in the unemployment rate and affect the LFPR indirectly because the unemployed are more likely to exit the labor force than the employed.

Note: Our methodology accurately quantifies the Participation Cycle for relatively small changes in the flow rates. Because these changes were very large during the Covid pandemic and our method is not reliable for those months, we do not report the results for that period.

Metadata, Data, Chart: (PDF/PNG/SVG).

Twelve-month change in the LFPR, actual and its participation cycle component. Monthly observations; seasonally adjusted; 12-month percentage point change; Total

Changes in LFPR by business cycle decomposed into flow sources

Cumulative effect on LFPR from every trough in the unemployment rate (dashed vertical lines). Entry is contribution from flows between N → U and N → E, exit is contribution from flows between U → N and E → N , and cycle from flows between U and E, i.e. U → E and E → U.

Note: Our methodology accurately quantifies the Participation Cycle for relatively small changes in the flow rates. Because these changes were very large during the Covid pandemic our method is not reliable for those months. For that period we approximate their contribution by their impact on the flow steady-state LFPR. This is why, for that period, the decomposition does not add up to the change in LFPR.

Metadata, Data, Chart: (PDF/PNG/SVG).

Spider chart of changes in LFPR by recession. Monthly observations; seasonally adjusted; cumulative change since unemployment trough; Total

Cyclical pressures on EPOP Ratio by cycle

Cumulative contribution to the change in the EPOP ratio coming from the unemployment and participation rates. The unemployment cycle is defined as the impact of the change in the unemployment rate relative to its trough to the change in the EPOP ratio. The participation cycle is the contribution of the cyclical pressures on the LFPR coming from E → U and U → E.

The Participation Cycle is well measured for small changes in the inflow and outflow rates. But during the Covid-pandemic the changes were very big and the approximation did not work very well. For that period we report results for changes in the flow steady state between the beginning of Covid and the spring of 2021.

Metadata, Data, Chart: (PDF/PNG/SVG).

Spider chart of changes in EPOP rate coming from unemployment rate and from participation cycle. Monthly observations; seasonally adjusted; contribution to change in EPOP since unemployment trough; Total

Fluctuations in Wages and Compensation

Exhibits related to the movement of wages and compensation over the business cycle.

This section contains links to the metadata, data (if available), and charts that are available to download from the site. If the data is directly taken from another source, we provide a link to that source. (Metadata)

Four main measures of wage/compensation growth

Four main measures of wage growth for the U.S.: Average hourly earnings (AHE), compensation per hour (CPH), Employment Compensation Index (ECI), and Median Usual Weekly Earnings (MWE).

There are many different ways to measure how much workers get paid, either per hour or week. As a consequence, there are many different measures of wage and compensation growth. This exhibit shows the four most commonly analyzed in real time for the U.S. economy as well as their average.

Metadata, Data, Chart: (PDF/PNG/SVG)

Four main measures of wage growth for the U.S.: Average hourly earnings (AHE), compensation per hour (CPH), Employment Compensation Index (ECI), and Median Usual Weekly Earnings (MWE).Monthly and quarterly observations; 12-month and 4-quarter growth rates

U.S. Wage Phillips curve

The Wage Phillips curve depicts the negative relationship between the unemployment rate and nominal wage growth.

This is an update of the relationship that Phillips (1958) documented for the U.K. and that inspired the extensive literature on the Phillips curve and the output-inflation tradeoff for macroeconomic stabilization policies that affect aggregate demand.

Note: The figure is plotted for 1986 onwards to focus on a period where inflation is relatively low. The wage growth measure on the y-axis is the average of the four main measures of wage/compensation growth.

Metadata, Data, Chart: (PDF/PNG/SVG)

Wage Phillips Curve - unemployment rate versus average wage growth across four main measures. Monthly observations; seasonally adjusted; Average 12-month wage growth across AHE,CPH,ECI,MWE

Median hourly wage growth by job movement

Median hourly wage growth for all workers and for those who stayed in the same job or switched. Job switchers, in a sense, are the marginal workers in the labor market and their wage growth is thus a useful indicator of overall wage pressures.

Metadata, Data, Chart: (PDF/PNG/SVG)

Median growth rate of hourly wages for all workers and split up by job switchers and job stayers. Monthly observations; seasonally adjusted; 3-month moving average; percent change from a year ago

Evolution of composition of compensation

A lot of emphasis is on wages and salaries paid to workers. But supplements to wages and salaries make up a much larger part of compensation now than in the 1950's. The above figure splits these supplements up into their main components. Most notable is the increased share of compensation paid by employers for health insurance. Payroll taxes for OASDI have also increased as a share of compensation.

Metadata, Data, Chart: (PDF/PNG/SVG)

Composition of compensation paid to workers split up into various categories. Annual observations; share of total compensation paid to workers

Wage Rigidity Meter and Downward Nominal Wage Rigidities (DNWR)

Nominal wages decline much less frequently than that they increase. In fact, nominal wage changes across different datasets, countries, and over time, show a spike at zero. This is generally considered as indicative of Downward Nominal Wage Rigidities. This section provides evidence on the prevalence of such zero wage changes over a year and the distribution of nominal wage changes.

(Metadata).

Spike at zero in distribution of nominal wage changes

The distribution of 12-month nominal wage changes shows a pronounced spike at zero, indicating a substantial share of workers experience no change in their nominal wages over the year. This discontinuity is a hallmark of downward nominal wage rigidity (DNWR)—the tendency for wages to resist cuts even in the face of economic slack. While positive wage changes are spread across a wide range of values, the sharp mass at zero suggests that employers often opt to freeze wages rather than reduce them. Such rigidity can hinder labor market adjustments, especially during downturns, by limiting firms’ ability to lower labor costs without layoffs. The presence of DNWR can also contribute to inflation persistence, as wage freezes delay downward pressure on prices (e.g. Akerlof, Dickens, and Perry (1996)). Though this is a common interpretation of the spike at zero, it is not uncontroversial (e.g. Elsby and Solon (2019)).

Metadata, Data, Chart: (PDF/PNG/SVG)

Distribution of nominal wage changes12-month moving average; 12-month percent change in nominal wage for those in same job as year ago

Share of workers with zero nominal wage changes over past 12-months

When labor-market conditions deteriorate a larger fraction of workers reports to have the same hourly wage as a year ago. This is consistent with downward nominal wage rigidities being more binding during downturns than when the labor market is tight. The figure above plots an estimate of the fraction of workers that are in the same job as a year ago who report the same hourly wage as a year ago as well.

Metadata, Data, Chart: (PDF/PNG/SVG)

Prevalence of zero wage changes in CPS over past 12 months for those imputed to be in the same jobMonthly observations; 12-month moving average; share of those in same job as year ago with same hourly wage

Composition Effect and Aggregate Wage Growth

Workers with lower wages experience larger fluctuations in employment and hours over the business cycle than higher-wage workers. During downturns, this shifts the composition of employment toward higher-wage workers, putting upward pressure on aggregate wages — whether measured by the mean or by a given percentile of the wage distribution. In strong labor markets, the reverse occurs, exerting downward pressure on aggregate wages. These shifts in workforce composition dampen the procyclicality of aggregate wage growth, a phenomenon known as the composition effect (Stockman, 1983, Bils, 1985, Solon, Barsky, and Parker, 1994).

Numerous studies have examined this effect. Here, we present real-time evidence based on data from the Current Population Survey.

(Metadata).

Median Usual Weekly Earnings Growth: All data and MIS4 to MISH8

Median Usual Weekly Earnings (MUWE), as published by the BLS, compares the median from the cross section of CPS full-time employed respondents at the end of one quarter with that at the end of the quarter before. The "All data" is the monthly equivalent of this calculated using the CPS microdata.

For our analysis, however, we compare the wages of respondents in the sample in a month (in CPS MISH8) and 12 months before (in CPS MISH4). We use a three-month rolling sample of this. The resulting wage growth measure is "Matched (MIS8 with MIS4 a year ago)". This is the wage-growth measure that we decompose.

The two measures differ due to: (i) sampling error, and (ii) sample attrition in the CPS between MISH4 and MISH8 (respondents drop out, for example because they move).

The decomposition that follows is based on Daly, Hobijn, Wiles (2011). To be more comparable with other exhibits on this site, we decompose nominal wage growth here, instead of real wage growth that was the focus of the paper.

Metadata, Data, Chart: (PDF/PNG/SVG)

MUWE 12-month growth based on MISH8 ORGs for last three months. Monthly observations; non-seasonally adjusted; 12-month moving average of 12-month percent change

MUWE growth: Wage growth and composition effects

This decomposes the Matched (MIS8 with MIS4 a year ago) into two parts:

Wage growth effect: Part of aggregate wage growth that can be attributed to the growth of wage of those who were full-time employed both at the beginning and end of the year.

Composition effect: Part of aggregate wage growth due to persons flowing into and out of full-time employment, to and from either part-time / self-employment, unemployment, or non-participation. Because inflows into full-time employment tend to happen at lower wages than outflows these flows generally pull down aggregate wage growth, i.e. the composition effect tends to be negative.

The decomposition that is based on Daly, Hobijn, Wiles (2011). To be more comparable with other exhibits on this site, we decompose nominal wage growth here, instead of real wage growth that was the focus of the paper.

Metadata, Data, Chart: (PDF/PNG/SVG)

MUWE 12-month growth based on MISH8 ORGs for last three months. Monthly observations; non-seasonally adjusted; 12-month moving average of 12-month percent change

Composition of Composition Effect

This shows the origins of the composition effect in terms of the origins and destinations of the in- and outflows into and out of full-time employment. The composition effect is countercyclical. It reduces the procyclicality of nominal wage growth. Flows between full-time and part-time/self-employment contribute the most to this. Flows into and out of unemployment are not that import for aggregate wage growth.

The decomposition that is based on Daly, Hobijn, Wiles (2011). To be more comparable with other exhibits on this site, we decompose nominal wage growth here, instead of real wage growth that was the focus of the paper.

Metadata, Data, Chart: (PDF/PNG/SVG)

Flow source of composition effect on MUWE 12-month growth based on MISH8 ORGs for last three months. Monthly observations; non-seasonally adjusted; 12-month moving average of 12-month percent change

Labor share of value added

The share of the value of goods and services produced that is paid to workers is known as the labor share. It is an important measure of the distribution of income across production factors and of the elasticity of output with respect to labor inputs. We studied the labor share in the U.S. in Elsby, Hobijn, and Şahin (2013). The following exhibits are based on that paper.

(Metadata).

Payroll, self-employment, and labor shares in nonfarm business sector

After having been relatively constant from 1950 through 1990, the labor share for the nonfarm business (NFB) sector has declined since. The published labor share is the sum of two parts: Compensation paid to workers on payroll and the imputed labor income of sole proprietors in the NFB sector. The latter is calculated assuming that sole proprietors earn the same compensation per hour as those who are on payroll. Both of these parts have contributed to the decline of the labor share.

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of nonfarm business GVA that goes to worker compensation, and imputed to go to proprietors. Labor share is the sum of these two.Quarterly observations; seasonally adjusted; nonfarm business sector; share of gross value added

Composition of nonfarm business value added

One problem with measuring the labor share is that some forms of income and payments by businesses are not clearly attributable to capital or labor. Compensation is clearly paid to labor and rental income, interest, and depreciation are paid to capital. One could also interpret profits as implicit payments to capital. However, to which production factor taxes and proprietors income are paid is ambiguous. So, part of the calculation of the labor share involves deciding on a way to split these ambiguous income sources between labor and other factors. This figure illustrates that this choice turns out to be important because these two categories make up a substantial share of gross value added of the nonfarm business sector. It includes the published labor share for reference.

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of nonfarm business GVA paid to different factors, owners, and government as well as published labor share.Quarterly observations; seasonally adjusted; nonfarm business sector; share of gross value added

Four measures of the NFB sector labor share

The answer to the question: "What share of income generated in the nonfarm business sector is paid to labor?" is not as clearcut as it might seem. This is because for several categories of national income generated in this sector, like proprietors' income, taxes, and profits, for example, it is unclear which production factor they should be attributed to. So, there are many different ways to calculate the labor share. This chart includes four of them: (i) labor basis assumes compensation per hour for the self-employed is the same as for those on payroll, (ii) all to labor assigns all proprietors' income to labor, (iii) economy-wide distributes proprietors' income across capital and labor at the same rate as the rest of the economy, and (iv) asset basis is the NFB labor share taken from the BLS Multifactor Productivity data. The latter also distributes taxes, profits, and other ambiguous sources of income across capital and labor. No matter which measure you look at, the labor share has declined substantially since the beginning of the 1990's

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of nonfarm business GVA paid to labor for different ways of dealing with ambiguous income components, especially proprietors income.Quarterly observations; seasonally adjusted; nonfarm business sector; share of gross value added

Income inequality and the labor share

The decline of the labor share is often considered as a measure of the increase in income inequality because it indicates workers are getting a "smaller piece of the pie". But that is not necessarily the right interpretation, because the imputation of the labor share also includes some proprietors income. A better way of looking at this is to look at the distribution of both compensation as well as proprietors' income. This figure shows these distributions and illustrates that "the share of income accounted for by both payroll wages and salaries and by proprietors’ income has been buoyed up since the 1980s by substantial rises in the shares accounted for by the very top fractiles of households in the United States" (Elsby, Hobijn, and Şahin (2013)) .

Metadata, Data, Chart: (PDF/PNG/SVG)

Share of nonfarm business GVA paid as compensation and proprietors income by percentiles of the income distribution.Quarterly observations; seasonally adjusted; nonfarm business sector; share of gross value added

Evolution of cross-industry distribution of payroll shares

There is a large variation in the level of payroll shares across industries. This figure plots this variation over time.

Industry classifications have changed over time. This figure uses SIC-72 for 1948-1986, SIC-87 for 1987-1997, and 2017 NAICS for 1998-.

Metadata, Data, Chart: (PDF/PNG/SVG)

Distribution of payroll shares across industries (Value-added weighted).Annual observations; nonfarm business sector; VA weighted distribution; shading is 10th, 25th, 75th, and 90th percentiles

Shift-share analysis of the change in the NFB payroll share since 1948

When we look at the industry-composition of economic activity, the aggregate payroll share changes for two possible reasons. First, because the payroll shares at the industry level change. Second, because economic activity, in terms of nominal value added, changes towards industries with different payroll shares. These two reasons can be quantified using a classic shift-share analysis. This is what we did in Elsby, Hobijn, and Şahin (2013). The above figure shows the main result.

The relative constancy of the aggregate payroll share from 1948 until the 1980's was the result of two offseting effects. Payroll shares at the industry level were rising while the the structure of economic activity was changing in favor of sectors with lower payroll shares. This aligns with Keynes’s (1939, p. 49) characterization of the “remarkable constancy” of the aggregate labor share as “a bit of a miracle”. The decline of the payroll share since the 1980's reflects that these two effects have not approximately offset each other over recent decades. Payroll shares at the industry level have started to decline (declining shift component), while economic activity has not moved towards sectors with higher payroll shares (constant share component).

Industry classifications have changed over time. This figure uses SIC-72 for 1948-1986, SIC-87 for 1987-1997, and 2017 NAICS for 1998-.

Metadata, Data, Chart: (PDF/PNG/SVG)

Cumulative change in the NFB sector payroll share since 1948 split into part due to changes in labor shares at the industry level (shift) and changes in the composition of nominal value added across industries (share).Annual observations; nonfarm business sector; cumulative pp change since 1948

Tech and banking sectors' contribution to change in payroll share

Bonuses are not directly reported in the NIPA data. But two sectors where bonuses make up a large part of pay are the tech and banking sectors. This figure shows the importance of fluctuations in the payroll shares in these two sectors for changes in the overall payroll share. It suggests that during the end of the 1990's and from 2006-2008, bonuses in these sectors played an outsized role for fluctuations in the payroll share.

Metadata, Data, Chart: (PDF/PNG/SVG)

Change in payroll share and shift contributions of the tech and banking sectors.Annual observations; nonfarm business sector; percentage point change in payroll share

TFP growth measured with actual and constant labor share

The labor share determines the weight of labor versus capital growth in the measurement of Total Factor Productivity (TFP) growth. So, one might think that the trend in the labor share has had a substantial impact on TFP growth. But this is not the case. The trend in the labor share is of second order for TFP growth measurement. To make this point, we constructed the figure above which shows TFP (for the business sector) calculated with the actual labor share and, instead, with a constant labor share set to its average over the sample period. The figure shows that these two time series are almost identical and that the difference between them is well within the margin of error with which other variables used to calculated TFP growth are measured.

Metadata, Data, Chart: (PDF/PNG/SVG)

TFP growth calculated with the actual time-varying labor share and with a constant labor share equal to the sample average.Quarterly observations; seasonally adjusted; business sector; annualized percent change

Dual Labor Market (DLM) rediscovered and uncovered

Part of our research agenda is related to the Dual Labor Market in the U.S. with an additional tertiary sector that covers persons who mainly do home production. We have used machine learning techniques to classify respondents in the CPS into these three DLM segments and our current research analyzes the reasons for segmentation and the existence of the DLM.

We will make exhibits and data available upon publication of the related research. (Metadata)

Great Seasonal Moderation in the U.S. labor market

Our current NSF-supported research focuses on the broad decline in seasonality in the U.S. labor market and its implications for the level and cyclicality of the headline numbers.

We will make exhibits and data available upon completion of the related research. (Metadata)

Slides and paper as teaching materials

We know it is hard to maintain a set of slides with the most recent data to teach about the labor market. We have both been there and worked on keeping our teaching materials up-to-date. To facilitate teaching about the macroeconomics of the U.S. labor market this page contains a link to the working paper that describes this website and provides the narrative of a typical labor market cycle. It also includes a PDF file with a slide deck that can be used to teach about the current state of the U.S. labor market.

Slides for teaching about the current state of the labor market.

Paper that explains how to use the exhibits on this webpage to assess the current state of the labor market.

(Metadata).

Presentations about current labor market conditions

• March 25, 2026: Forecasters Club of New York

Focused on facts: We provide evidence, not policy interpretation

Evidence, broadly defined, is information that supports the formation of conclusions. In January 2019, the President signed into law the Foundations for Evidence-Based Policymaking Act of 2018, a measure that passed with strong bipartisan support. This website is created in the spirit of that legislation.

Our goal is to provide evidence that contributes to discussions about labor market developments—among academics, researchers, analysts, and policymakers. All figures and indicators are based on the most recent available vintage of the underlying data. We do not offer policy conclusions or interpretations here.

Sometimes the evidence may disappoint, as during the sluggish labor market recovery following the Great Recession. At other times, it may exceed expectations—such as during the rapid rebound after the Covid-19 pandemic.

(Metadata).

Handling the missing October 2025 data

The October 2025 releases from several federal statistical agencies were never produced due to the extended government shutdown. Because the underlying surveys were not fielded, these observations do not exist and will not be released in the future. As a result, any time series that normally rely on October 2025 data now contain a permanent gap.

Our approach to handling this missing observation depends on the needs of each exhibit and the specific statistical properties required for the analysis. In most cases, we use a simple, transparent interpolation that preserves the overall trajectory of the series without introducing additional modeling assumptions. For data for which the source data is non-seasonally adjuste, we have the Census-X13 method impute the missing values for both the non-seasonally adjusted as well as the seasonally adjusted time series.

Why missing data matter

To maintain Labor Market Update, we rely heavily on public statistics and work to make their interpretation easier and more accessible. The challenges created by the missing October 2025 data underscore the importance of sustained investment in the nation’s statistical infrastructure—an asset whose benefits extend well beyond our own work and across the U.S. and global economy.

(Metadata).