Use the menus below to change the range of data and individual data series you'd like to view. To save this chart as an image or PDF document, select an option from the "Export" menu. See our modifications to the spider chart, noted in red in the first passage under the Indicators tab, in response to the impact of COVID-19 on the labor market.
1. Why does the Federal Reserve care about employment?
Section 2A of the Federal Reserve Act states, "The Board of Governors of the Federal Reserve System and the Federal Open Market Committee shall maintain long-run growth of the monetary and credit aggregates commensurate with the economy's long run potential to increase production, so as to promote effectively the goals of maximum employment, stable prices, and moderate long-term interest rates." This part of the Federal Reserve Act is often referred to as the Fed's "dual mandate." Basically, it states that the Federal Reserve's monetary policy has the goals of stable prices and maximum employment. The gap between the unemployment rate and the estimated normal rate of unemployment is the most popular statistic that measures the degree to which the Federal Reserve has achieved the goal of maximum employment.
2. What is the basic idea behind the spider chart?
The spider chart uses 15 measures of labor market activity. Where necessary, the indicators are transformed so that they do not have a clear upward or downward trend, either by dividing by the size of the labor force or, in the case of the two wage/compensation measures, conversion to 12-month growth rates. Indicators like the unemployment rate, where larger values correspond to a weaker labor market, are multiplied by -1.
After these transformations, the indicators are rank-ordered over a fixed sample period and assigned to the value of their cumulative distribution function. For example, the maximum value is assigned 100, the minimum is assigned zero, and the median is assigned 50. The values of the cumulative distribution function are then plotted in the spider chart. The outer- and inner-black circles correspond to the maximum and minimum values of the indicators, respectively, while the fainter gray circle in between corresponds to the median values of the indicators.
3. How do you handle ties?
This is best illustrated by example. Suppose the data considered are the 11 numbers consisting of each of the counting numbers from zero to 7 and the number 8 repeated three times. For each of the numbers k between zero and 7, k is assigned the value 10 times k. The repeated three values of 8 are arbitrarily ordered and preliminary assigned their three percentiles in the cumulative distribution: 80, 90, and 100. Each of the three 8s is then assigned to the average of these three percentiles: 90 = (80+90+100)/3. In this example, none of the numbers are assigned to the largest possible value in the spider plot of 100 since there are ties at the maximum value.
4. How did you choose the start dates of the sample periods for the distributions of the labor market indicators?
We do not use data prior to January 1994, as a major redesign of the Current Population Survey was introduced at this date so that methodologically consistent time series for both marginally attached workers and part-time workers for economic reasons cannot be constructed over a period beginning before this month. The earliest and default sample start date is March 1994. Other sample start dates are the beginning months of the last two recessions as dated by the National Bureau of Economic Research (March 2001 and December 2007) and the ending months of those recessions (November 2001 and June 2009). Each of these two recessions was followed by periods of continued declines in payroll employment and increases in the unemployment rates. By these criteria, the troughs in the labor market following these recessions were right around August 2003 and December 2009. We allow users to select these months as sample start dates. Finally, we allow users to choose sample start dates so there are either exactly five, 10, 15, or 20 years of data over which the distributions are calculated. These start dates will move up a month with each additional month of labor market data. In total, there are up to 10 possible sample start dates that can be chosen.
5. How can you have sample start dates as far back as 1994 when data from the Job Openings and Labor Turnover Survey (JOLTS) begins in December 2001?
We extend the private hires and quits rates back to 1994 using data constructed by Steven J. Davis, R. Jason Faberman, and John Haltiwanger for their article "Labor Market Flows in the Cross Section and over Time" published in the January 2012 issue of the Journal of Monetary Economics. We extend the private openings rate back to 1994 using the Composite Help-Wanted Index constructed by Regis Barnichon for his article "Building a Composite Help-Wanted Index" published in the December 2010 issue of Economics Letters. The exact details of how these data are spliced together with the JOLTS data are provided in the "Indicators" section of this webpage.
6. Why not simply look at the unemployment rate?
Commentary on the labor market tends to focus on the unemployment rate as the summary measure of the health of the labor market. However, while trends in the unemployment rate over the medium term are a pretty good gauge of changes in overall labor market conditions, over short periods of time the unemployment rate can be influenced by factors that make it a less reliable directional gauge. For example, it is possible that the unemployment rate could rise for a while as conditions improve as those currently out of the labor force enter at a faster rate but fail to secure a job immediately.
7. Why do you use the three-month change in nonfarm payroll employment rather than the three-month growth rate?
In typical applications, when considering economic variables that are presumed to grow exponentially, transformations using growth rates or log differences are much more commonly used than raw differences. For example, real gross domestic product (GDP) in 2014 was more than eight times larger than it was in 1947. Consequently, comparing changes in the level of real GDP in recent years with changes in the level of real GDP in the late 1940s is not at all useful as the former are, on average, much larger and much more volatile due to exponential growth of real GDP. Comparing growth rates of real GDP is much more informative. An analogous treatment for payroll employment is not necessarily appropriate due to changes in the growth rate of the working-age population (ages 16 to 64) over the past 20+ years. In 1994, the annualized growth rate of the working-age population was around 1.0 percent while in 2015 it appears to be around 0.5 percent. Thus, the growth rate of payroll employment needed to keep the unemployment rate constant is probably about twice as large in 2015 as it was in 1994. Using raw differences in payroll employment mitigates this problem to some extent. In January 1994, nonfarm payroll employment needed to increase by about 90,000 jobs a month to pace with the growth rate of the working-age population. In January 2015, they only needed to increase about 60,000 per month. This is still a large difference, but, proportionately, it's only about half as large as the difference between 1.0 percent growth and 0.5 percent growth.
8. Why wasn't labor force participation used in the set of indicators? Why is the age 25 to 54 employment-population ratio used instead of the standard ratio for all civilians age 16-plus?
Over the 10 years ending in December 2015, the labor force participation rate declined from 66.0 percent to 62.6 percent. However, roughly two-thirds of this decline (2¼ percentage points) can be accounted for by demographic changes in the age/sex distribution of the population (primarily reflecting the aging of the baby-boomer generation into retirement ages). Hence, an apples-to-apples comparison between today's labor force participation rate and the rate 10 years ago cannot be made. Using the age 25 to 54 employment-population ratio is a standard way to adjust for the aging of the population, since this "prime-age" population is commonly thought of as old enough to have completed school (in most cases) but too young for retirement. However, even this adjustment is quite imperfect. According to data from the Atlanta Fed's Labor Force Participation Dynamics website, the proportion of the age 26-to-55 population who did not want a job increased from 14.4 percent in 1998 to 16.9 percent in 2014. About 60 percent of this increase was due to a higher frequency of persons self-reporting that they are "ill" or "disabled," while the remaining increase was equally split between increases in "retirement" and "in school/training." These data imply that one should be very careful when comparing the employment-population ratio for prime-age workers over long periods of time.
9. How are data released after the Employment Situation release handled?
The JOLTS data series (quits, hires, and openings) for a given month are released about five and a half weeks after the BLS's Employment Situation. Therefore, on the day of the Employment Situation release for month t, only JOLTS data through month t-2 will be available. In this case, the spider plot values of the JOLTS series in months t-1 and t will be set equal to their corresponding spider plot values for month t-2.
Also, on the day of the Employment Situation release for month t, NFIB survey data will only be available through month t-1. In this case, the spider plot values of the NFIB series in month t will be set equal to their corresponding spider plot values for month t-1.
10. Initial unemployment insurance claims are weekly; how are they converted to monthly?
Daily claims are assumed to be constant within each week. Monthly claims are taken to be the average of the daily claims for all the weekdays in the month.
11. Do larger values always correspond to outward movement on the chart?
No. There are four variables that move inversely to payroll employment. These are the unemployment rate, initial unemployment insurance claims, part-time for economic reasons, and marginally attached. The latter three variables are all divided by the size of the labor force. For all four variables, the indicator is inverted so that a decline is represented by outward movements on the chart.
12. Are there other ways to visualize labor market conditions? Or are there uses of spider charts in other applications?
Besides the early version of the earlier version of the labor market spider plot introduced in Macroblog entries here and here, the Atlanta Fed has a page dedicated to labor force participation dynamics that uses visualization tools to provide insight as to why labor force participation has been declining since at least the start of the 2007–09 recession. The New York Fed's "Eight Different Faces of the Labor Market" provides time-series plots of labor market indicators that are assigned to one of eight different categories like "job loss" and "wages." The Federal Reserve Board's discontinued Labor Market Conditions Index and the Kansas City Fed's Labor Market Conditions Indicators combine roughly 20 labor market indicators into one, or several, summary statistics using statistical factor model methods. Finally, a number of Federal Reserve Board economists used almost exactly the same methodology used here to construct a spider plot visualizing vulnerabilities in the U.S. financial system.
13. The marginally attached workers category is not seasonally adjusted by the BLS; do you seasonally adjust it yourselves?
Yes. The series is seasonally adjusted using the default (X12-ARIMA) settings in Haver Analytics.
Note: Effective with the May 8, 2020, update, we have ceased using three-month averages to smooth out the data so as not to understate the impact of COVID-19 on the labor market. Percentiles of the 12-month growth rates of both the Employment Cost Index and average hourly earnings are used for the spider chart.
Payroll employment (CES)
The monthly change in the level of employment reported in the payroll survey less the monthly change in federal government Decennial Census temporary and intermittent workers.
Private job openings rate (JOLTS, spliced with data from researcher Regis Barnichon prior to December 2000)
Job openings are positions (not filled) on the last business day of the month; a job is "open" only if it meets all three of the following conditions:
- A specific position exists and there is work available for that position. The position can be full- or part-time, and it can be permanent, short term, or seasonal.
- The job could start within 30 days, whether or not the establishment finds a suitable candidate during that time.
- There is active recruiting for workers from outside the establishment location that has the opening.
Private openings from JOLTS start in December 2000; we extend it back to 1994 using the Composite Help-Wanted Index (Composite HWI) constructed by Regis Barnichon. The Composite HWI combines the Conference Board's discontinued HWI of print advertising with its HWI of online advertising. To put the pre-December 2000 values of the Composite HWI on the same basis as JOLTS private openings, we multiply them by the December 2000 ratio of JOLTS private openings to the Composite HWI. The spliced series of private openings is converted to a rate by dividing by the sum of (spliced) private openings and private payroll employment.
Private hires rate (JOLTS, spliced with data from Steven J. Davis, R. Jason Faberman, and John Haltiwanger prior to December 2000)
Private hires are the number of additions to private payrolls during the month. It is converted to a rate by dividing by the level of private payroll employment. The JOLTS hires rate starts in December 2000; we extend the series back to 1994 using data constructed by Davis, Faberman, and Haltiwanger (DFW) available here at Steven Davis' website. The DFW private hires rate is quarterly; we convert it to monthly by dividing it by three and assigning the resulting value to all three months of the quarter. We put the pre-December 2000 values of this monthly hires rate on the same basis as JOLTS by adding the difference between the December 2000 JOLTS private hires rate and the December 2000 DFW monthly hires rate.
Hiring plans (NFIB)
The share of surveyed firms that plan to increase total employment over the next three months.
Job availability (Conference Board)
Percentage of survey respondents who say they find "jobs plentiful."
Private quits rate (JOLTS, spliced with data from Steven J. Davis, R. Jason Faberman, and John Haltiwanger prior to December 2000)
Private quits measure the number of nongovernment employees who left voluntarily, with the exception of retirements or transfers to other locations. It is converted to a rate by dividing by the level of private payroll employment. As with the private hires, JOLTS data start in December 2000. We extend this data back to 1994 with the Davis, Faberman, and Haltiwanger (DFW) quarterly quits rate in exactly the same way we extend the private JOLTS hires rate prior to December 2000 with the DFW hires rate.
Firms unable to fill job openings (NFIB)
Share of surveyed firms reporting at least one job opening they are currently not able to fill.
Unemployment rate (CPS)
Percentage of persons in the civilian labor force who are unemployed.
Employment-population ratio, ages 25–54 (CPS)
The proportion of the civilian noninstitutional population ages 25 to 54 years who are currently employed.
Marginally attached workers (percent of labor force) (CPS)
Marginally attached workers are individuals not in the labor force who want and are available for work, and who have looked for a job sometime in the prior 12 months, but were not counted as unemployed because they had not searched for work in the four weeks preceding the survey. The number of marginally attached workers is divided by the size of the civilian labor force.
Work part-time for economic reasons (percent of labor force) (CPS)
This category includes the number of persons who indicated that they would like to work full-time but were working part-time (one to 34 hours a week) because of an economic reason, such as their hours were cut or they were unable to find full-time jobs. The number of these workers is divided by the size of the civilian labor force.
Employment Cost Index growth (NCS)
The Employment Cost Index (ECI) for total compensation measures the change in the cost of labor, free from the influence of employment shifts among occupations and industries. The index covers workers in the private nonfarm economy except those in private households, and workers in the public sector, except the federal government. Total compensation includes wages, salaries, and employer costs for employee benefits.
The 12-month percent change in the ECI is used for the spider plot. The ECI is released quarterly and covers the third month of each quarter. To convert the series to monthly, we assume that the 12-month compensation growth rate for the first and second months of a particular quarter equals the 12-month growth rate for the third month of that quarter.
Average hourly earnings growth (CES)
This category measures the 12-month percent change in average hourly earnings of private-sector production and nonsupervisory employees. Workers in this group include production and related employees in manufacturing and mining and logging, construction workers in construction, and nonsupervisory employees in private service-providing industries.
Initial claims (percent of labor force) (UI)
Initial claims measure the number of new claims for unemployment insurance. The number of claims is divided by the size of the civilian labor force.
Job finding rate (CPS)
This series is constructed using data from the research series on labor force status flows from the CPS. The job finding rate is the total number of employed persons who were unemployed in the prior month (UE) divided by the total number of persons who were both unemployed in the prior month and part of the age 16-plus civilian noninstitutional population in the current month. Data on labor force status flows are available here.
The U.S. Bureau of Labor Statistics, through the Atlanta JOLTS Data Collection Center, collects data from a sample of approximately 16,000 U.S. business establishments. The JOLTS survey covers all nonagricultural industries in the public and private sectors for the 50 states and the District of Columbia. JOLTS collects data on total employment, job openings, hires, quits, layoffs and discharges, and other separations.
Through the establishment survey, which is formally called the Current Employment Statistics (CES) program, the BLS surveys approximately 145,000 nonfarm businesses covering about 557,000 work sites, asking employers about employment, hours, and earnings of their workers. The establishment survey is commonly referred to as the payroll survey.
The survey's total employment number reflects an estimate of the number of people in the United States who received a paycheck for work during the pay period that includes the 12th day of the month. This count most accurately gives the total number of jobs in the country at a given point in time.
The BLS also surveys about 60,000 households each month to obtain estimates of employment and nonemployment activity, total income, and demographics of the population of the United States. The reference period for activity is the same week as the establishment survey: the week that includes the 12th day of the month. The household survey is officially called the Current Population Survey (CPS).
The U.S. Department of Labor produces the UI report.
This survey is conducted monthly of members of the NFIB. There are generally between 500 and 2,500 responses each month and they are used to construct the Index of Small Business Optimism.
This survey uses a probability sample design to select each month's random sample from the household universe frame. There is a target of 3,000 responses per month.
The NCS is a quarterly survey by the U.S. Bureau of Labor Statistics of over 8,000 establishments on employee salaries, wages and benefits. The survey provides the source data for both the Employment Cost Index and Employer Costs for Employee Compensation news releases.
Release times shown are from the original source. The chart is usually updated within a few hours following these times.
||Date of release
||Time of release
|Job Openings and Labor Turnover Survey (JOLTS): November 2022
|Employment situation: December 2022
|Job Openings and Labor Turnover Survey (JOLTS): December 2022
|Employment situation: January 2023
|Employment situation: February 2023
|Job Openings and Labor Turnover Survey (JOLTS): January 2023
|Job Openings and Labor Turnover Survey (JOLTS): February 2023
|Employment situation: March 2023
|Job Openings and Labor Turnover Survey (JOLTS): March 2023
|Employment situation: April 2023
|Job Openings and Labor Turnover Survey (JOLTS): April 2023
|Employment situation: May 2023
|Job Openings and Labor Turnover Survey (JOLTS): May 2023
|Employment situation: June 2023
|Job Openings and Labor Turnover Survey (JOLTS): June 2023
|Employment situation: July 2023
|Job Openings and Labor Turnover Survey (JOLTS): July 2023
|Employment situation: August 2023
|Job Openings and Labor Turnover Survey (JOLTS): August 2023
|Employment situation: September 2023
|Job Openings and Labor Turnover Survey (JOLTS): September 2023
|Employment situation: October 2023
|Job Openings and Labor Turnover Survey (JOLTS): October 2023
|Employment situation: November 2023
Download a spreadsheet of these release dates