US non-farm payrolls background


19 Nov 2009

The non-farm payrolls (NFP) data series, from the Bureau of Labor Statistics (BLS), is the best proxy available for total employment in the country, and is thus one of the most basic indicators of the overall employment situation. NFP are based on a monthly sampling, interpreted and reported in a matter of days, against a background of more comprehensive, but less timely, quarterly survey results. The monthly results necessarily involve some extrapolation (including the much-commented “net birth-death adjustment” concerning new businesses not yet reporting). While these extrapolations are reasonable, the MTM changes are quite noisy, and should not be given too much weight. The YOY change, with a neutral value of 1%, shows more clearly the trend. The YOY change is strongly correlated with, and lags, recession.


Entries below covered through 6 Feb 2009.

Data sources

  • Coverage (19 Nov 2009) “The data exclude proprietors, the unincorporated self-employed, unpaid volunteer or family workers, farmworkers, and domestic workers. Salaried officers of corporations are included. Government employment covers only civilian employees; military personnel are excluded. Employees of the Central Intelligence Agency, the National Security Agency, the National Imagery and Mapping Agency, and the Defense Intelligence Agency also are excluded.”
  • Jobs, not individuals (19 Nov 2009) Note that the NFP is based on surveys of jobs, not individuals. Someone with two jobs may be counted twice.
  • QCEW (19 Nov 2009) The Quarterly Census of Employment and Wages (QCEW) program is carried out jointly by the BLS and the state employment security agencies. The main limitation of the QCEW data is the delay in publication: about 7 months after the quarter for quarterly results, and about 10 months after year end for the annual results.
  • CES (1 Aug 2009) The BLS carries out the Current Employment Statistics (CES) survey monthly, covering about 1/3 of all NFP workers. Preliminary results are available a few days after the end of the month in question.

Methodology, overview (19 Nov 2009) The primary methodological challenge is that companies that go out of business ('deaths') stop reporting immediately, but new companies ('births') are not included in the survey universe for about 7 months. The first order method is to calculate employment for any given month from the employment for the previous month by using the ratio for those businesses that respond in both months. So, for the first order approximation, total employment is assumed to behave like employment in continuously existing companies or, equivalently, each business death is assumed to have a corresponding business birth that does not change the sample behavior. A smaller, second order adjustment is the much-commented 'birth-death model', that adds a term extrapolated from long-term differences between actual counts (from the QCEW) and the first-order adjusted series. Each fall, when the March QCEW results come out, the March values are re-benchmarked and the following months re-calculated.

Methodology, strengths and weaknesses (19 Nov 2009)

  • At the annual benchmarking, (1) the net revision has average absolute value 0.2% of NFP, and (2) the revision would be significantly larger without the net birth/death adjustment, showing that overall the extrapolation methods provide a reasonable approximation to the actual data. 0.2% of NFP is also typically small compared to the actual YOY change, showing that the YOY change is not too noisy.
  • A frequent complaint is that the methodology, by extrapolating from long term data, does not reflect changes in business conditions. It is true, of course, that some radical changes cannot be reflected in the extrapolation – clearly mortgage lending companies were not being replaced as many went out of business in 2008. However the main trend in business conditions is reflected, since the main component of the reported number is simply scaling up from continuously existing businesses. Also, the net birth/death adjustment is largely independent of business conditions.
  • One frequent complaint is that the birth/death adjustment is too large to make any sense. This is based on a misunderstanding – the complaint is based on summing the monthly entries in a table the BLS publishes of the net birth/death adjustment, but the table entries are the cumulative net adjustment to date for each month, i.e., already summed.
  • the MTM changes are very often significantly revised. Below is a typical period. Mid-2007 was chose because there was considerable controversy about whether the nation was about to enter recession, and every month the release of the NFP generated intense arguments. Yet, as the table makes clear, the first number out was often quite significantly revised a month or two later. For Aug 2007 the MTM change was even revised from negative to positive. Note also that even the final revisions (the rightmost column of the above table) jump around a great deal. (Numbers are thousands of jobs; source is
Month Initial report of MTM change First revision Second revision
2007-Jun 132 126 69
2007-Jul 92 68 93
2007-Aug -4 89 93
2007-Sep 110 96 44
2007-Oct 166 170 159

Long term, MTM change (25 Jun 2009) It is often stated that roughly 100k non-farm jobs need to be added each month just to keep pace with growth in the workforce. Presumably over the long term the growth in payrolls should roughly match the growth in the working-age population. The total population of those aged 20-64 was estimated in the 2000 census at 166.5m, and projected to be 185.5m in 2010. This is an 11.4% change in 10 years; giving a CAGR of 1%. 1% of the current estimated non-farm payrolls of 137,846,000, divided by 12 months, works out to 115k/month.

Drop of YOY change below neutral rate correlates with recessions (4 Aug 2009) On average, the first drop in NFP YOY below 1% occurs about 1.6 months after the recession starts. The first rise above 1%, after an extended period below the neutral rate, occurs on average 12 months after the recession ends.

Recession Drop < 1% Drop to recession Rise > 1% Rise to recession end
Feb 1945 - Oct 1945 Feb 1944 12 months Jun 1946 -8 months
Nov 1948 - Oct 1949 Jan 1949 -2 May 1950 -7
Jul 1953 - May 1954 Nov 1953 -4 Mar 1955 -10
Aug 1957 - Apr 1958 Jun 1957 +2 Feb 1959 -10
Apr 1960 - Feb 1961 Jul 1960 -3 Nov 1961 -9
Dec 1969 - Nov 1970 Jun 1970 -6 Oct 1971 -11
Nov 1973 - Mar 1975 Nov 1974 -12 Jan 1976 -10
Jan 1980 - Jul 1980 May 1980 -5 Jun 1981 -11
Jul 1981 - Nov 1982 Nov 1981 -4 Sep 1983 -10
Jul 1990 - Mar 1991 Oct 1990 -3 Dec 1992 -17
Mar 2001 - Nov 2001 Mar 2001 0 May 2004 -29
Dec 2007 - Aug 2007 5 - -


See also

Selected commentary

Projected population change

Undated (based on 2000 census). Census web site.

The estimated/projected total population, in the US, of those aged 20-64, was 166.5m in 2000 and 185.5m in 2010. This is an absolute growth of about 158,000/mo, or a CAGR of 1.01%.

QCEW background

18 Dec 2003. BLS web site.

The Quarterly Census of Employment and Wages Program is a cooperative program involving the Bureau of Labor Statistics (BLS) of the U.S. Department of Labor and the State Employment Security Agencies (SESAs). The QCEW program produces a comprehensive tabulation of employment and wage information for workers covered by State unemployment insurance (UI) laws and Federal workers covered by the Unemployment Compensation for Federal Employees (UCFE) program. Publicly available files include data on the number of establishments, monthly employment, and quarterly wages, by NAICS industry, by county, by ownership sector, for the entire United States. These data are aggregated to annual levels, to higher industry levels (NAICS industry groups, sectors, and supersectors), and to higher geographic levels (national, State, and Metropolitan Statistical Area (MSA)).

The QCEW program serves as a near census of monthly employment and quarterly wage information by 6-digit NAICS industry at the national, State, and county levels. At the national level, the QCEW program publishes employment and wage data for nearly every NAICS industry. At the State and area level, the QCEW program publishes employment and wage data down to the 6-digit NAICS industry level, if disclosure restrictions are met. In accordance with BLS policy, data provided to the Bureau in confidence are not published and are used only for specified statistical purposes. BLS withholds publication of UI-covered employment and wage data for any industry level when necessary to protect the identity of cooperating employers. Totals at the industry level for the States and the Nation include the nondisclosable data suppressed within the detailed tables. However, these totals cannot be used to reveal the suppressed data.

Employment data under the QCEW program represent the number of covered workers who worked during, or received pay for, the pay period including the 12th of the month. Excluded are members of the armed forces, the self-employed, proprietors, domestic workers, unpaid family workers, and railroad workers covered by the railroad unemployment insurance system. Wages represent total compensation paid during the calendar quarter, regardless of when services were performed. Included in wages are pay for vacation and other paid leave, bonuses, stock options, tips, the cash value of meals and lodging, and in some States, contributions to deferred compensation plans (such as 401(k) plans). The QCEW program does provide partial information on agricultural industries and employees in private households.

Data from the QCEW program serve as an important input to many BLS programs. The QCEW data are used as the benchmark source for employment by the Current Employment Statistics program and the Occupational Employment Statistics program. The UI administrative records collected under the QCEW program serve as a sampling frame for BLS establishment surveys.

In addition, data from the QCEW program serve as an input to other Federal and State programs. The Bureau of Economic Analysis (BEA) of the Department of Commerce uses QCEW data as the base for developing the wage and salary component of personal income. The Employment and Training Administration (ETA) of the Department of Labor and the SESAs use QCEW data to administer the employment security program. The QCEW data accurately reflect the extent of coverage of the State UI laws and are used to measure UI revenues; national, State and local area employment; and total and UI taxable wage trends.

BLS publishes data from the QCEW program every quarter in the County Employment and Wages press release. This is usually released 6 to 7 months after the end of the quarter. The QCEW program also publishes a subset of its quarterly data through the Create Customized Tables system, and full quarterly industry detail data at all geographic levels in ASCII files through our FTP server.

In addition, QCEW publishes the annual bulletin Employment and Wages, Annual Averages about 10 months after the end of the year. The 2002 edition is the first to appear on the internet in its entirety. For more information about this, please see the Publications section of our home page. Also, the cooperating SESAs issue quarterly and/or annual reports of QCEW data.

In addition to yearly benchmark, quarterly differences are monitored

11 Oct 2005. BLS FAQ page on birth/death model.

“Q: Does BLS ever revise birth/death factors once a new quarter of QCEW [Quarterly Census of Employment and Wages] universe employment data becomes available?

A: There is no formal refitting of the birth/death factors with each new quarter of QCEW data. However, the monthly estimates are tracked against universe employment counts. If there were indications that the model was misspecified, then the model would be reviewed and potentially changed for the relevant industries. Since the introduction of the birth/death model in June 2000, there have not been any mid-year interventions.”

BLS sampling and modeling for non-farm payroll numbers

May 2006. Monthly Labor Review, p28, 31.

“Impact of business births and deaths in the payroll survey. by Kirk Mueller”

“The Current Employment Statistics (CES) survey, conducted by the Bureau of Labor Statistics (BLS), is a monthly survey of more than 400,000 business establishments. The CES program provides estimates on employment, hours, and earnings by industry detail for the Nation, States, and metropolitan areas. … The CES sample-based employment estimates for March of each year are benchmarked, or reanchored, annually to the March universe count derived principally from the Quarterly Census of Employment and Wages (QCEW) program. These QCEW population counts are much less timely than sample-based estimates and are used to provide an annual point-in-time census for employment. For national series, only the March sample-based estimates are replaced with the population counts. … The new survey design [phased in 2000-2003] includes both the new sample composition [probability-based rather than quota-based] and the use of a two-step process to account for the employment associated with business births. This process first imputes a portion of the birth employment from the employment associated with business deaths. The second step models the historical difference between the imputation and the actual relationship between business birth and business death employments; this step is referred to as the net birth/death model. … Like many establishment surveys, the CES has difficulty with obtaining and developing a timely sample frame for business births. This is because a lag exists between an establishment opening for business and its appearance on the universe frame, where it would be available for sampling. This lag currently is about 7 months. In contrast, the lag from the reference month to the CES first publication of the employment estimates for that reference month is 3 weeks. … the net birth/death model. The Bureau’s Longitudinal Database (LDB) is the basis for developing the historical relationship to be modeled. The LDB links establishments over time, which allows for the identification of the continuous establishments, establishments that go out of business, and births of establishments. To develop the history for modeling, the same handling of business deaths as described for the CES sample data is applied to the population data. Establishments that go out of business have employment imputed for them based on the rate of change of the continuous units. The employment associated with continuous units and the employment imputed from deaths are summed. The difference is compared with the actual population level to create the series modeled by the birth/death models. To date, the residual net birth/death component has shown to be a relatively stable portion of the population employment regardless of the point in the business cycle.”

ÂEc  =  estimate of all employees for the current month  
ÂEp = estimate of all employees for the previous month 
i = the i-th sample unit 
wi = the weight for the i-th sample unit 
aec,i = i-th sample unit that reports for the current month 
aep,i = i-th sample unit that reports for the previous month"

Employment baseline

15 Sep 2007. Economist p15.

“America needs to create at least 100,000 jobs a month merely to absorb the growing working population.”

BLS sampling and methodology for non-farm payroll numbers

4 Jan 2008. BLS web page “CES Net Birth/Death Model”

“In 2007, the CES sample includes about 160,000 businesses and government agencies drawn from a sampling frame of Unemployment Insurance tax accounts which cover approximately 400,000 individual worksites. The active CES sample includes approximately one-third of all nonfarm payroll workers. … BLS is implementing an estimation procedure with two components: the first component uses business deaths to impute employment for business births. This is incorporated into the sample-based link relative estimate procedure by simply not reflecting sample units going out of business, but imputing to them the same trend as the other firms in the sample. … The second component is an ARIMA time series model designed to estimate the residual net birth/death employment not accounted for by the imputation. The historical time series used to create and test the ARIMA model was derived from the UI universe micro level database, and reflects the actual residual net of births and deaths over the past five years. The ARIMA model component is updated and reviewed on a quarterly basis. … The most significant potential drawback to this or any model-based approach is that time series modeling assumes a predictable continuation of historical patterns and relationships and therefore is likely to have some difficulty producing reliable estimates at economic turning points or during periods when there are sudden changes in trend. BLS will continue researching alternative model-based techniques for the net birth/death component; it is likely to remain as the most problematic part of the estimation process. … The table below shows the net birth/death model adjustment used in the published CES estimates since the establishment of the most recent benchmark level for March 2006.”

Absolute size of current non-farm payrolls

4 Apr 2008. BLS press release.


Non-farm employment, Mar 2008, 137,846,000.

Multiple jobholders

8 Aug 2008. BLS methodology overview page.

“Monthly Employment Situation Report: Quick Guide to Methods and Measurement Issues”

“The CES employment series are estimates of nonfarm wage and salary jobs, not an estimate of employed persons; an individual with two jobs is counted twice by the payroll survey. …

The CPS estimate of employment is for the total number of employed persons. Included are categories of workers that are not covered by the Current Employment Statistics (CES) survey: self-employed persons, private household workers, agriculture workers, unpaid family workers, and workers on leave without pay during the reference period. Multiple jobholders are counted once in the estimate of total employed.”

BLS on the B/D adjustment

3 Oct 2008. BLS paper (thanks to Jeffrey Miller at A Dash of Insight)

“Issues in Labor Statistics: How the Business Birth/Death Model Improves Payroll Employment Estimates”

“Each year, the Current Employment Statistics (CES) program adjusts its sample-based estimates of industry employment for the previous year to universe employment counts derived primarily from the Quarterly Census of Employment and Wages (QCEW). CES recently announced a preliminary benchmark revision of -21,000 for its March 2008 total nonfarm employment estimate, an adjustment of less than one-tenth of one percent; the average absolute revision over the past decade is two-tenths of one percent. (See table below.)

The preliminary 2008 benchmark calculation indicates that the CES estimate of March 2007 – March 2008 employment change (+500,000) is nearly identical to that reflected in QCEW universe counts (+479,000). … The birth/death adjustment has consistently resulted in a smaller benchmark revision than would have been the case if the sample alone were used to measure payroll employment change. (See chart 1.)

Actual net employment change from business births and deaths follows a relatively stable trend across years and is not highly correlated with changes in the business cycle, or even with changes in overall employment. …

Note: The QCEW is a nearly complete count of employment for all businesses in the US and is based on administrative records of the Unemployment Insurance tax system that most businesses are required to file with their state. The benchmark adjustment procedure replaces the March sample-based employment estimates with the employment levels from the universe for March of each year. For more information on the QCEW, see”

CES does not cover farmers; covers some but not all government workers

6 Feb 2009. BLS background page.

“Technical Notes to Establishment Survey Data Published in Employment and Earnings”

“BLS collects data each month on employment, hours, and earnings from a sample of nonfarm establishments. The sample includes about 150,000 businesses and government agencies, which cover approximately 390,000 individual worksites drawn from a sampling frame of roughly 8.9 million Unemployment Insurance tax accounts. The active CES sample includes approximately one-third of all nonfarm payroll workers. …

The data exclude proprietors, the unincorporated self-employed, unpaid volunteer or family workers, farmworkers, and domestic workers. Salaried officers of corporations are included. Government employment covers only civilian employees; military personnel are excluded. Employees of the Central Intelligence Agency, the National Security Agency, the National Imagery and Mapping Agency, and the Defense Intelligence Agency also are excluded.”

'Neutral rate' of change in NFP varies with workforce participation

28 Aug 2009. Calculated Risk.

“Unemployment and Net Jobs. by CalculatedRisk”

“graph shows the monthly change in net jobs (on the x-axis) as a percentage of the civilian workforce, and the change in the unemployment rate on the y-axis.

The data is for the last 40 years: 1969 through July 2009. …

the economy needs to be adding about 0.33 percent of the civilian workforce per quarter to keep the unemployment rate from rising. That is about 170 thousand net jobs per month.

Note that the trend line is a 2nd order polynomial (equation on graph). When the economy starts to add jobs, more people start looking for work - and the relationship between net jobs and unemployment rate is not linear.”

[i.e. as sentiment improves, people come back into the labor force, and so more jobs are needed to keep unemployment at the same level.]

Wikipedia says farmers not covered due to seasonality

6 Nov 2009. Wikipedia.

“Nonfarm payrolls”

“The farming industry is not included because of its seasonal hiring which would distort the number around harvest times (as farms add workers and then release them after the harvest is complete).”

Update on birth-death model