Where Do National Employment Numbers Come From?
(October 8, 2012) Last Friday, the U.S. Bureau of Labor Statistics released the national employment report for September 2012. As South by North Strategies, Ltd. (SBN) noted in its monthly analysis, the American economy gained an estimated 114,000 jobs in September, while the share of the labor force that was unemployed fell to 7.8 percent. That 7.8 percent figure was the lowest national monthly unemployment rate recorded since January 2009.
Normally, the release of national employment data for a given month, which occurs at 8:30 am on the first Friday of the following month, is a relatively straightforward event of interest mainly to policymakers, journalists, business executives, and analysts like SBN. At the same time, the report is one of the most important economic indicators to appear in a month, and the data hold the potential to drive markets, politics, and public policies. That explains why the release of information is tightly controlled, with accredited journalists subject to “lock-up conditions.”
Is The Fix In?
Due to the stakes associated with the upcoming presidential election, the September employment report has become enmeshed in political controversies and surrounded by conspiracy theories, some of which have been percolating at the state and local level for months.
As John Cassidy of The New Yorker noted on Friday in his Rational Irrationality blog, the first allegations of number fixing occurred within five minutes of Friday’s data release, when on-air personalities at CNBC raised the issue. The claims were repeated throughout the day and were amplified through various print, broadcast, and online outlets, as well as through social media. The conspiracy theories seemingly received a further boost when Jack Welch, the retired CEO of General Electric (and beloved mentor of Jack Donaghy of 30 Rock fame), tweeted that the data were suspect. (He later attempted to clarify his views.) More recently, Donald Trump expressed similar concerns on national television.
The past few days have seen a flurry of claims and counterclaims about the BLS figures, especially in the blogoshpere. Two points deserve mention. First, the numbers were not altered for political reasons and accusations to the contrary demonstrate an unfamiliarity with the processes used to generate these important statistics. Second, the September numbers are, by any objective measure, unimpressive and paint a picture of an economy that is generating too few jobs. The numbers are “good” only in relation to the disastrous conditions experienced since the onset of the “Great Recession” in December 2007.
As University of Oregon economist Tim Duy wrote on his well-regarded blog Fed Watch:
The jobs numbers are out, and they are reasonably solid. Reasonably solid as long as you weren’t expecting miracles. Very strong if you thought the economy was heading into recession. But just about where they should be if you think the economy is just sort of grinding along at a slow but steady pace.
The professional view of SBN is broadly consistent with that of Professor Duy. At the same time, the recent flap is fascinating in how it has pushed an issue on which the firm works into the national spotlight and highlights just how unfamiliar many prominent people and ordinary citizens are with some of the nation’s most important economic statistics. Inspired by the remark long attributed to former U.S. Sen. Daniel Patrick Moynihan that “everyone is entitled to their own opinions, but they are not entitled to their own facts,” SBN would like to offer a quick, admittedly simplified, overview of from where national labor market statistics come.
Responsibility for measuring the labor market rests with the U.S. Bureau of Labor Statistics (BLS), an agency within the US Department of Labor. Founded in 1884, the BLS is the oldest permanent major federal statistical agency. Consistent with its founding in the late 19th century, the agency’s mission is to collect “information upon the subject of labor, its relation to capital, the hours of labor and the earnings of laboring men and women, and the means of promoting their material, social, intellectual, and moral prosperity.”
For most of its early history, the BLS had little more than fact-finding authority, but that changed during the 1930s when the magnitude of the Great Depression pushed labor market issues to the center of the national debate and led to the development of social scientific methods of measuring the labor market and associated phenomena like unemployment. Since 1959, the BLS has had sole responsibility for preparing national estimates of labor market conditions.
Today, the BLS, which has its headquarters in the District of Columbia, is responsible for collecting information on employment, unemployment, jobs, wages and benefits, prices, productivity, workplace conditions and injuries, consumer expenditures, time use, and international comparisons. The agency is staffed by professional civil servants with specialized knowledge of such fields as economics and statistics. With the exception of the Commissioner of Labor Statistics, who is nominated by the president and subject to confirmation by the U.S. Senate, none of the agency’s employees are political appointees; even the current commissioner is a career civil servant who is performing the job on an acting basis.
What Do They Count?
The United States is a large country with a sizable labor market. In September, almost 244 million Americans were old enough to be potentially in the civilian labor market, and 155 million persons were either employed or unemployed; the remainder were not participating in the labor force (seasonally unadjusted figures). Similarly, employers had nearly 134 million positions on their payrolls, of which 83.4 percent were in the private sector (again, seasonally unadjusted figures).
These estimates come from two different random surveys sponsored each month by the BLS. The first survey is the Current Population Survey (CPS), which is conducted for the BLS by the US Census Bureau. Each month, the CPS contacts 60,000 households chosen as representative of the civilian noninstitutional population ages 16 and older. A household selected for inclusion in the study participates eight times over an 18-month period. Interviews occur during the week containing the 19th of the month and ask for information pertaining to the week containing the 12th of the month. Participation is voluntary, and all responses are treated as confidential. (Note that different methods are used for state and local estimates.)
The second survey is the Current Employment Statistics (CES) program, which is a federal-state partnership that began in 1915. Each month, the BLS surveys some 486,000 individual business establishments and solicits information about wages, hours, and the total number of jobs on their payrolls during the week that includes the 12th of the month. The survey covers all types of payroll positions (supervisory, non-supervisory and production) and inquires into both full-time and part-time posts.
Because the survey frame is derived from unemployment insurance records, the CES only covers establishments subject to the unemployment insurance system, meaning that agricultural workplaces, domestic employers, and unincorporated small businesses not subject to unemployment insurance taxes are excluded. That said, almost all employed Americans work for establishments covered by unemployment insurance. Furthermore, the connection of survey coverage to the unemployment system means that is is possible to check the survey data against administrative tax records, which is done through a third BLS program called the Quarterly Census of Employment and Wages. Participation in the CES is voluntary, and responses are confidential. Also, the BLS suppresses any information that might permit the identification of individual business establishments.
As hopefully is clear from the preceding discussion, the CPS and CES measure two different things. The CPS measures persons, or individual human beings, typically those age 16 and older. The CES, meanwhile, is a measure of jobs, or positions on the payrolls of individual business establishments. There is not a 1-to-1 relationship between the number of persons and the number of jobs, due partly to persons who hold multiple jobs. Additionally, the two surveys have different geographic frames of reference. The CPS groups people by their place of residence, while the CES categorizes jobs by the place of work. So, a person who lives in Wake County, North Carolina but works in neighboring Durham County would be counted as employed in Wake County under the CPS, but his/her job would be included in the Durham County tally in the CES. It is because the surveys measure different things that they sometimes yield seemingly contradictory results, such as a large jump in employed persons occurring alongside seemingly modest growth in the number of payroll positions, which is what happened in September.
How Do They Count?
Both the CPS and CES data are derived from statistical sampling, a process by which analysts select and study a subset of a population. If chosen properly, the results from the sample will apply to the entire population of interest. The figures nevertheless are estimates that are subject to various forms of sampling and non-sampling errors. For instance, the BLS estimates that the margin of error surrounding the monthly change in the payroll employment level reported in the CES is +/- 100,000 positions. Because the number of jobs gained from August to September exceeded that amount (+114,000), the change was a statistically significant one. Similarly, the monthly change in the unemployment rate is subject to a margin of error of approximately +/- 0.2 percentage points. That means that the 0.3 percentage point decline in the unemployment rate between August and September was a statistically significant fall. Remember that if a monthly change in a given value is not outside of the applicable margin of error, it is not possible to conclude statistically that a real change occurred.
The BLS relies on statistical surveys primarily for reasons of practicality; it simply would not be possible to contact every single American in the labor force and every single business establishment each month. And, while seemingly counter-intuitive, the results of a well-designed sample survey may by less prone to error than the results obtained through a headcount or enumeration.
Recognize that the BLS also makes statistical adjustments to the data collected through surveys like the CES. For instance, the BLS uses statistical imputation to estimate the number of payroll jobs generated and lost each month by new and closing business establishments–important data that the survey is not able to collect. Similarly, the agency revises the CES sample (and past estimates) annually to align or “benchmark” CES estimates with administrative records derived through the unemployment insurance system. For instance, the BLS currently is on track to revise its 2012 payroll estimates upward by 386,000 positions (+0.3 percent); the final revisions will occur on February 1, 2013.
Another factor to keep in mind is that the BLS must navigate a tension between timeliness and completeness, especially in relationship to the CES. If the agency waited until it received all the surveys back from business establishments, it would not be able to release estimates on a predictable monthly schedule. At the same time, releasing figures earlier means that the estimates do not reflect all relevant data points. As a result, the BLS releases payroll figures on a preliminary basis and then revises the data in subsequent months to reflect improved information. These adjustments sometimes can be sizable, so it is best not to read too much into any monthly report and instead view the numbers in light of broader trends spanning more time.
A final adjustment made by the BLS is designed to account for issues related to seasonality. As the BLS states in each month’s national employment report:
Over the course of a year, the size of the nation’s labor force and the levels of employment and unemployment undergo regularly occurring fluctuations. These events may result from seasonal changes in weather, major holidays, and the opening and closing of schools. The effect of such seasonal variation can be very large…. Because these seasonal events follow a more or less regular pattern each year, their influence on the level of a series can be tempered by adjusting for regular seasonal variation.
To accomplish such adjustments, the BLS employs various statistical methods, Yet the adjustment process is imperfect, and the resulting data are subject to periodic revisions and refinements. And as the adjustment factors are based on past trends, they may be slow to capture new, unexpected developments in the labor market (e.g., a re-structuring in the composition of seasonal holiday hiring from retail to trade, transportation, and warehousing owing to the growth of e-commerce).
While the techniques used by the BLS are complex, the agency does endeavor to make those methods transparent. In fact, the agency’s entire Handbook of Methods is available online.
One of the most important but most misunderstood labor market concepts is unemployment. For statistical purposes, the term has a very specific definition that traces its roots to the 1940 creation of the CPS. For the purposes of the monthly employment report, an unemployed person is a member of the civilian labor force who fulfills all three of the following conditions:
- Lacked a job during the survey reference week (the week containing the 12th of the month)
- Was available for work
- Had made at least one active effort to find a job during the previous four weeks
The unemployment rate, then, is the share of the labor force that satisfies those three conditions. Computationally, the rate is obtained by dividing the number of unemployed persons by the total number of persons in the labor force and multiplying the quotient by 100. In September 2012, the national unemployment rate equaled 7.8 percent.
Also known as the U-3 measure, the unemployment rate is just one gauge of the extent to which available labor is not being utilized. In an economy as large and complex as the American one, it is ridiculous to think that one statistic, no matter how rigorously calculated, could adequately capture all economic conditions. The unemployment rate therefore must be interpreted in light of all the other data contained in the monthly employment report to discern larger trends.
Additionally, bear in mind that the CPS definition of unemployment has no relation to the receipt of unemployment insurance compensation; in fact, the vast majority of unemployed Americans never collect unemployment insurance benefits, either because they never apply or because they fail to meet the eligibility criteria set by their respective states. In 2008, for instance, just 36 percent of unemployed persons collected insurance payments. This means that state administrative data on the number of persons receiving unemployment insurance, like those released each Thursday morning by the US Department of Labor, are not interchangeable with CPS data related to unemployment.
People sometimes accuse the BLS of not reporting the “real” or “true” unemployment for political purposes. While such critics rarely specify exactly what they mean, they often seem to be talking about underemployment, which is a broader conceptualization of labor under-utilization. The category of underemployed persons includes the following three groups:
- Persons who are unemployed
- Persons who are marginally attached to the labor force, meaning those who have stopped looking for jobs due to discouragement AND those who are conditionally interested in work but are not seeking it for reasons other than discouragement, such as a lack of childcare
- Persons who would prefer to work on a full-time basis but are working involuntarily part-time
The underemployment rate (also known as the U-6 measure) is the share of the adjusted labor force that is effectively jobless. Computationally, the rate is obtained by dividing the sum of unemployed persons, marginally attached workers, and involuntary part-time workers by the sum of the civilian labor force and marginally attached workers and then multiplying the quotient by 100. In September 2012, the underemployment rate totaled 14.7 percent.
While the unemployment rate is the official measure of joblessness and the one that attracts the attention of journalists each month, the underemployment rate is not hidden by the BLS. In fact, the agency publishes the official unemployment rate, the underemployment rate, and four other alternative measures in both seasonally adjusted and unadjusted formats. These measure are found in table A-15 in each monthly report on the employment situation.
Allegations of data manipulation on the part of the BLS are ill-informed ones that point to a lack of understanding of how important economic numbers are computed, what they measure, and how to interpret them properly. While the accusations are troubling, they hopefully can create an “educable moment” to improve understanding of the problems facing the American labor market and the ways in which statistical data like those compiled by the BLS can inform policymaking.
For, when all is said and done, the numbers contained in the September employment report are by any measure disturbing ones that point to a labor market that simply is not creating enough jobs for all those Americans who want and need work. This is a crisis that has shown no sign of abating anytime soon and that will confront whichever candidate wins the presidency in November.
(Updated on October 10, 2012 to include the CES preliminary benchmark announcement. Updated on October 11, 2012 to clarify the relationship between CPS measures of unemployment and the receipt of unemployment insurance compensation.)