Handbook of Methods Current Employment Statistics - State and Metro Area History
Current Employment Statistics - State and Metro Area: History
From its origins in 1915 to today, the Current Employment Statistics - State and Area (CES-SA) program expanded and evolved gradually over time. The program has been shaped by advances in technology and statistical methods, user demand, recommendations from numerous presidential commissions and special panels, and changing resource levels.
Hover over the red dot to see historical information.
Major milestones
1915: In October 1915, the Bureau of Labor Statistics (BLS) first began to collect employment and payroll data for four manufacturing industries—boots and shoes, cotton goods, cotton finishing, and hosiery and underwear. Data from this survey were published as month-to-month changes in employment and payroll. Before BLS started collecting employment and payroll data, several states conducted surveys of industry employment; New York, New Jersey, and Massachusetts were among the first to do so.
1916: BLS and the State of New York signed the first cooperative agreement, allowing BLS to use sample data collected by New York in its national estimates.
1916: BLS expanded the CES program to publish data for 13 industries at a national level with a sample size of 574 establishments covering 519,185 jobs.
1920: Wisconsin became the second state to sign a cooperative agreement with BLS.
1923: BLS expanded the program to publish 29 additional manufacturing industries at a national level bringing the total to 52 industries. This expansion was largely attributable to recommendations of the Committee on Employment Statistics commissioned by President Warren G. Harding and a subsequent increase in Congressional funding.
1927: Sample size grew to 11,000, and employment and total payroll data were published for 54 manufacturing industries nationally.
1928: Seven additional states signed cooperative agreements with BLS.
1929: Total sample size expanded to 34,000 establishments. Publication started for 11 nonmanufacturing industries. These advances followed recommendations of committees appointed by the American Statistical Association in 1922–23.
1930: A new 6-month shuttle schedule (collection form) replaced the single-month schedule.
1931: The Advisory Committee on Employment Statistics commissioned by President Herbert Hoover issued its report, urging benchmarking to Census of Manufacturing, addition of city and state data, collection of hours data in manufacturing, and improved sample coverage for new industries and new firms. Congress appropriated additional funding to these expansions beginning in 1932.
1932: BLS publishes data on monthly employment and payroll for all 48 states, the District of Columbia and 13 cities.
1932: Additional published series were added (38 industries in manufacturing and 15 in nonmanufacturing), and collection of data on production worker hours began.
1932: Federal civil service employment reporting began.
1932: For production workers in 15 manufacturing industries, BLS published hours and earnings data for the first time.
1932: Sample size increased to 70,000 establishments covering 4.5 million jobs.
1935: Congress passed the Social Security Act establishing the unemployment insurance (UI) program. Employers were required to regularly report employment and payroll data to the states. BLS gained access to these data 5 years later and adopted UI employment data as its primary CES benchmark source.
1936: BLS published an estimate of nonfarm wage and salary employment for the first time at a national level.
1936: Twelve states signed cooperative agreements with BLS.
1938: A new 12-month shuttle schedule (collection form) replaced the 6-month shuttle schedule.
1940: CES-N published its first UI-based benchmark.
1940: Total nonfarm employment data for 48 states and the District of Columbia were published for the first time. States with cooperative agreements began publishing their own estimates. BLS regional offices prepared estimates for all other states.
1940: Sample size increased to 148,000 establishments covering 8.4 million workers.
1947: The 1945 Standard Industrial Classification (SIC) system was implemented for manufacturing industries.
1947: Sample coverage increased to cover 180 industries—covering 148,000 establishments and 12.5 million jobs.
1949: BLS signed cooperative agreements with 48 states for the first time. States began preparing their own estimates.
1949: First comprehensive UI-based benchmark was completed and published. Newly expanded UI-based data made it possible to benchmark all employment levels of industry detail directly.
1954: Publication of the monthly periodical Employment and Earnings began.
1958: SIC was adopted for all industry series published by BLS. Benchmarks by size became available, allowing for size class stratification of the sample.
1962: The President’s Commission to Appraise Employment and Unemployment Statistics (also known as the Gordon Committee) was created by President John F. Kennedy following charges of political manipulation in a Reader’s Digest article. The commission’s report, “Measuring Employment and Unemployment,” dismissed all charges of political manipulation. It recommended improvements in benchmark data, improved sampling techniques, and publication of estimates of standard error.
1964: A major redesign of the CES sampling method was issued as a technical memorandum to the cooperating state workforce agencies (SWAs). It was the first attempt to put a structure and set of requirements around the CES sample needed for national estimates. Sampling requirements were in the form of sampling ratios by industry and size class. The requirements were designed to approximate a sampling proportional to size design but did not constitute a probability sample.
1964: CES started collection of women workers and hours and earnings for production and nonsupervisory workers in all private-sector industries.
1965: Employment, hours, and earnings series were published monthly for the nation, states, the District of Columbia., and 170 metropolitan areas.
1966: Implementation of the 1964 sample design was completed.
1966: The difference link and taper became the standard estimator for hours and earnings.
1967: First publication of average earnings at the total private level was issued. This was made possible by a sample expansion begun in 1964 in nonmanufacturing industries.
1968: Bias adjustment factors to account for business births not captured by the sample was started.
1975: Sample increased to 160,000 establishments.
1978: CES series were converted from the 1967 SIC to the 1972 SIC with some limited historical reconstruction.
1978: A significant expansion of UI coverage to most of state and local government caused a large benchmark revision but improved employment series for these industries.
1979: The National Commission on Employment and Unemployment Statistics (also known as the Levitan Committee) issued its report “Counting the Labor Force” with recommendations for many employment and unemployment statistics programs. For the CES, it criticized the sample design and other statistical underpinnings of the program but urged caution in making changes that might disrupt the time series and other series that used CES as an input. It cited inadequate sample size, lack of quality control, lack of a consistent annual benchmarking process, weak bias factor adjustment techniques, and poor documentation as primary concerns. It recommended expanded publication in the service sector industries, expansion of publication to include all MSAs, addition of supplements, collection of hours worked data, and collection of data on part-time workers.
1980: CES-SA began publishing over 24,000 state and metropolitan series on LABSTAT.
1982: CES-N began the practice of benchmarking every year to a UI-based benchmark. Prior to 1982, benchmarking was not done consistently every year, mainly due to delays in UI data processing. This accomplishment fulfilled one of the Levitan Committee recommendations.
1982: Cooperative agreements were signed with all 50 states, the District of Columbia., Puerto Rico, and the U.S. Virgin Islands.
1983: BLS was given full financial responsibility for the labor market information system; UI trust fund money was transferred from the Employment and Training Administration (ETA), and BLS became responsible for administering LMI grants to state workforce agencies for BLS programs. Previously states had received some of their CES program funding directly from BLS, and the remainder of their CES funding from ETA. This transfer of funding for all BLS federal-state programs was a specific recommendation of the Levitan Committee.
1983: Sample increased to 190,000 establishments.
1984: CES addressed some of the Levitan Committee recommendations by expanding the sample size, issuing a comprehensive revision to the state operations manual and improving training for state and regional staff. In addition, CES began experimenting with a new data collection method, computer assisted telephone interviewing (CATI), as a precursor to a probability sample redesign, which would require better solicitation and collection methods.
1985–93: A comprehensive CES-SA processing system known as Automated Current Employment System (ACES) was developed by the Iowa state work agency and gradually exported to all states.
1986: CES began touchtone data entry (TDE) sample collection for selected reporters.
1988: CES tested voice recognition (VR) as a sample collection method for selected reporters. It did not prove feasible and was stopped after the test period.
1989: Sample increased to 425,000 establishments covering nearly one-third of universe employment.
1990: CES converted estimates from the 1972 SIC to the 1987 SIC basis including extensive historical reconstruction work.
1991–96: CES received additional permanent funding through the federal economic indicators (FEI) initiative. It was used to fund conversion to automated data collection techniques and expand the sample size and publication detail in the service-producing sector.
1993: Statewide employment series first published on a seasonally adjusted basis.
1993: The U.S. Government Accountability Office (GAO) conducted a review of the CES and Quarterly Census of Wages of Employment and Wages (QCEW) programs after charges of political manipulation followed an unusually large 1991 benchmark revision at the national level and in many states—most notably California. The focus was on CES series as an input to the gross domestic product estimates. GAO issued a report titled “Gross Domestic Product: No evidence of manipulation in first quarter 1991 estimates” dismissing the charges.
1993: CES undertook an extensive Response Analysis Survey (RAS) of CES and QCEW microdata reporting via payroll processing firms, payroll software providers, and individual respondents. This survey was a follow-up from the large 1991 benchmark revision that had been traced to reporting problems by payroll processing firms. The RAS identified minor issues in reporting in both the CES and QCEW.
1994: The two-step seasonal adjustment procedure was adopted to control for the effects of differing seasonal patterns between current monthly sample-based estimates and final benchmarked employment series.
1994: CES began collecting sample data via electronic data interchange (EDI) for selected large firms which had large number of worksites.
1994: A special panel of the American Statistical Association, requested by BLS, issued its report titled “A research agenda to guide and improve the CES program,” with 26 comprehensive recommendations for CES and ES-202 (now known as QCEW) program improvement. The genesis of the review was an unusually large national benchmark revision for 1991.
1995: BLS announced a wide-ranging redesign of the CES program based largely on the 1994 American Statistical Association panel’s recommendations. It was conducted in three sequential phases: research, production testing, and implementation.
1995: CES began FAX sample data collection for selected reporters.
1995–97: As part of the probability sample redesign effort, BLS completed pilot tests of sampling new business births more rapidly, from first-cut UI registration files. Results indicated this was not feasible and efforts were instead focused on modeling for new business births. In addition, BLS worked with experts from outside the Bureau to complete a probability-based sample design and explore alternative estimators. Other outside experts worked with BLS to research improvements to solicitation and collection methods that would help raise survey response rates.
1996: CES began sample data collection by email and internet for selected reporters.
1997: The sample redesign research phase was completed; BLS began production test phase of the redesign. BLS took over sample solicitation from states and newly selected probability sample units were solicited by the three BLS regional collection centers. Test estimates were produced from the new sample and assessed for benchmark revision and variance properties.
1997: Research was conducted to test of the feasibility of estimating federal employment using standard benchmark and sample-based estimators rather than the then-current method of using end-of-month counts from the Office of Personnel Management.
1997: ACES service center began operating as the primary estimation system for state and area series. Some states continued to utilize individual state installations of ACES.
1998: A pilot test of collecting all employee hours and earnings was completed. Though successful, a decision was made to postpone this initiative until after the CES redesign was complete, to avoid overwhelming the program resources.
2000–03: Implementation phase of the redesign began. A new probability-based sample design and resultant estimates began publication on an SIC major division basis with the wholesale trade industry. Mining, construction, and manufacturing followed in 2001, transportation and public utilities, and finance, insurance, and real estate, and retail trade in 2002 and services in 2003. Government was not converted due to the already existing very high sample coverage.
2000: CES-SA introduced small domain model (SDM) estimates for state and area level industries with small sample sizes.
2003: CES-SA fully implemented the program redesign. Changes included implementation of a probability sample design, introduction of business birth-death modeling, conversion from an SIC to a 2002 North American Industry Classification System (NAICS) basis, and the introduction of a new method for estimating federal government employment. Historical time series were reconstructed based on NAICS.
2005: CES expanded its geographic coverage to publish all Office of Management and Budget-designated metropolitan statistical areas (MSAs) and metropolitan divisions. Previously about three-quarters of the MSAs were published.
2005: Following Hurricane Katrina, CES implemented temporary changes to its monthly estimation techniques for national estimates and affected states and metropolitan areas to better capture the full impact from the storm. Prior user notification and full documentation on the changes were provided on the BLS website.
2006: BLS began collecting data on all employee payroll and hours.
2007: MSA employment series first published on a seasonally adjusted basis.
2007: CES began including tips in its payroll definition for the private service-providing industries, excluding education.
2007: Nearly all remaining state workforce agencies sample data collection was transferred to BLS.
2008: CES-SA began publication of experimental all employee hours and earnings series.
2008: CES-SA ceased publication of the 65 smallest metropolitan areas to accommodate a reduction in funding to BLS. Publication of these areas was resumed in 2009.
2008: BLS completed an extensive response analysis survey to explore sources of reporting error in the CES and QCEW microdata and their possible contributions to benchmark error and differing seasonal patterns between the two programs. Results contributed to CES decision not to change benchmark methods or frequencies until error sources in the QCEW could be better understood.
2008: CES-SA updated data series from NAICS 2002 to NAICS 2007.
2009: Adoption of a “top-down” estimation approach, in which state and area employment is estimated at the super sector level and is used to control the sum of detailed industry estimates, following large differences between sum-of-state and national employment during the 2007–09 recession. The top-down approach included a robust estimation process to reduce the weight of highly influential reports and a Fay-Herriot (FH) model for statewide super sectors with small sample size.
2009–11: Introduction of ACES web and transition away from ACES mainframe system for producing state and area estimates
2010: CES-SA began publication of official all employee hours and earnings.
2011: CES-SA introduced a special seasonal adjustment technique to control for the variation in intervals between surveys (also known as the 4- versus 5-week effect).
2011: Responsibility for producing state and area employment estimates was transferred from state workforce agencies to BLS to reduce costs and standardize estimation methodology.
2011: CES began updating the business birth-death model on a quarterly basis rather than the previous procedure of annual updates.
2012: CES-SA converted to NAICS 2012 which resulted to minor content changes within the manufacturing and the retail trade sectors, as well as minor coding changes within the utilities and the leisure and hospitality sectors. Several industry titles and descriptions were also updated.
2013: CES-SA introduced a modified Fay-Herriot (MFH) model for estimating employment in select metropolitan areas with small sample sizes.
2014: CES changed from annual to quarterly sample rotation to improve response rates and estimates by more quickly phasing in newly solicited sample.
2014: CES completed the implementation of a probability sample redesign for all private industries in Puerto Rico.
2015: CES-SA transitioned from using X‑12‑ARIMA to X‑13ARIMA‑SEATS to produce seasonally adjusted series and forecasts of birth-death residuals.
2017: Introduction of generalized variance functions (GVF) to estimate sampling variance
2018: CES-SA adopted concurrent seasonal adjustment.
2018: CES-SA updated data series to reflect changes from the 2012 NAICS to 2017 NAICS
2020–21: CES-SA implemented temporary adjustments beginning with March 2020 estimates to better capture unprecedented employment changes resulting from the COVID-19 pandemic. These adjustments were discontinued in October 2021.
2020: Diffusion indexes of state and area employment were first published.
2022: CES-SA introduced a third-generation small area model (SAM Gen3), replacing existing models (SDM, FH, FH2) used for estimating state and area employment cells with small sample sizes.
2023: CES-SA updated to NAICS 2022 which resulted in major changes to the retail trade sector codes, as well as minor coding changes within the information and financial services sectors. Several industry titles and descriptions were also updated.
Last Modified Date: October 27, 2023
United States Department of Labor
