The ACS is a household survey developed by the Census Bureau to replace the long form of the decennial census program. The ACS is a large demographic survey collected throughout the year using mailed questionnaires, telephone interviews, and visits from Census Bureau field representatives to about 3.5 million household addresses annually. Starting in 2005, the ACS produced social, housing, and economic characteristic data for demographic groups in areas with populations of 65,000 or more. (Prior to 2005, the estimates were produced for areas with 250,000 or more population.) The ACS also accumulates sample over 5-year intervals to produce estimates for smaller geographic areas, including census tracts and block groups.
Topics covered by the ACS are virtually the same as those covered by the census long-form survey data. Estimates are produced for demographic characteristics (sex, age, relationship, households by type, race, and Hispanic or Latino ethnicity), social characteristics (school enrollment, educational attainment, marital status, fertility, grandparents caring for children, veteran status, disability status, residence one year ago, place of birth, U.S. citizenship status, year of entry, world region of birth of foreign born, language spoken at home, and ancestry), economic characteristics (employment status, commuting to work, occupation, industry, class of worker, income and benefits, and poverty status), and housing characteristics (housing occupancy, units in structure, year structure was built, number of rooms, number of bedrooms, housing tenure, year householder moved into unit, vehicles available, house heating fuel, utility costs, occupants per room, housing value, and mortgage status and costs). Data availability for geographic areas differs by population size: 1-year estimates are available for areas of population 65,000 or more, while 5-year estimates are available for all areas. (Three-year estimates had been produced for areas with populations of 20,000 or more through the 2011-13 period.)
Yes, the ACS includes questions about work and search for work so that some measures of labor force activity are available, along with the other social and economic data collected in the survey.
The Current Population Survey (CPS), jointly sponsored by BLS and the Census Bureau, is a monthly sample survey of about 60,000 households designed specifically to produce the current monthly employment and unemployment data and annual data on income and poverty for the nation. CPS monthly employment and unemployment estimates are available within a few weeks of the end of the reference period. ACS 1-year employment and unemployment estimates are available about nine months after the end of the reference year, while 5-year estimates typically are issued early in December of the following year. No monthly estimates are available from ACS.
The monthly CPS estimates are a key input to the Local Area Unemployment Statistics (LAUS) program conducted by BLS, which produces the official labor force statistics for states and local areas. Employment and unemployment estimates from the ACS and CPS can differ because the surveys use different questions, samples, and collection methods.
The CPS reference period is typically the week including the 12th of the month, with interviews being conducted the following week (typically the week including the 19th of the month). CPS data are produced and published monthly. Annual-average data are also developed at the end of the calendar year. The CPS uses a fixed reference period, as compared to the ACS, where the reference period is the week prior to when a respondent answers the survey. CPS interviews are conducted in the course of a single designated week each month, whereas respondents answer the ACS at times that vary throughout the month and year. ACS respondents are initially contacted by mail. If they do not return their survey within a month of receiving it, they are then contacted by phone. Approximately 1 in 3 households that still do not provide answers is sub-sampled for an interviewer to contact them in person in the third month.
Yes. In 2018, the numbers of people the ACS classified as “employed,” “unemployed,” and “not in the labor force” for the nation were all higher than the official CPS estimates. The ACS unemployment rate was 4.9 percent, compared to the CPS annual average of 3.9 percent.
Yes. In 2018, the numbers of employed individuals in the ACS were higher than the LAUS estimates in 31 states and lower in 19 states and the District of Columbia. The employment level differences ranged from +5.9 percent in New Jersey to -3.1 percent in Idaho, with an average difference of +0.6 percent. The ACS-based unemployment levels and rates were higher than the corresponding LAUS estimates in 46 states and the District of Columbia and lower in 4 states. The unemployment level differences ranged from +64.0 percent in Hawaii to -3.3 percent in Montana, with an average difference of +27.4 percent. The unemployment rate differences ranged from +2.3 percentage points in Mississippi to -0.2 point each in Montana, Washington, and Wyoming.
A number of factors may account for the difference in the estimates, including the following: overall questionnaire differences, differing requirements in the two surveys with regard to whether an individual is actively looking for work, and differing reference periods, modes of collection, and population controls.
The underlying concepts and definitions of labor force data developed from the LAUS program are consistent with those of the Current Population Survey (CPS). Monthly estimates for all states and the District of Columbia are produced using estimating equations based on time-series and regression techniques. These “signal-plus-noise” models combine current and historical data from the CPS, the Current Employment Statistics (CES) survey, and state unemployment insurance (UI) systems. A tiered approach to estimation is used. Model-based estimates are produced for the nine census divisions that geographically exhaust the nation. (These models use inputs only from the CPS.) The division estimates are benchmarked to the national levels of employment and unemployment each month. The benchmarked division model estimate is then used as the benchmark for the states within the division. In this manner, the sums of the state estimates of unemployment and employment equal the national estimates for these measures.
The ACS questions relating to labor force activity are less detailed than those in the CPS. For example, the ACS uses seven questions in determining labor force status, while the CPS uses sixteen. There are more detailed, probing questions in the CPS regarding employment status. In addition, the CPS information is always collected by trained interviewers and never through mail questionnaires.
Effective 2008, changes were made to the ACS questionnaire that modified and improved existing questions for several subject areas. In particular, revised labor force questions were introduced to better capture data on employment status. The modifications had the impact of increasing the estimated number of employed persons from the ACS relative to CPS and LAUS estimates.
The ACS instrument asks people if they are actively looking for work and available to take a job if offered one, but does not ask about the nature of the job search. The CPS questionnaire probes to see if people are actively looking for work (interviewing, calling contacts, etc.) versus passively looking for work (for example, looking at want ads in the newspaper). In the CPS, a person is unemployed only if that person responded affirmatively to having engaged in one or more active methods of job search.
ACS responses can relate to any weekly period throughout the year and reflect different economic events. Respondents can choose to delay completion of the ACS form.
ACS data are collected over a range of time periods. In the ACS, the reference is to activity in the “last week” whenever the respondent fills out the survey. In the CPS, the reference period generally is fixed for the calendar week including the 12th of the month. A varying reference week and time of data collection could be particularly problematic for shorter, transitory statuses or activities that could be influenced by seasonal variation. Unemployment, for example, is a state that is subject to both seasonal and cyclical variability.
The mode of collecting data also may affect the labor force estimates. All CPS interviews are conducted through personal visits or telephone calls by Census Bureau field representatives using laptop computers for data entry. ACS data are collected primarily by mail using “paper and pencil” questionnaires, with telephone and personal visit collection used as follow-up to mail nonresponse. Data collected using paper forms do not have interviewers assisting respondents in interpreting questions.
Both the ACS and CPS are sample surveys used to make estimates for a larger population. Each person in the survey represents a larger number of similar individuals in the population. To do this, both surveys utilize population estimates produced by the Population Estimates Program (PEP) of the Census Bureau. Each year, PEP publishes population estimates by demographic characteristics (age, sex, race, and Hispanic or Latino ethnicity) for the nation, states, and counties. The reference point for population estimates is July 1st.
Although the programmatic source of the population controls is the same, the ACS, CPS, and LAUS programs each incorporate these PEP data in different ways. For the 1-year datasets, the ACS uses the latest post-censal estimates vintage, such that the controls for the 1-year estimates for 2018 were consistent with the PEP estimates for July 1, 2018, from Vintage 2018. For the national data from the CPS, there is a lag between the reference year and the underlying PEP-based controls, with a break between vintages in January of each year. (So, for example, the CPS national controls for December 2018 reflect Vintage 2017, while the CPS national controls for January 2019 reflect Vintage 2018.) Neither the ACS (at any level of geography) nor the CPS national estimates are subject to subsequent population recontrolling. A given PEP vintage, however, includes both new estimates for the latest July-1 reference point and revised estimates back to the decennial base period for post-censal estimation, which currently is April 1, 2010. The LAUS program, per its annual model re-estimation, recontrols the CPS employment and unemployment inputs for the prior five years, based on the latest PEP vintage available at the time. Hence, for the cycle of LAUS model re-estimation completed in February 2019, the CPS inputs to the state models were recontrolled from January 2014 forward based on Vintage 2018 for states. Only for the latest year, then, do the ACS 1-year estimates and the LAUS annual averages reflect the same control basis, during the 6-month window between the release of the ACS 1-year estimates and the next LAUS model re-estimation cycle.
Below the state level, LAUS data do not reflect independent controlling to PEP estimates. Estimates from the ACS, on the other hand, are controlled down to the county level.
Yes. With the implementation of the 2015 LAUS program redesign, several substate estimation inputs that historically had come from the decennial long-form survey, most recently for Census 2000, were updated with inputs developed from ACS 5-year estimates. Information on the 2015 LAUS program redesign is available at www.bls.gov/lau/lauschanges2015.htm
Last Modified Date: December 6, 2019