After interviews for the Current Population Survey (CPS) are conducted, the U.S. Census Bureau processes the raw data files (as submitted by interviewers) to create a microdata file that can be used to produce estimates. This processing includes removing all personally identifiable information, assigning standardized occupation and industry classifications, editing the data for completeness and consistency, and creating new data elements based on responses to multiple survey questions.
Data from field interviewers and centralized call centers are transmitted to Census Bureau headquarters daily during the survey interview period through secure electronic communications. When the Census Bureau’s processing activities are completed, the microdata file is securely transferred to the U.S. Bureau of Labor Statistics (BLS) for estimation of labor force statistics.
For additional details, including imputation methods, see chapters 8 and 9 of Design and methodology: Current Population Survey, Technical Paper 66.
In order to produce national labor force estimates from the survey microdata, a statistical weight for each person in the sample is developed. A series of procedures adjusts the weights for sample records in order to ensure that these sample-based estimates of the population match independent population controls for a number of geographic and demographic subgroups. On average, a person in the CPS sample represents about 2,500 people in the population. The estimation methodology also includes adjustments to account for people who do not respond to the survey and to minimize the error range of survey estimates.
The estimation methodology for the CPS is a highly complex set of statistical procedures. A complete discussion of the estimation process can be found in chapter 10 of Design and methodology: Current Population Survey, Technical Paper 66.
Labor force levels, employment, unemployment, and other labor market measures fluctuate sharply over the course of a year because of seasonal events such as weather, major holidays, and the opening and closing of schools. Seasonal adjustment is a statistical procedure used to remove seasonal fluctuations from data series, thereby making it easier to observe cyclical and other economic trends. A wide range of seasonally adjusted labor market measures is available from the CPS. However, not all measures are available on a seasonally adjusted basis.
Seasonally adjusted CPS data for the current year are produced with a technique known as concurrent adjustment. Under this practice, the current month’s seasonally adjusted estimate is computed with the use of all relevant original data up to and including data for the current month. Revisions to estimates for previous months, however, are postponed until the end of the calendar year, at which time BLS re-estimates the seasonal adjustment factors for CPS series in order to include the latest 12 months of data in the estimation process. On the basis of this annual reestimation, the most recent 5 years of seasonally adjusted CPS data are subject to revision. The new seasonal factors and the revised seasonally adjusted series are introduced with the publication of December estimates in January of each year. Data series that are not seasonally adjusted, including annual averages, are not part of this revision process. Articles describing the annual seasonal adjustment process over time are available in the CPS technical documentation.
Population controls are independent estimates of the population that are used to weight the CPS sample results so that estimates reflect the civilian noninstitutional population ages 16 and older. The Census Bureau develops the CPS population controls, which are based on decennial census population counts, supplemented with birth and death data and estimates of net international migration.
The Census Bureau reviews and adjusts the population estimates every year. BLS introduces the Census Bureau’s annual population control adjustments into CPS estimates for January. The adjustments may increase or decrease estimates of the population level, depending on whether the latest information indicates that the population estimates have trended high or low. Articles describing the population control adjustments over time are available in the CPS technical documentation.
Two types of error are possible in an estimate based on a sample survey: sampling error and nonsampling error.
Sampling error. When a sample, rather than an entire population, is surveyed, estimates differ from the true population values that they represent. The component of this difference that occurs because samples differ by chance is known as sampling error, and its variability is measured by the standard error of the estimate.
A sample estimate and its estimated standard error can be used to construct an approximate confidence interval, or range of error. Standard errors and confidence intervals are used to help determine if the difference between two estimates is statistically significant or just part of the sampling variability associated with survey estimates.
There is about a 90-percent chance, or level of confidence, that an estimate based on a sample will differ by no more than 1.645 standard errors from the true population value because of sampling error. BLS analyses generally are conducted at the 90-percent level of confidence.
Estimated standard errors for key CPS data series published by BLS are available in the CPS technical documentation.
Nonsampling error. Nonsampling error is due to factors that are not related to sample selection. This type of error in surveys can be attributed to many sources, including
The full extent of nonsampling error is unknown.
In addition to containing the estimated standard errors for key CPS data series published by BLS, the CPS technical documentation includes more information about the reliability of CPS estimates, along with guidance on calculating approximate standard errors and confidence intervals.