Technical Notes to the Current Population Survey

The estimates of secondary artist employment presented in Arts Data Profile #3 were produced by the NEA's Office of Research & Analysis using the Current Population Survey (CPS), public file. The annual figures reported here were calculated by averaging results for each month in the 2013 CPS public data set. Due to factors such as "top coding," estimates from the public CPS file will differ somewhat from official CPS estimates calculated and reported by the Bureau of Labor Statistics (BLS).

The CPS is a nationwide survey of 60,000 households, and is conducted monthly by the U.S. Census Bureau for the Bureau of Labor Statistics. The CPS collects information about the demographic characteristics and employment status of the civilian non-institutionalized population 16 years of age and older.

Since January 1994, employed CPS respondents have been asked if they have more than one job. In the CPS, a multiple jobholder is defined as workers who (1) had a job as a wage and salary worker with two or more employers; (2) combined a wage and salary job with self-employment; or (3) combined a wage and salary job with unpaid family work. The definition of multiple jobholders excludes self-employed persons with multiple businesses and persons who hold multiple unpaid jobs as family workers.

As reported by the BLS, more than 90 percent of multiple jobholders hold two jobs (with the remainder holding three or more jobs). The primary job is defined as one at which the greatest number of hours were worked; the secondary job is the one at which the second greatest number of hours were worked.

Each month, additional questions on the industry and occupation of the secondary job are asked of one-quarter of the CPS respondents—those in the outgoing rotation group, who are respondents in their fourth or eighth month of the sample.

Using CPS Weights to Measure Multiple Jobholding

Given that secondary jobholding is collected from outgoing CPS respondents, it is important that users of the CPS public file apply the outgoing rotation weight to CPS multiple jobholding variables. Within the CPS public file, that weight is labeled "PWORWGT."

In addition, CPS weights are fractional, or probability weights. In order to use CPS weights as the more conventional frequency weights, data users can multiply CPS weights by 10,000. Doing so, however, results in level estimates (e.g., number of secondary jobholders) with four implied decimal places.

Alternatively, users of statistical software packages can designate CPS weights as probability weights. In Stata, for example, "pweight" may be designated and used with "svy" code, such as "svy: tabulate."

Historical Comparability of CPS Estimates

Historical comparability of estimates from the CPS is affected by revisions to occupational and industry classification and by revisions to population controls used to weight CPS sample results.

While the 11 artist occupations presented in this Profile were unaffected by changes in the CPS occupational classifications, CPS "level" estimates, including primary and secondary employment, are influenced by revisions to CPS population controls.[1]

Population controls, which are used to weight CPS sample estimates to the U.S. population, are revised annually by the U.S. Census Bureau to reflect results from the decennial population census and from population changes indicated by administrative records (e.g., birth and death records).

Because level shifts in CPS labor force and employment series can reflect revisions to population controls, it is difficult for CPS data users to evaluate level changes over time.

Revised population controls, however, do not affect rates or percentages derived from the CPS. Consequently, rates such as the unemployment rate or the share of workers holding second jobs from one period to the next are not influenced by these revisions.

High Margin of Error for Estimates of Secondary Artists

Due to the small population size of secondary artists, estimates of their numbers and characteristics result in large margin of errors.

Caution should be exercised when considering fewer than 3,000-4,000 weighted CPS respondents. On average, weighted respondents in that range represent roughly one or two un-weighted CPS respondents.

For technical information on the CPS, please see the BLS' documentation:

Labor Force Statistics from the Current Population Survey, Technical Documentation

Notes on Using Current Population (CPS) Subnational Data