*** BE SUBMITTED BY EMAIL TO NSCW@FAMILIESANDWORK.ORG ***
WE URGE USERS OF THE PUBLIC-USE FILES TO ADVISE US OF ANY ERRORS THEY ENCOUNTER THAT MIGHT REQUIRE CORRECTIONS IN THE FILES. NSCW FUNDERS 2008 NSCW Alfred P. Sloan Foundation (lead sponsor) IBM Corporation
The many survey respondents who donated their honoraria in support of the NSCW
2002 NSCW Alfred P. Sloan Foundation (lead sponsor) IBM Corporation
Johnson & Johnson
The Ford Foundation
Salt River Project
The many survey respondents who donated their honoraria in support of the NSCW. 1997 NSCW KPMG Peat Marwick LLP (lead sponsor) Alfred P. Sloan Foundation
I. NATIONAL STUDY BACKGROUND ............................................1
II. TECHNICAL INFORMATION ...............................................2
III. CONTENTS OF THE NSCW PUBLIC-USE FILES CD-ROM VERSION................9
IV. COPYING/OPENING/READING FILES ON THE CD-ROM ........................12
V. SAMPLE WEIGHTING ....................................................13
VI. COMPARING DATA FROM DIFFERENT SURVEY YEARS .........................15
VII. IRB COMPLIANCE.....................................................18 I. NATIONAL STUDY BACKGROUND
Beginning in 1969, the Department of Labor funded three national studies of the United States workforce as part of the Quality of Employment Survey (QES). The last survey in this series, which was conducted in 1977, marked the first time that research on a large, representative sample of U.S. workers collected information about not only the work lives of employees, but their personal lives as well.1 When the QES program was halted for a variety of reasons in 1977, a 15-year gap ensued during which there were small-scale studies of life on and off the job, but no large-scale nationally representative studies were undertaken.
In 1990, the Families and Work Institute (FWI) obtained private support for the National Study of the Changing Workforce (NSCW) as an ongoing research program of the Institute. The Institute’s program is more explicit and comprehensive than the QES in addressing issues related to both work and personal life. It also reflects a strong business perspective, in addition to the broader social and economic perspectives that shaped the QES.
The NSCW surveys representative samples of the nation’s labor force every five - six years, with findings on important and timely issues released during the intervening years through Institute reports, publication in academic journals and books, media coverage, and presentations to audiences of private- and public-sector decision-makers. The first NSCW survey was conducted in 1992;2 the second, in 1997;3 the third, in 2002;4and the fourth, in 2008.5 The 2008 NSCW survey includes many questions used in the 1977 Quality of Employment Survey, providing a unique opportunity to examine workforce issues through the lens of history with 30-year comparisons to the 1977 QES as well as comparisons with earlier data from the 1992, 1997,and 2002 NSCW surveys.
1 Quinn, R.P. & Staines, G.L. (1979). The 1977 Quality of Employment Survey. Ann Arbor, MI: Institute for Social Research, University of Michigan.
2Galinsky, E., Bond, J.T., & Friedman, D.E. (1993). The Changing Workforce: Highlights of the National Study. New York: Families and Work Institute.
3 Bond, J.T., Galinsky, E., & Swanberg, J.E. (1998). The 1997 National Study of the Changing Workforce. New York: Families and Work Institute.
4 Bond, J.T. with Thompson, C., Galinsky, E., & Prottas, D. (2003). Highlights of the 2002 National Study of the Changing Workforce. New York: Families and Work Institute.
5 No summary report was published for the 2008 NSCW. Rather FWI researchers have prepared and continue to develop topical reports of findings, which are available as free PDF downloads on FWI’s Web site: familiesandwork.org. Although data collection for the survey began in the last quarter of 2007, most data were collected in 2008. Thus, we refer to the survey as the 2008 NSCW.
Sample sizes are large enough to support analyses of many subgroups of interest: 2008 NSCW total sample = 3,502 (2,769 wage and salaried workers); 2002 NSCW total sample = 3,504 (2,810 wage and salaried workers); 1997 NSCW total sample = 3,552 (2,877 wage and salaried workers); 1992 NSCW total sample = 3,718 (2,958 wage and salaried workers); 1977 QES total sample = 1,515 (1,298 wage and salaried workers).
II. TECHNICAL INFORMATION: NSCW 2008, 2002, 1997, 1992, AND QES 1977
Table 1 provides brief descriptions of the five surveys/samples discussed here.
Table 1: Overview of the Surveys/Samples Referenced in this Guide
* The 1977 Quality of Employment Survey cross-section data file is not included in the Public-Use Files CD-ROM but is available from the Inter-University Consortium for Political and Social Research at the University of Michigan. These data are also archived at most major universities. The Public-Use Files CD-ROM does, however, contain SPSS syntax files for creating all analytic files containing 1977 QES data described in the following pages.
** Three different weighting variables were developed for the 2008 NSCW. The third weight (*WT3) is described in this table. See full discussion, which follows, for more information.
*** Total sample weighted by SAMPWT3. 2008 National Study of the Changing Workforce
The 2008 NSCW survey was conducted by Harris Interactive, Inc. using a questionnaire developed by the Families and Work Institute. A total of 3,502 interviews were completed with a nationwide cross-section of employed adults between November 12, 2007 and April 20, 2008. Interviews, which averaged 50 minutes in length (47 minutes for substantive questions and 3 minutes for the screener), were conducted by telephone using a computer-assisted telephone interviewing (CATI) system. Coding of open-ended responses was done by interviewers, with the exception of occupation and industry which were coded by the U.S. Bureau of the Census using 1990 three-digit occupation (SOC) and industry (SIC) classifications. 1990 Census classifications were used in 2008 to facilitate comparisons with the 1997 and 2002 NSCW surveys, which used the same codes.
Telephone calls were made to a stratified (by region) unclustered random probability sample generated by random-digit-dial methods. Up to 60 calls were made to each telephone number that appeared to represent a potentially eligible household — busy signal obtained, voice mail pickup, answered by a non-eligible with some indication of a potential eligible in household, or answered by a potential eligible who wanted a callback. When eligibles were identified and requested callbacks, additional calls were made. If twenty-five consecutive calls were made to a number where there was no answer, busy signal, or other dialing outcome, these numbers were considered non-residential, non-working numbers, or non-voice-communication numbers. Cash incentives were offered to increase cooperation. Incentives began at $25 and increased to $50 if necessary to convert initial refusals. Three to five attempts were made to convert each refusal.
Despite the fact that the level of effort of 2008 interviewers went substantially beyond the efforts made in 2002, 1997, and 1992, the overall response rate was only slightly higher, indicating that it has become significantly more difficult to complete telephone interviews in recent years. Our experience has been widely confirmed by other researchers.
Sample eligibility was limited to people who 1) worked at a paid job or operated an income-producing business, 2) were 18 years or older, 3) were employed in the civilian labor force, 4) resided in the contiguous 48 states, and 5) lived in a non-institutional residence—i.e., household—with a telephone. In households with more than one eligible person, one was randomly selected to be interviewed.
Of the total 42,000 telephone numbers called, 24,115 were found to be non-residential or non-working numbers and 6,970 were determined to be ineligible residences (1,389 because no one spoke English or Spanish well enough to be interviewed). Of the remaining telephone numbers, 3,547 were determined to represent eligible households, and interviews were completed for 3,502 of these——a completion rate of 99 percent. However, eligibility or ineligibility could not be determined in the remaining 7,368 cases. Among those contacts for which eligibility could be determined, the eligibility ratio was 0.3886 [3,547/(3,547+5,581)]. Thus, we estimate that potentially 38.86 percent of the 7,368 cases for which eligibility could not be determined——2,863 cases in all——might have been eligible households. Dividing the number of completed interviews (3,502) by the number of known eligibles (3,547) plus the number of estimated eligibles (2,863) yields an overall response rate of 54.6 percent for potentially eligible households. [This method of response rate calculation follows the conservative CASRO and AAPOR recommendations.]
Three weights were developed for the 2008 NSCW sample:
Weight 1: The sample was weighted by the number of eligibles in the respondents’ households in relation to the percentage of households in the U.S. population (March 2007 Current Population Survey, CPS) with the same number of eligibles (i.e., number of employed persons 18 and older per household with any employed person 18 or older); eligible men and women in the U.S. population; and eligibles with different educational levels in the U.S. population (less than high school/GED, high school/GED, some postsecondary, 4-year college degree, post-graduate degree). This is the same weighting algorithm used for the 2002 NSCW. The average Design Effect for Weight 1 is estimated to be 1.359. Applying this Design Effect, the maximum sampling error for total sample statistics (n=3,502) is approximately +/- 1.00 percent.
Weight 2: This weighting algorithm incorporates the same variables as Weight 1 plus race/ethnicity to equal the CPS distributions (Hispanic, non-Hispanic White, non-Hispanic Black, non-Hispanic other). The average Design Effect for Weight 2 is estimated to be 1.530. Applying this Design Effect, the maximum sampling error for total sample statistics (n=3,502) is approximately +/- 1.05 percent.
Weight 3: The weighting algorithm for Weight 3 incorporates the same variables as Weight 2 plus age to equal the CPS age distribution (18 – 27, 28 – 42, 43 – 61, 62 or more). [Weight 3 provides the most accurate estimates of population statistics when generalizing from the NSCW sample.] The average Design Effect for Weight 3 is estimated to be 1.705. Applying this Design Effect, the maximum sampling error for total sample statistics (n=3,502) is approximately +/- 1.10 percent.
Of the total sample of 3,502 interviewed, 2,769 are wage and salaried workers who work for someone else, while 733 respondents work for themselves — 255 business owners who employ others and 478 independent self-employed workers who do not employ anyone else.