The GSS has employed a number of different sampling methodologies over time. Since 1975, the GSS used full-probability sampling of households. The 1982 and 1987 GSSs applied this methodology but also included oversamples of black respondents. Beginning in 2004, the GSS began to use a two-stage sub-sampling design. For this reason, users must determine which weight is appropriate based on the years and population being analyzed.
WTSSALL is suitable for most common analyses across all years. However, below are special situations in which the use of a different weight should be considered.
| Sample Year | Exception | Appropriate Weight |
|---|---|---|
| 1978, 1980, 1982-1985 | A certain set of variables in these samples are based on questions that only appeared on one form and as a result were only asked of a subset of participants. Therefore, users must use a FORMWT when analyzing variables that appear on this list: Appendix P. | FORMWT |
| 1982; 1987 | The 1982 and 1987 GSSs included oversamples of black respondents. To adjust statistical results for this oversampling, users should exclude cases in the black oversamples (codes 4, 5, and 7 on variable SAMPLE). Alternatively, if only analyzing the 1982 or 1987 samples users can weight statistical results using the OVERSAMP weight. | OVERSAMP |
| 2004-2014 | Beginning in 2004, the GSS began to use a two-stage sub-sampling design for nonresponse. To obtain weighted estimates adjusted for this non-response, users should use the WTSSNR weight in these years. | WTSSNR |
For a thorough discussion of the GSS Sample Design and weight construction, see the Codebook’s Appendix A: Sampling Design & Weighting.
For users interested in generating robust standard errors, please see Calculating Design-Corrected Standard Errors for the General Social Survey.
