How were the variables for the Key Trends feature selected?
Trend Variables
Module topics were selected based on information included in previously developed GSS Trend Reports, as those
were vetted by and deemed topics of interest by the GSS research community. Coupled with these reports and an
extensive review of all GSS data, variables were selected based on those which were: relevant to module topics,
spanned at least three years, and included respondent opinion data. Variables with fewer than three years of
data and/or variables that did not pertain to module topics were excluded. Additionally, several variables
required multiple adjustments in order to account for measurement variation over time and were thus excluded.
Response categories were retained for most variables. Any variable recodes or adjustments are detailed in the footnotes.
Future feature releases could include data on additional module topics or International Social Survey Program (ISSP) data.
Breakdown Variables
Trends may be viewed for all GSS respondents or broken down by subgroup. The list of breakdown variables currently
available in the Key Trends feature includes:
- Age (AGE)
- Sex (SEX)
- Race (RACE)
- Hispanic specified (HISPANIC)
- Highest educational degree (DEGREE)
- Labor force status (WRKSTAT)
- Subjective class identification (CLASS)
- Marital status (MARITAL)
- Political affiliation (PARTYID)
- Region of interview (REGION)
- Citizenship status (USCITZN/FUCITZN)
- Condition of health (HEALTH)
With the exception of Hispanic specified, citizenship status, and health, these demographic items are available for all GSS data collection years.
Note that some GSS variables have been recoded in order to produce consistent time series. “Don’t Know” responses
to breakdown variables were excluded from calculations.
Several demographic variables, such as income, required multiple adjustments in order to account for measurement
variation over time and were thus excluded. Other breakdown variables, such as subjective class identification (CLASS),
highest degree (DEGREE), and labor force status (WRKSTAT) may be used as indicators of socioeconomic status.
Future feature releases could include additional breakdown variables.
How do I know which GSS variable was used to create the trend visualization?
The GSS variables used to create the trend visualization are listed beneath the chart, along with the variable description and the original question text. In most cases, the trend data were produced directly from the GSS variable. In some cases, GSS variables have been recoded, combined, or adjusted in order to produce a consistent time series.
Please refer to the footnotes in each chart for details about variable recodes and adjustments.
Are the trend data weighted?
The default weight
WTSSALL
was applied for all years. Oversampled black respondents (SAMPLE=4,5,7) were excluded to adjust for the oversampling procedures in 1982 and 1987.
Please refer to the footnotes for each trend chart, as these may specify additional weighting considerations
(e.g., use of FORMWT or WTSSNR) that may be applied for more in-depth analyses. See this page
for more detail about GSS weighting procedures.
Calculations were performed using SAS Software. Due to slight variations in how different statistical software
calculate weight, the estimates presented in this report may differ slightly from reported percentages in other GSS publications.
For years 1975 onward, variables
vstrat
(for stratum) and
vpsu
(for primary sampling units) were applied to generate design-corrected standard errors. These complex standard errors are only valid from 1975 onward.
See this page for more information.
The standard errors are available in the downloadable data files and may also be viewed within the visualization
by hovering over the data point or segment of interest.
Were any data excluded?
Unless otherwise specified in the chart footnote, “No answer” and “Not applicable” values were excluded from calculations.
“Don’t Know” responses were excluded from calculations for breakdown variables and included in calculations for trend variables.
Please refer to the footnotes in each chart for additional details about exclusions.
How should I interpret the trends for “Don’t Know” or “Other” responses?
Some response options, such as “Don’t Know” or “Other,” aren’t selected often enough to provide consistent,
reliable data; this can produce misleading trend lines.
Therefore, when viewing these and other trends, keep the following considerations in mind:
-
The context in which the questionnaire was administered:
In most cases, the “Don’t Know” or “Other” option is not read aloud to the respondent and is only therefore
selected if the respondent volunteers this response. For more information on administration procedures,
refer to the GSS Cumulative Codebook and
GSS Questionnaires.
-
It is more difficult to draw accurate conclusions when there are few observations:
One way of measuring the accuracy of a particular estimate is by referring to the standard error value.
Standard error measures the extent to which an estimate from a sample is likely to deviate from the true
population. Data points with high standard error values relative to the percent estimate are less reliable.
Take caution when interpreting estimates with a relative standard error greater than 30%. To view the
standard error value associated with a data point, hover your mouse over the data point or download the data.
