Following is a quick walkthrough of how to use the General Social Survey without the aid of any outside statistical packages.
We’ll enter two variables to cross-tabulate, “wordsum” and “degree”. For contemporary relevance, we will only look at responses from the year 2000 to the year 2018. To do that, we’ll enter “year(2000-2018)”. Additionally, we’ll check the hyperlinked box “Summary statistics” a little further down the page. The page should look like this:
Now either hit enter on your keyboard or click the “Run the Table” button below the data entry field box. The new page that opens contains our results. On the left (the row), running from top to bottom, we have respondent wordsum scores. Along the top (the column), running from left to right, we have the highest educational degree respondents have earned.
Let’s focus on a wordsum score of 10. We see that .3% of respondents with less than a high school education scored a 10 on the wordsum test, while 13.4% of those with a graduate degree did. The green circle shows what percentage (3.6%) of the entire respondent pool scored a 10. The non-bolded numbers below each of the bolded numbers–the bolded numbers actually show percentages–reveal the weighted number of actual respondents who fall into each cell. So 5.1 respondents with less than a high school education scores a 10.
How do 5.1 people do something? They don’t, unless the results are adjusted to correct for under-/over-sampling as they are in this case! Just above the “Summary statistics” box we checked, notice two boxes under the “N of cases to display:” hyperlink. The GSS will default to weighted responses, but that can be toggled to unweighted responses if preferred.
A bit further down shows the average (mean) wordsum scores among respondents by degree attained.
Respondents with less than a high school education earned an average score of 4.36 on the wordsum test. Respondents with a graduate degree averaged a 7.52. The average for all respondents was 5.99.
One standard deviation in wordsum score for the entire respondent pool is 2.00. There is thus a 1.58 standard deviation difference (7.52 – 4.36 = 3.16; 3.16 / 2.00 = 1.58) between the average high school dropout and the average graduate degree holder in wordsum performance.
Return to the home page where we initially entered our variables. How do we find them in the first place?
The “Search:” field (green circle) functions as a sort of search engine for variables. If we are looking for questions on religion, for example, we can type “belief” into the field and either hit enter or click the view button.
Or at least we should be able to, but for some reason this function does not work in the 2018 iteration of the survey at the time of this writing. However, for all variables except for those added for the first time in 2018, we can go to the 2016 page and find variables this way (notice the link now ends in +gss16 instead of +gss18). This page looks nearly identical. Let’s type “belief” into the search field and then hit enter or press the “Go” button.
Returned to us are several variables that have something to do with “belief” in them. We can then go back to the homepage and enter those specific variables into the “Row:” and “Column:” fields to discover how they cross-tabulate with other variables like we did early with wordsum and degree.
Returning to the “Variable Selection:” section, we can type specific variables–they must be verbatim–into the “Selected:” box (red circle). Let’s type the variable “hell” into the selected box and push the “View” button.
This shows us the question presented to respondents (“Do you believe in … d. Hell”?) and the distribution of all responses through all years the question has been asked. Parenthetically, questions are often asked as part of modules, so each item in the module will have a letter in front of it. This module presumably asked respondents whether they believed in a host of different things of which Hell was one.
There is also a drill down database below the “Selected:” and “Search:” fields that can be used to look through the entire field of variables since the survey’s inception in 1972.
I think that’s enough to get started playing around with the survey if you’re so inclined. Enjoy!