SOCIAL STRATIFICATION
General Social Survey Internet Assignment

Robert E. Wood
Department of Sociology, Anthropology and Criminal Justice
Spring Semester 1999

Rationale: The purpose of this exercise is twofold. First, it aims to teach and/or reinforce skills in elementary data analysis. Multivariate analysis looks at how different variables are related to each other. In the class lab session and in this assignment, we will explore how to produce and interpret cross-tabulations of two variables, and how to introduce a third, control, variable. Furthermore, we will learn how to do these types of multivariate analysis over the internet using the data from the General Social Survey. Knowing how to do this has many potential practical applications. When writing papers for other classes, you can easily call up and print out data on attitudes and behavior that can support or test your arguments. In business and service agency settings, you can immediately access data to help guide decisions about marketing, program design, and many other things. With not only the General Social Survey but several other major data sets at this web site at your instant command, you can become an essential resource person in a variety of group and organizational settings.
Due Date: Tuesday, March 9th

Preparation: We will learn about several interactive websites that allow you to select and analyze data over the internet in the lab sessions on Thursday, Feb. 25th and Tuesday, March 2nd.

Part of these sessions will involve a review of the basic steps in quantitative data analysis, including:

• Formulating a two-variable hypothesis
• Operationalizing variables
• Creating a cross-tabulation
• Determining whether the data are statistically significant
• Deciding whether your hypothesis is supported or not
• Introducing a third, control variable, and seeing how it affects the original relationship.

Before proceeding with the computing part of this exercise, you should plan ahead what you are going to do. This involves two steps:

1) I want you to explore two ways in which social class background is related to attitudes, behavior, or social attributes using data from the General Social Survey. As you undoubtedly know from other sociology courses, social class is one of the most central concepts in sociology, and one of the most important predictors of individual behavior and life chances. To measure social class, we will use a self-identification question in which the respondents are asked to choose between different class labels. The code for this variable is "class". This will be your independent variable.

To choose your dependent variables, consult the "Standard Variable List" for the 1972-1994 General Social Survey Cumulative File. This is available online, but I have printed it out and appended excerpts to these instructions for your convenience. You can probably figure out what most of them involve, but you can always call up the full question by going to the codebook site and clicking on the question you are interested in. Choose two variables (from different sections of the list) that you think are likely to be related to class, and think about what kind of relationship you expect to find. For each of these, write down a hypothesis. For example, if you were to choose "happy", think about whether you would expect happiness to increase or decrease as you go up the class scale. Write down the underlined code names for these two variables and your hypothesis about how each is related to social class.  In most cases, your hypothesis should be in the form of: The higher a persons's social class position, the less likely (more likely) is the person to.....

Using the instructions below, you will generate cross-tabulations of these two variables with the "class" variable and interpret the results.

2) In the second part of this exercise, I want you to introduce a "control" variable. Controlling for a third variable allows you to see if the original relationship between two variables is altered when you compare people who have similar characteristics on the third. In this exercise, the control variable you will use is "race". What you want to see is whether controlling for race alters the association you found between "class" and one of the variables you listed above.

Running the Cross-Tabulations: If you are reading a hardcopy version of these instructions, now proceed to a computer with internet access.   Access this page from the course home page at http://camden-www.rutgers.edu/~wood/332gss.html, then click on
SDA Survey Data Archive for Survey Documentation and Analysis
You are now ready to proceed with your analysis.

You will note that there are actually several data bases you can also use, but for the purposes of this analysis, click on GSS Cumulative Data File 1972-1994.    At the webpage that appears, under "Select an action," the default is to browse the codebook, but we don't have to do that since you've already chosen your variables. So with the mouse, click on the icon to the left of "Run Crosstabulation." Then click on the box marked "Start" and then on "Continue Submission" if an intermediate screen appears.

You should now be at a screen that will read:
CSA Tables Program
(Selected study: GSS 1972-1994 Cumulative Datafile)

In the "Horizontal" box, type "class" as your independent variable.
In the "Vertical" box, type the code name of the first dependent variable you chose. Then:
Click on "Vertical" for "Percentaging"
Click on "Yes" for "Statistics"
Click on "Yes" for "Question Text?"
Then click on the box, "Run the Table."

In a few seconds, you will see the cross-tabulation you have produced. Make sure that your independent variable is listed as the horizontal variable and your dependent variable as the vertical variable.  To print out your table, you follow the usual browser instructions, except for the fact that you may want to print the table out sideways (what is called "landscape" format) if it extends beyond the screen to the right.  

Now click on "Back" and run and print out the table for your other variable, entering your second variable code name in the "Vertical" box and keeping "class" in the "Horizontal" box. Be sure that vertical percentaging, statistics, and question text? are specified properly (as above).

Before proceeding further, examine the Chisquare "p" value to make sure that at least one of your tables is statistically significant (that is, meets the standard of there being less than a one-in-twenty probability that the results could be obtained by chance); "p" must equal .05 or less.  If neither of your hypotheses produced statistically-significant results, experiment with alternative dependent variables until you get one relationship that meets the test of statistical significance.

You now have just one more computer task to perform, which is to rerun one of your cross-tabulations using a control variable. Choose one of your two variables; it should be one where the differences were statistically significant (p=0.5 or less). Set up your your variables exactly as before (using your chosen variable and class again), but this time type in "race" in the "Control" box. Now run and print out this table (which will come out as four, one for each racial category and one for all combined). Print these tables out also. Assuming you have three sets of tables in hand, you are now finished with the computer work necessary for this project. If you are doing this at the computer lab, be sure to exit fully out of your account.

Interpreting and Writing Up Your Findings: In your written report, I want you to provide a brief summary of your findings, table by table. Label your printouts Tables 1-3 (the last table should include the two sub-tables for whites and for blacks--omit the others) and cut and paste them into your report (scissors and tape are fine; advanced computer users may download the tables and insert them into a word-processing program if they wish). In the body of your report, for each of the first two tables, be sure to include the following:

1. State your hypothesis and explain briefly your reasoning behind it. [For example: The higher the social class of a person, the more likely he or she is to be happy. This seems likely because people of higher social classes are more likely to have the material and other resources to secure the things that are important to them.]
2. Paste in the relevant table.
3. State whether the results were statistically significant. [That is, is the Chisquare p less than 0.05?]
4. Was your hypothesis supported? If the differences in the data were not statistically significant, the hypothesis is rejected. If the data were statistically significant, you still need to examine the table to see if the data go in the direction of your hypothesis. State your conclusion about whether your hypothesis was supported. Summarize briefly what the data say.
5. Repeat this for your second variable.

For the final set of tables (with the control variable), examine the tables for whites and blacks to see if controlling for race produces different results, and respond to these questions:

1. Do the class differences hold when race is held constant? Or does race significantly modify the relationship you found in your earlier table?
2. If the introduction of race does modify your original findings, explain how. Are the findings changed for both whites and blacks or just for one group?
3. Based on your findings when you control for race, which variable would you conclude is more important for explaining variation in the dependent variable: class or race?   Summarize your interpretation of the data.

Return to Social Stratification Course Homepage

Feb. 24, 1999