We can use the GSS2016 data to learn how causal hypotheses can be evaluated with nonexperimental data.
1. Specify four hypotheses in which CAPPUN is the dependent variable and the independent variable is also measured with a question in the 2016 GSS. The independent variables should have no more than 10 valid values (check the variable list).
a. Inspect the frequency distributions of each independent variable in your hypotheses. If it appears that one has little valid data or was coded with more than 10 categories, substitute another independent variable.
b. Generate cross-tabulations that show the association between CAPPUN and each of the independent variables. Make sure that CAPPUN is the row variable and that you select “Column Percents.”
c. Does support for capital punishment vary across the categories of any of the independent variables? By how much? Would you conclude that there is an association, as hypothesized, for any pairs of variables?
d. Might one of the associations you have just identified be spurious because of the effect of a third variable? What might such an extraneous variable be? Look through the variable list and find a variable that might play this role. If you can’t think of any possible extraneous variables, or if you didn’t find an association in support of any of your hypotheses, try this: Examine the association between CAPPUN and WRKSTAT2. In the next step, control for sex (gender).
The idea is that there is an association between work status and support for capital punishment that might be spurious because of the effect of sex (gender). Proceed with the following steps:
i. Select Analyze/Descriptive statistical/Crosstabs.
ii. In the Crosstabs window, highlight CAPPUN and then click the right arrow to move it into Rows, Move WRKSTAT2 into Columns and SEX into Layer 1 of 1.
iii. Select Cells/Percentages Column/Continue/OK.
Is the association between employment status and support for capital punishment affected by gender? Do you conclude that the association between CAPPUN and WRKSTAT2 seems to be spurious because of the effect of SEX?
2. Does the association between support for capital punishment and any of your independent variables vary with social context? Marian Borg (1997) concluded that it did. Test this conclusion by reviewing the association between attitude toward African Americans (HELPBLK) and CAPPUN. Follow the procedures in SPSS Exercise 1d, but click HELPBLK into columns and REGION4 into Layer 1 of 1. (You must first return the variables used previously to the variables list.) Take a while to study this complex threevariable table. Does the association between CAPPUN and HELPBLK vary with region? How would you interpret this finding?
3. Now, how about the influence of an astrological sign on support for capital punishment? Create a cross-tabulation in which ZODIAC is the independent (column) variable and CAPPUN is the dependent (row) variable (with column percents). What do you make of the results?