Exercise 7. Gender, presidential vote, and marital status
Another reason for introducing a control variable into the analysis is to see if the relationship between the independent and dependent variable is much stronger for some groups of people than it is for others. If we find that such is the case, we commonly refer to that as a conditional relationship. A good example of a conditional relationship involves the association between gender and presidential vote, which is explored in Exercise 4. In Table 4A, you should have observed a modest relationship between gender and voting, with men being about seven percentage points more Republican in their vote than were women (or women being seven points more Democratic than men).
We also saw in Exercise 2 that marital status was related to the presidential vote. We might hypothesize that the relationship between gender and vote would be much stronger for those who are not married than for those who are. For one thing, married couples might have a tendency to agree politically, although this certainly is not always the case. To examine the possible conditional effect that marital status has on the relationship between gender and vote, generate the appropriate three-variable table. You should use the recoded version of presidential vote that you used in previous exercises, and you should use the recoded version of marital status that you created in Exercise 2 (Married/widowed versus Single/divorced/separted).
In this example, our original two-variable relationship almost disappears in one category of the control variable and becomes much stronger in the other category.