Exercise 4. Gender and presidential vote
One of the interesting features of American politics over the past 20-25 years is the existence of a gender gap in voting behavior. Specifically, women have been more likely than men to vote for Democratic candidates. To examine this relationship for 2004, you should start by generating a table that shows how an individual's gender (V143) was related to the individual's presidential vote. For the reasons suggested in Exercise 1, you should use the recoded version of V002 that you created for that exercise, so that you examine only the major-party vote (i.e., only the Bush and Kerry voters).
Table 4A raises as many questions as it answers. Why should gender be related to the vote? What explains the greater tendency for women to vote Democratic (or men to vote Republican)? We can attempt to explain this relationship by examining possible intervening variables (i.e., those that are influenced by the independent variable and in turn affect the dependent variable). In this case, we would might want to identify the attitudes that create a gender gap in voting. We might speculate that attitudes toward abortion produce the relationship. Kerry was generally identified as supporting legalized abortion, while Bush was largely opposed to it, except in some narrowly defined circumstances. Perhaps women were more likely to vote for Kerry because they were more supportive of abortion and cast their ballots on the basis of that issue.
As we saw in exercise three, in order to examine a potential intervening variable, you should run the original two-variable relationship with the potential intervening variable added as a control variable. In this exercise, you should use V121 (attitude toward abortion) as your control variable, and you should recode V121 so that there are just two categories (favor and oppose), which will make the three-variable table easier to interpret.
In this example, the relationship between gender and presidential vote is just as strong when we control for abortion attitude as it is when we do not. This tells us that the effect that gender has on presidential vote is not due to attitudes on abortion. If the original two-variable relationship had largely disappeared in the three-variable table (which is what happened in exercise 3), then we would have concluded that the control variable was a key intervening variable in this relationship, but that is not what happened.