Conducting Your Own Research

One way of studying voting behavior might be to have your computer generate all possible tables that could be constructed from these data. Such a strategy is neither practical nor useful. Tables by themselves tell you very little. They have to be interpreted and explained. This interpretation usually is done in terms of some theory or generalization about how people behave.

To make research efforts more meaningful, and to avoid wasted effort, we need to carefully frame the question we are studying and to be sure that we have some sound reason for studying it. That reason usually is directly related to some theoretical concepts or ideas that we have learned from the work of others. The general rule is to apply what we already know to discover something new.

It is important to explicate why we are looking at a particular set of relationships, because that justification becomes part of the explanation of what the contingency tables tell us. As we have said earlier, political scientists are far more interested in the relationship among variables than they are in the actual percentages of some group that voted one way or another. Relationships help to explain behavior. We especially want to know why the variables are related the way that they are. We also want to know why two variables that our theory predicts should be related might wind up not to be related when we look at the data. Often, the lack of a finding or the fact that two variables are unrelated is just as important as finding that two variables are strongly related, but only when we have a pre-existing theory or hypothesis that predicts that the two variables should be related.

Most research involves hypothesis testing. We start from a theory developed by reading other people's work; we generate hypotheses from that theory; and we test those hypotheses by comparing our predicted relationships to those resulting from data analysis. If the data bear out our hypotheses, we can claim some support for the theory; if the data do not bear out our hypotheses, we can conclude that the theory is not supported. In either event, the conclusions are valuable.

Using what you already know about voting behavior, as well as your intuition, you should be able to design a research project that examines some aspect of voting behavior or public opinion. Specifically, you should:

  • formulate a question about the relationship among a set of variables;
  • justify the selection of variables and the expectations you have about the connection between the variables;
  • test your ideas by obtaining the necessary tables;
  • carefully interpret the tables and write up your conclusions.

Although we have tended to focus on one dependent variable--who people vote for--there are of course other dependent variables. One might be interested in issue differences between men and women, as suggested above, or in why people think one candidate for office is honest and another one is not, or in what attributes respondents are considering when they report feeling "warm" or "cool" towards a candidate for office on a feeling thermometer. The data set contains a wealth of variables that can be considered as either dependent or independent variables according to the type of research project your are pursuing.

Some possible research topics are suggested by the above exercises:

  1. The impact of perceptions of candidate characteristics on presidential voting: Exercise 9 examined the relationships between perceptions of the leadership abilities of the presidential candidates and voting behavior. There are other candidate characteristics that voters may find relevant, and this dataset contains items on a number of such characteristics. One possible research topic is the effect that these assessments of candidate characteristics have on voting? Are some assessments more strongly related than others to the vote? Overall, how did the voters assess the personal characteristics of the two candidates? Did one candidate have an overall advantage when it came to these perceptions? Can we be confident that these assessments have a direct effect on voting, as opposed to just being associated with vote choice? What voter characteristics influence how the voters assess the personal qualities of the candidates?

  2. The impact of economic evaluations and attitudes on presidential voting. Exercise 3 and Exercise 10 examined the effect of several economic evaluations and attitudes on presidential vote choice. This topic could be explored in more depth. How are evaluations of personal financial situation, perceptions of the national economy, and evaluations of the president's handling of the economy related to each other and to the vote? What is the causal order among these variables (i.e., what affects what)? What voter characteristics affect economic evaluations and attitudes?

  3. The impact of moral issues on presidential voting. The effect of several issue orientations on the vote is examined in Exercise 4, Exercise 5, and Exercise 6. In particular, we might be interested in the impact that so-called moral issues, such as abortion or gay marriage, had on the presidential vote in 2004. Some analysts speculated that these issues played a significant role in Bush's victory. This suggests several interesting research questions: Did these issues have a significant effect on the vote? Did Bush have an advantage over Kerry when it came to these issues? What voter characteristics affect attitudes on these moral issues? One could also investigate the effect that other issue orientations had on voting, such as the effect of attitudes on economic or social welfare issues or the effect of attitudes on questions of taxation, spending, and deficits.