Democracy and Social Science II

Winter 2003 (Prepared January 28, 2003)

Homework #3. Due February 4, 2003

Use the data set (Excel format) and codebook (rich text format) for an abridged version of the 1996 National Election Study from the last homework assignment.

Your assignment for this week is as follows:

For reasons that will remain obscure, if you were to use a dichotomous variable like vote choice as the dependent variable in a regression model estimated using OLS, your regression coefficients estimated with OLS will have undesirable properties. As a result, political scientists will often use an alternative to OLS such as logit analysis or they will try to measure the dependent variable in a slightly different way. For your homework, you are to pursue the latter approach, using the difference between the respondent’s placement of Clinton and Dole on the feeling thermometer scales as a measure of net affect towards the candidates, based on the assumption that people who “like” Clinton better than Dole are more likely to vote for Clinton and that people who “like” Dole better than Clinton are more likely to vote for Dole. Further, the greater the difference in affect toward the two candidates, the more likely it is that the respondent will vote for their more preferred candidate.

1. Construct a reasonable test using a t-test or t-tests for whether a respondent’s placement of candidates on feeling thermometer scales provides a reasonable alternative to vote choice. (Note that depending on the way that you set up the test(s), it is possible that you will have a negative critical value, and your t-statistic would have to be less than the critical value in order to reject the null hypothesis).

2. Once again trot out two of your hypotheses about the continuous variables (so please don’t use dichotomous variables like gender or race) that you think are related to vote choice. Then perform the following steps.

a. Compute the covariance between each of those variables and a variable measuring the difference between Clinton and Dole’s feeling thermometer scores.

b. Compute the correlation between each of those variables and a variable measuring the difference between Clinton and Dole’s feeling thermometer scores.

c. Perform a bivariate regression with the difference between Clinton and Dole’s feeling thermometer scores as the dependent variable and the variables that you hypothesized to influence vote choice as the independent variables. Report your estimates of the intercept, the regression coefficient, and the R².

3. Using your results from question 2, interpret the effect of each of the independent variables on relative candidate affect and explain how well each of these variables explains respondent’s attitudes.

For each question in the homework assignment, you must report enough of your steps that the grader can follow what you are doing. Except for simple calculations like the mean, you are not to use the “canned” functions in Excel (in a week or so, this will no longer be taboo).

Hint. I suggest that you perform the following steps to avoid problems with missing values. Step 1. Copy all of the observations for the two variables into a separate page of the spreadsheet, being careful to ensure that each row of data lines up the same way on both pages of the spreadsheet. Step 2. Sort one column of data so that all of the missing values are either at the top or the bottom of your spreadsheet. Step 3. Delete the rows of data with missing values. Step 4. Sort the second column of data in the same fashion. Step 5. Delete the rows of data with missing values.