t: time trend
y: year of presidential election
m: annual growth rate of real personal disposable income
p: cpi-based annual rate of inflation
u: annual unemployment rate
v: Democratic share of the two-party vote

 t   y      m      p    u    v 
=================================
 1 1932 -13.8936 -9.9 23.6 0.592 
 2 1936  11.2831  1.5 16.9 0.625 
 3 1940   5.3778  0.7 14.6 0.550 
 4 1944   2.6156  1.7  1.2 0.538 
 5 1948   3.7344  8.1  3.8 0.523 
 6 1952  1.33760  1.9  3.0 0.446 
 7 1956  2.92265  1.5  4.1 0.423 
 8 1960  0.14933  1.7  5.5 0.501 
 9 1964  5.47193  1.3  5.2 0.613 
10 1968  2.86171  4.2  3.6 0.496 
11 1972  2.88392  3.2  5.6 0.382 
12 1976  2.58274  5.8  7.7 0.511 
13 1980 -1.08862 13.5  7.1 0.447 
14 1984  4.92447  4.3  7.5 0.409 
15 1988  3.34549  4.0  5.5 0.461 

(1) Add the following three political dummy variables to the dataset (dummy variables are discussed in Ch. 14):
i:  1 if the incumbent is a Democrat
   -1 if the incumbent is a Republican
d:  1 if a Democratic incumbent is running
   -1 if a Republican incumbent is running
    0 otherwise
b:  1 if the incumbent party has been in control for two terms or more
    0 otherwise

(2) Enter the data into SPSS for Windows.  Use the single-letter variable names given above but provide variable labels of your choice. 

(3) Compute the variable mi=m*i and regress v on mi.  Summarize the regression results.  How do you interpret the results?  Specifically, how do you interpret the estimated slope coefficient?

(4) Compute the variables t=t+11 and pi=?p?*i.  Regress v on i, d, t, mi, and pi.  Interpret the results.

(5) Compute the variable bi=b*i.  Regress v on mi, bi, and d.  Interpret the results.  Based on this equation, how much annual growth rate of real personal disposable income is needed in 1992 for President Bush to win reelection?
 

B.  The following results were reported in Young et al., "Personal Agenda and the Relationship Between Self-Interest and Voting Behavior" (Social Psychology Quarterly, 1987, Vol. 50, No. 1, 64-71).  Pretend that you have a friend who doesn't understand statistics (we'll call her "Stefanie").  How would you explain the meaning of the following passage to Stefanie?

"To compare policy beliefs with self-interest directly, as well as the traditional party identification variable, all indices were entered into a multiple regression equation using vote choice as the dependent variable. For the high-importance condition, as expected, the beta for self-interest was .115, p<.001; for the policy belief index, the beta was .088, p<.025; and for the party identification, the beta was .704, p<.0001.
 

C.  Explain when multicollinearity occurs, and why it is a problem.
 

D.  Do review problem 13-24 in Wonnacott and Wonnacott, on page 429.
 

E. Read Rahn et al., "A Social-Cognitive Model of Candidate Appraisal" (in Ferejohn & Kuklinski, eds., Information and Democratic Processes, 1990). This article is also based on the 1984 Gallup Survey data.

(1) Try (if you can) to replicate the "whole sample" multiple regression results reported in Tables 2, 3, and 4 of the article. Make sure you appropriately define missing values and reverse scales. Summarize the results in your report. Attach SPSS programs only.

(2) Comment on the use of R-Squared in this article. You should read and refer to at least two of the following articles which debate the use (and misuse) of R-Squared:

1. C. H. Achen, 1982, Interpreting and Using Regression. Especially Section 5, "Choosing a Specification." And 1990, "What Does 'Explained Variance' Explain?: Reply." Political Analysis, Vol. 2, 173-184.

2. M. S. Lewis-Beck and A. Skalaban, 1990, "When to use R-Squared." Political Methodologist, 3: 9-11. And 1990, "The R-Squared: Some Straight Talk." Political Analysis, Vol. 2, 153-171.

3. G. King, 1990, "When Not to Use R-Squared." Political Methodologist, 3: 11-12. And 1990, "Stochastic Variation: A Comment on Lewis-Beck and Skalaban's 'The R-Squared.'" Political Analysis, Vol. 2, 185-200.

4. G. King, 1986, "How Not to Lie with Statistics: Avoiding Common Mistakes in Quantitative Political Science." American Journal of Political Science, 30: 666-687. And 1991, "'Truth' Is Stranger than Prediction, More Questionable than Causal Inference." American Journal of Political Science, 35:1047-1053.

5. R. C. Luskin, 1991, "Abusus Non Tollit Usum: Standardized Coefficients, Correlations, and R2s." American Journal of Political Science, 35:1032-1046.