Exercises to Accompany the Instructional Subset 1996 American National Election Study
John L. Korey,
Calif State Polytechnic University Pomona
JeDon Emenhiser, Humboldt State University
© The Authors, 1998; Last Modified 16 August 1998
- Weight the data by the WEIGHT variable. Run frequency distributions for several variables and compare the results with those found in your codebook. How much difference does the weighting make?
[For exercises 2 through 5, prepare variable and value labels for new or modified variables if your software (such as SPSS) has this capability. Also, don't forget to allow for missing values.]
- Using the YEARBORN variable, create a new AGE variable as follows:
- The G.I. Generation (born 1925 or earlier)
- The Silent Generation (born 1926-1945)
- Early Boomers (born 1946-1955)
- Late Boomers (born 1956-1964)
- Gen Xers (born 1965 or later)
- Create a new variable that recodes EDUC into a smaller number of categories (3 to 5) of about equal size. Do the same for INCOME. Be sure not to replace the original variables -- -you'll need them later on.
- Create a new ETHNIC variable that combines the RACE and HISPANIC variables.
- Combine RELIG1 and RELIG2 into a single variable (RELIGION) that, for example, classifies a respondent as "Protestant" if he or she either attends a Protestant church or, though not a church-goer, considers him- or herself to be Protestant.
- Cross-tabulate POLVIEWS with the AGE variable to see if there is a "generation gap" when it comes to ideology. Repeat this several times, but substitute views on specific issues for POLVIEWS.
- Different generations may define terms like "liberal" and "conservative" differently. Crosstabulate POLVIEWS with measures of attitudes on specific issues, controlling for AGE. Do some issues correlate more strongly in some age groups than in others with self-identified ideology?
- Cross-tabulate VOTE with SEX. Repeat with a control for AGE. How do you explain the resulting patterns?
- Cross-tabulate PARTYID with VOTE, VOTEPRES, and VOTEHSE. Based on the results, are the independent leaners (codes 2 and 4) true independents, or are they really subconscious partisans? That is, do they behave more like pure independents (code 3) or weak partisans (codes 1 and 5)? If you were to recode PARTYID into three categories, which would be the most appropriate categories?
- Several variables in this dataset provide 7-point scales on various issues. Prepare a correlation matrix among all of these variables. Which scales tend to cluster together (that is, have strong correlations, either positive or negative)? Which scales tend not to be closely associated?
- Several variables provide "feeling thermometers" regarding governmental institutions. Are there significant differences in the mean values of these variables when broken down by different background variables (SEX, ETHNIC, REGION, PARTYID, RELIGION, AGE, MARITAL, and the recoded education and income variables)?
- Use bivariate regression (least squares) analysis to assess the separate impact of EDUC and INCOME (unrecoded) on each of the feeling thermometers. Use multiple regression to measure the combined impact of the two independent variables.