In lieu of an abstract, here is a brief excerpt of the content:

CHAPTER 9 Multi-variate Analysis of Political Cleavages W E now have a general overview of the relationship between partisan choice and various background variables. So far, our analysis has been based entirely on simple percentage tables, usually tables showing the relationship between voting intention and only one other variable. These tables tell us to what extent the more educated, for example, prefer the parties of the Right; but by themselves they tell us nothing conclusive about what causes an individual to vote for the Right or the Left. For higher levels of education, higher incomes and middle-class occupations all tend to go together. Furthermore, they all tend to be associated with Post-Materialist values. To obtain a clearer sense of whether a given variable actually influences party choice we must carry out multi-variate analyses. We will use two complementary computer programs for this purpose: the Automatic Interaction Detection (AID) technique and Multiple Classification Analysis (MCA). The AID procedure splits a survey sample into progressively smaller groups, on the basis of the relative power of a given predictor variable (religion, education, occupation, age, sex, and so on) to "explain" variance in the dependent variable (political party preference).1 This form of analysis is particularly useful in indicating whether the effects of given predictor variables are additive or interactive. Examination of the AID output for each of our eight countries indicates that interaction effects seem to have relatively little importance . This is a negative finding, but an important one. For status inconsistency theory suggests that interactions between two or more variables might play an important role in shaping political party choice. Our AID analyses reveal few if any instances of groups giving disproportionate support to the Left (or the Right) 1 For a more complete description of the AID analysis, see John A. Sonquist and James N. Morgan, The Detection of Interaction Effects (Ann Arbor: Institute for Social Research, 1964). Multi-variate Analysis — 245 as a result of interaction effects; in the few cases in which there is a hint of interaction according to the status inconsistency model, the effects are so weak that they could easily be due to sampling error. These results corroborate a series of recent findings that status inconsistency explains little, if anything, beyond what might be attributed to the impact of the respective variables.2 Given a set of essentially additive predictor variables, we can move on to a more conclusive form of analysis: Multiple Classification Analysis (MCA). MCA could be considered a form of dummy variable multiple regression. Like AID, the MCA technique is based on non-metric assumptions about the predictor variables. But whereas AID merely shows us major lines of cleavage based on the strongest predictor of party preference at given points in the breakdown, MCA gives us an indication of the explanatory power of each predictor variable across the sample as a whole. The MCA output provides two useful statistics for each predictor variable. The first is the Eta coefficient, an indicator of how much of the variation in party preference can be explained by the given predictor. The second is the Beta coefficient, which indicates whether the given predictor can still explain a significant portion of the variation when we control for the effects of all the other predictors. The output also specifies how much of the variation in the dependent variable can be explained by an entire set of predictor variables.3 Table 9-1 shows the results of MCA analyses for each of the eight countries. The predictor variables are ranked from top to bottom, according to the relative strength of their Beta coefficients. The dotted line across each table indicates a threshold below which the given predictors are considered to have a negligible effect on political party preference, when the effects of other variables are taken into account. We have (somewhat arbitrarily) set this threshold at a Beta coefficient below .075. The difference between Eta and Beta coefficients is often substantial. Thus, although certain predictors have a reasonably strong zero-order relationship with political party preference, the relationship may largely disappear when we take other variables into account. 2 For an excellent interpretation of the literature on this subject, see David R. Segal, Society and Politics: Uniformity and Diversity in Modern Democracy (Glenview, 111.: Scott, Foresman, 1974), 91-97. 3 See John A. Sonquist, Multivariate Model Building: The Validation of a Search Strategy (Ann Arbor: Institute for Social Research, 1970). TABLE 9 - 1 . Predictors of Political...

pdf

Additional Information

ISBN
9781400869589
MARC Record
OCLC
933516258
Pages
496
Launched on MUSE
2016-01-01
Language
English
Open Access
No
Back To Top

This website uses cookies to ensure you get the best experience on our website. Without cookies your experience may not be seamless.