Using confirmatory factor analyses to identify attitude groups
There is now widespread agreement that foreign policy beliefs among U.S. leaders as well as the public are structured in terms of multiple dimensions, and that these beliefs cannot be represented adequately by a single "internationalist-isolationist" continuum. However, no consensus exists regarding the optimum number or nature of these basic attitudinal dimensions for either the leaders or the public. Past comparative analyses of the CCFR surveys relied on exploratory factor analyses to establish the dimensions underlying foreign policy preferences. None extended these exploratory analyses to generate statistically testable models of the attitude structures of the leaders or the public using confirmatory factor analyses. The attitudinal models resulting from confirmatory factor analyses of the 1994-2002 CCFR surveys have been successfully replicated using matched data sets from the Times Mirror/Pew Research Center surveys of the general public (1993 and 1997) and the Foreign Policy Leadership Project (1992 and 1996).
Generally, these analyses have shown that a four-factor model for the public (five factors for the leaders) provides a probabilistic close fit to the observed correlation matrix in the data set, and that the accuracy of the models-that is, closeness of fit-declines as we move from the relatively complex four-factor model for the public (five-factor model for the leaders) to the simpler three-factor and two-factor structures. See the additional reading list for other articles on these analyses. |
