Applying a Two-Step Strategy to the Analysis of Cross-National Public Opinion Data

W. Phillips ShivelyDepartment of Political Science, University of Minnesota, 1414 Social Sciences Bldg. 267 19th Ave. South, Minneapolis, MN 55455 email: kjlong{at}umich.edu (corresponding author) email: shively{at}polisci.umn.edu. In recent years, large sets of national surveys with sharedcontent have increasingly been used for cross-national opinionresearch. But scholars have not yet settled on the most flexibleand efficient models for utilizing such data. We present a two-stepstrategy for such analysis that takes advantage of the factthat in such datasets each "cluster" (i.e., country sample)is large enough to sustain separate analysis of its internalvariances and covariances. We illustrate the method by examininga puzzle of comparative electoral behavior—why does turnoutdecline rather than increase with the number of parties competingin an election (Blais and Dobryzynska 1998, for example)? Thisdiscussion demonstrates the ease with which a two-step strategyincorporates confounding variables operating at different levelsof analysis. Technical appendices demonstrate that the two-stepstrategy does not lose efficiency of estimation as comparedwith a pooling strategy.