Understanding Public Opinion in Post-Communist States: The Effects of Statistical Assumptions on Substantive Results

Abstract

FOLLOWING THE COLLAPSE OF COMMUNIST RULE in Eastern Europe and the Soviet Union, a large number of scholarly works appeared analysing public opinion in the region.' Most of the research has focused on identifying the sources of support for current political leaders, independence, democracy, or the market in the late Soviet period or the early post-Soviet period.2 Developing a systematic understanding of the factors that influence public attitudes about politics in post-communist states can contribute to the long-term success of democratisation, marketisation and state-building efforts. However, the statistical assumptions that guide much of this literature may undermine the validity of the substantive conclusions. In this article we investigate the effects of prevailing assumptions about causality on the substantive results of quantitative survey data analysis in the post-communist mass attitudes literature. Most research about post-communist mass attitudes relies on the specification of a single-stage causal model, usually employing OLS linear regression estimation. Although social scientists have long recognised that such models may be faulty under certain conditions,3 there is little recognition of the problem in the growing literature about post-communist mass attitudes. In addition to model specification, measurement error is a problem that post-communist studies scholars have often ignored. We compare results from a single-equation statistical approach to those of a structural equation model (SEM) in an analysis of support for the government and political system of Kazakhstan by Russian-speaking minorities. The SEM analysis in this study involves estimating a complex (partially mediated) causal structure, employs maximum likelihood estimation (MLE), and includes estimation of a measurement model for attitudinal latent constructs. The results demonstrate that