Testing Measurement Equivalence across Nations: An Empirical Investigation of Environmental Concerns

Wednesday, 18 July 2018: 17:30
Oral Presentation
Sandra MARQUART-PYATT, Sociology & Environmental Science and Policy, Michigan State University, East Lansing, MI, USA
Chloe QIAN, HUI, Michigan State University, USA
In seeking to describe trends in public opinion about environmental concerns cross-nationally, social scientists use a variety of measures and analytical techniques. Although an understudied topic, the question of measurement comparability across contexts is important for this line of inquiry. Only a handful of studies assess the extent to which measures of environmental concern are comparable in cross-cultural surveys using an appropriate analytical technique. This research examines public opinion on the environment in cross-national context using data from thirty-seven countries in the 2010 International Social Survey Program (ISSP) Environment dataset. We use structural equation modeling with latent variables (SEMLV) to construct a latent variable of environmental risk perception. We examine the composition, level, and distribution of the latent construct environmental risk perception cross-nationally, with comparisons across theoretically-informed regional groups as well as individual countries. Particular attention is given to how structural equation modeling can be used as a tool for testing measurement equivalence in cross-national research.

We conduct confirmatory factor analyses (CFAs) across many axes of comparison (pooled, by region, and for individual countries) given their salience in setting the stage for multigroup comparisons using CFAs (MGCFAs) (Bollen 1989). Establishing equivalence, defined in this study as configural and metric invariance for hierarchy of invariance tests, is vital for this study as MGCFAs enable the testing of to what extent environmental risk perception is invariant across the regions and the 37 countries in the 2010 ISSP data. After establishing measurement equivalence of the latent construct, we investigate how a core set of predictors including individual education, socio-demographics, environmental beliefs and attitudes, willingness to make environmental sacrifices, and personal efficacy shape environmental risk perception across nations. Results from structural equation modeling with latent variables (SEMLV) reveal that the model is robust across nations, yielding insights for future comparative work.