Causes of generalized social trust: An innovative cross-national evaluation

In this chapter we want to establish whether the 3-item scale to measure generalized social trust (GST) in the European Social Survey (ESS) can be meaningfully used in comparative research. If so, we also want to study the causes of GST cross-nationally; building upon the work by Delhey & Newton (2005). The standard procedure to assess the comparability of a survey measure is by means of measurement equivalence tests, a specific form of multi-group confirmatory factor analyses (MG-CFA). In general it is quite difficult to evaluate these models, i.e. it is not easy to decide when a model fits the data, how and where a model should be improved, and when the improvements are sufficient. In this chapter we introduce two innovations for testing measurement equivalence of survey measures. One innovation involves an alternative procedure to evaluate structural equation models (Saris, Satorra, and Van der Veld, 2009). This procedure is implemented in a software program called JRule which is developed to detect misspecifications in SEM models taking into account the power of the test. The other innovation concerns the separation of error components and the substantive components in the response, which makes it possible to test for measurement equivalence after correction for random and systematic measurement error. Applying these innovations in our study resulted in evidence that the 3-item measure for GST is scalar invariant in all 19 countries that we analyzed