Measurement Invariance in Multilingual Survey Research: The Role of the Language of the Questionnaire

This paper examines whether or not survey measures asked to two linguistic groups exhibit invariance within a country. It tests configural, metric and scalar invariance in a model that distinguishes the response component and the cognitive component of the measurement process using survey items. Our findings show that when differences in the response process are allowed, concepts are (partially) invariant across groups. The analysis is conducted for items measuring some aspects of political trust and satisfaction with politics for six linguistic groups including French, Dutch, Estonian, German, Ukrainian and Russian in four countries. Data comes from the European Social Survey.