Testing for measurement equivalence of individuals’ left-right orientation

Abstract

Subjective variables such as opinions, attitudes or preferences cannot be measured directly. Researchers have to rely on the answers people give in surveys, and whenever those answers shall be compared it is required that people answer these questions in the same way. Only then a concept can be used in different contexts. This paper deals with the measurement of the left-right concept: it analyses whether people make a distinction between a scale labelled 0 left and 10 right to one which is labelled 0 extreme left and 10 extreme right and tests whether the instrument is equivalent across groups. Following the three steps of invariance testing, configural, metric and scalar invariance, we find that the left-right response scale is on average equivalent across groups with different levels of political interest and different levels of education. This finding holds also in 23 of the 25 European countries tested, with the exception of the eastern part of Germany, Finland and France. In order to estimate how serious the difference between these two groups of countries is, we compare the observed means (which are affected by the difference) to the latent means (which are free of those effects), and the effect of the observed variable “attitude towards government’s intervention in the economy” on the observed variable “left-right self-placement” with the effect between these variables after correcting for scale difference. It was found that countries’ means can be compared but that the relationship with other variables might not be comparable among East Germany, Finland, France and the remaining countries.