Designing Better Questions for Complex Concepts with Reflective Indicators

Abstract

There are many concepts in the social sciences that are measured using multiple indicators. Such concepts have been called by Blalock (1968) concepts-by-postulation because one needs a theoretical argument to define them. Within the set of concepts-by-postulation, a distinction has been made between concepts with reflective indicators and concepts with formative indicators. This distinction refers to the assumption whether the latent concepts determine the observed indicators (reflective) or that the indicators together determine the latent concept of interest (formative). Blalock complained that developing measurement procedures for complex concepts, researchers think mainly about questions not about concepts that these questions measure. In this way, the questions used contain unique components which reduce the quality of the composite score based on these questions as measure for the complex concept of interest. Saris and Gallhofer have shown how alternative formulated questions can be developed to measure so-called concepts-by-intuition. In this paper, we will show that the same procedure can be used to avoid unique components in the measurement of complex concepts with reflective indicators and that in this way the quality of the composite score for complex concepts can considerably be increased.