Harmonizing Single-Question Instruments for Latent Constructs With Equating Using Political Interest as an Example.

Abstract

Many latent constructs in the social sciences, such as political interest, are measured with single-question instruments. Furthermore, survey programs often differ in the wording and response format of those instruments. This is problematic if we want to compare or combine data across different surveys or across changing instruments within a survey program. Consequently, we need robust methods to establish comparability in existing data (i.e., ex-post harmonization). In this paper I demonstrate the usefulness of an approach from psychometry: Observed score equating with a random groups design. Using two existing instruments for political interest, I show that equating works well in transforming the numerical scores of instruments so that they can be compared across instruments. Since random groups equating needs data for both instruments from the same population, I also demonstrate the feasibility of equating using an online nonprobability sample as well as using two probability samples of the adult German population.