Harmonizing substantive instruments

Are you currently combining or planning to combine data from different survey programs? Or are you involved in a survey which plans to change a question (e.g., in a time series)?

Then you are probably aware of a central methodological challenge: How can data measured with different instruments be made comparable? Or in other words: How can responses collected with different instruments be transformed so that they form a seamlessly harmonized variable?

To support you in this, GESIS offers information and consulting on the harmonization of substantive variables measured with different instruments.

Special focus: Harmonizing instruments for latent constructs

A special focus of methodological research and consulting in this area is the harmonization of instruments which capture latent constructs, such as values, attitudes, intentions, interests, emotions, or personality traits. Unlike manifest constructs, intangible constructs have no natural units. We can measure income in currency units and behavior in times per month. But how much political interest is “very interested”; and how does that compare to “very strongly”? In short, we want to compare apples and oranges.

Equating is a very promising approach to overcome that challenge. Equating originates from psychometric performance and aptitude diagnostics, where it is used to make results from different tests comparable. Despite equating’s psychometric roots, it can also be applied to instruments with only one question with observed score equating. The advantages of equating lie in its solid mathematical and empirical foundation and its usefulness as a general framework to evaluate the quality of instrument harmonization. However, in addition to equating, other approaches are available on which we consult you as well.

Consultation service in a nutshell:

  • Assessing the comparability of different instruments.
  • Weighing the potential consequences of a lack of comparability.
  • Harmonizing substantive variables measured with different instruments; especially latent constructs.

Dr. Ranjit K. Singh

ranjit.singh(at)gesis(dot)org

Department: Survey Design and Methodology