Using qualitative methods for attribute development for discrete choice experiments: issues and recommendations

TLDR

Attribute generation for discrete choice experiments is poorly reported and faces challenges such as aligning with random utility theory, requiring a two‑stage conceptual and refinement process, and balancing the reduction of complex qualitative insights into precise terms. The study investigates attribute‑development challenges in DCEs, compares qualitative methods, and recommends an iterative constant‑comparative approach with clear reporting guidelines. The authors analyze eight studies—four developing attributes for measurement instruments and four for ad hoc policy questions—to illustrate the attribute‑development process. Findings indicate that the choice of qualitative data collection depends on the question’s sensitivity and the availability of prior qualitative material. © 2011 John Wiley & Sons, Ltd.

Abstract

SUMMARY Attribute generation for discrete choice experiments (DCEs) is often poorly reported, and it is unclear whether this element of research is conducted rigorously. This paper explores issues associated with developing attributes for DCEs and contrasts different qualitative approaches. The paper draws on eight studies, four developed attributes for measures, and four developed attributes for more ad hoc policy questions. Issues that have become apparent through these studies include the following: the theoretical framework for random utility theory and the need for attributes that are neither too close to the latent construct nor too intrinsic to people's personality; the need to think about attribute development as a two‐stage process involving conceptual development followed by refinement of language to convey the intended meaning; and the difficulty in resolving tensions inherent in the reductiveness of condensing complex and nuanced qualitative findings into precise terms. The comparison of alternative qualitative approaches suggests that the nature of data collection will depend both on the characteristics of the question (its sensitivity, for example) and the availability of existing qualitative information. An iterative, constant comparative approach to analysis is recommended. Finally, the paper provides a series of recommendations for improving the reporting of this element of DCE studies. Copyright © 2011 John Wiley & Sons, Ltd.