Researchers' Stated Preferences for Stated-Preference Methods

BACKGROUND: New and better methods are constantly being developed in applied research. Eventually, state-of-the-art procedures become state-of-the-practice procedures that are required for peer-reviewed publications. These methods have been extensively used and tested for validity in environmental economics and market research. Health economists are increasingly employing these methods to elicit patients' and physicians' preferences. However, the quality of data collected and the validity of results are sensitive to a number of researcher decisions. OBJECTIVES: The aim of this study is to quantify the degree of current consensus in the rapidly evolving area of stated-preference research in health valuation. METHODS: Stated-preference (SP) methods include both contingent-valuation and conjoint-analysis approaches. Conjoint analysis asks respondents to choose, rate or rank hypothetical or existing alternatives. Designing such a study requires numerous judgments involving such considerations as what features of the treatments to include, how many alternatives to show, how many groups of alternatives to show, how to construct alternatives to satisfy particular statistical criteria, and how to estimate preference parameters to obtain valid, unbiased results. The resulting tradeoff patterns reveal the underlying importance weights respondents use to evaluate alternatives. Employing the same method, we have developed a meta-survey of this same type to elicit researchers' preferences for the methods themselves. We surveyed researchers who have published manuscripts in peer-reviewed journals on stated preference methods in the last 10 years. We asked these researchers to evaluate hypothetical studies, where one of the stated preference methods, such as contingent valuation or stated choice approach was used. The researchers compared a series of pairs of hypothetical studies with specified characteristics in health applications. They were first asked which study is better and then asked which study they would recommend to use to inform policy decisions by an agency of the government where they reside. RESULTS: We used bivariate probit model to estimate researchers' preferences. Around 75% of the researchers in the sample were academic and about 40% reported that they reside in the US. Researchers were more likely to recommend a study to inform policy decisions if the survey is pre-tested or conducted by mail, there is a scope test, and the econometric model involves advanced modeling, such as random-parameters model. Researchers who have used contingent valuation studies were likely to recommend a study used contingent valuation and researchers who have used conjoint analysis studies were likely to recommend a study used conjoint analysis. While younger researchers were likely to recommend surveys conducted through mail, older researchers were likely not to recommend either study. The location of the researcher did not have any significant effect on researchers' preferences. CONCLUSION: Using an SP method to evaluate SP methods was a convenient and valid way of investigating current research. Our study results state that it is hard to reach a consensus among the researchers and there is room for improvement in health applications