Evidence of preference construction in a comparison of variants of the standard gamble method

An increasingly important debate has emerged around the extent to which techniques such as the standard gamble, which is used, amongst other things, to value health states, actually serve to construct respondents' preferences rather than simply elicit them. According to standard theory, the variant used should have no bearing on the numbers elicited from respondents, i.e. procedural invariance should hold. This study addresses this debate by comparing two variants of standard gamble in the valuation of health states. It is a mixed methods study that combines a quantitative comparison with the probing of respondents in order to ascertain possible reasons for the differences that emerged. Significant differences were found between variants and, furthermore, there was evidence of an ordering effect. Respondents' responses to probing suggested that they were influenced by the method of elicitation.