The nature of individual preferences: a prologue to Johannesson, Jonsson and Karlsson.

In their recent Guest Editorial in this journal, Johannesson et al. provide descriptions of the different types of outcome measure that can be used in economic evaluation and recommendations about which measures should be used. Their central argument in the normative sections of the paper appears to be that the chosen outcome measure should reflect individual preferences, which they note is ‘in accord with the individualistic foundations of welfare economics’ (p. 282). For example, the authors discuss the assumptions that are necessary for life years gained to be viewed ‘as a valid cardinal utility function that will rank treatments according to individual preferences’ (p. 281) and suggest that the problems associated with using composite scores derived directly from quality of life instruments are that they ‘may not even rank health states according to individual preferences’ (p. 281). The entire discussion of the QALY-HYE debate is focused around the extent to which these measures can be assumed to be valid representations of individual preferences and the lengthy discussion of the contingent valuation (CV) method is principally to alert researchers to the ‘best practices’ in eliciting willingness-to-pay (WTP) responses. It is somewhat surprising, therefore, that Johannesson et al. do not consider the nature of the individual preferences that they wish to judge the various outcome measures by. It is important to do so however, particularly as the recommendations on how to proceed with the measurement of health outcomes, which is clearly a central part of the paper (for example, see pp. 288 and 293), will to a large extent depend on which how individual preferences are viewed and which paradigm is adopted. The received wisdom amongst economists has been that individuals have clear, well-defined preference functions which can be ‘tapped into’ by appropriate questions: in the words of Fischhoff, ‘if we’ve got questions, then they’ve got answers.’ This is referred to by Fischhoff as the philosophy of articulated values. An implication of this viewpoint is that if a particular respondent’s answers are different on two separate occasions, then implicitly the questions must have been different. Proponents of this paradigm focus on ensuring that questions are formulated and understood as intended, arguing that any ‘slip’ could invoke a precise, thoughtful answer to a ‘wrong’ question. In their recommendations for outcome measurement in cost-utility analysis (CUA), Johannesson et al. argue that it is ‘important to continue the work on testing to what extent QALYs are consistent with individual preferences’ (p. 288). This statement, and indeed the whole discussion in this section, implies that individual’s possess a unique set of (consistent) preferences over health outcomes. However, if such preferences are either shaped by the elicitation procedure or simply illdefined, it is plausible that a valuation task (as involved in the calculation of QALYs) may imply a different ranking of alternatives than that revealed from a direct choice task (which is how Johannesson et al. judge the validity of the valuation methods used in CUA). The preference