• METHODESEditerSupprimer

    Revisiting Cost Vector Effects in Discrete Choice Experiments2019

    AKAICHI F., GLENK K., MARTIN-ORTEGA J., MEYERHOFF J.Journaux et Revues (scientifiques)

    aide à la décision, analyse conjointe / dichotomique, évaluation contingente, incertitudes / biais, préférences déclarées

    Resource and Energy Economics
    Available online 11 May 2019

    • Three different cost vector treatments are compared in a choice experiment study.

    • Willingness to pay increases as the magnitude of cost vectors increases.

    • Systematic choice of cheapest non-status quo option increases with cost vector magnitude.

    • Attribute non-attendance is found to be higher for lower and higher cost vector magnitudes.

    • Choice experiment practitioners should routinely include different cost vectors in the study design.

    JEL Classifications

    1. Introduction
    The estimation of marginal utility of income in choice experiments is of crucial importance for the estimation of willingness to pay (WTP) and welfare estimates. It requires the inclusion of a monetary or cost attribute with a series of levels – the cost vector – to be defined by the researcher. Numerous studies have been concerned with (optimal) bid selection in closed-ended contingent valuation (e.g., Cooper and Loomis 1992; Boyle et al. 1998; Veronesi et al. 2011). Regarding cost vector design in choice experiments, Carlsson and Martinsson (2008, 167) noted: “[a] similar discussion on which attribute levels to attach to the cost attribute is relatively absent in the choice experiment literature”. This statement is still valid after a decade that has seen a surge in choice experiment applications for non-market valuation. Specific information on the process of selecting the cost vector (i.e. on the number of levels to use and their values) is rarely reported in environmental (public good) applications of the discrete choice experiment literature. According to Hanley et al. (2005, 228), in discrete choice experiments “[t]he researcher typically specifies [the] levels [of the cost vector] based on an educated guess as to the underlying distribution of WTP” (page 228). Mørbak et al. (2010) have been somewhat more specific and report, citing Garrod and Willis (1999), that the highest level of the cost vector should be chosen using a ‘rule of thumb’ that alternatives with the highest cost level should not be selected in more than 5%-10% of the choice tasks in which it is present. While a clear explanation for this rule is lacking, it appears to have been used to recognise that measurement of demand for a good requires identification of the choke price (i.e. the price at which demand is zero). Recent guidance on the use of stated preference methods focuses on the type of information that ought to be conveyed via the payment vehicle, but does not comment on the choice of cost or bid vector magnitude beyond the statement that “[a]mounts and payment vehicles must be credible and salient to respondents” (Johnston et al. 2017, p.328)....

    Under a Creative Commons license

  • Retour