Does Attribute Cut-off Elicitation Affect Choice Consistency? Contrasting Hypothetical and Real-Money Choice Experiments

The use of threshold values (cut-offs) is a well-recognized simplifying strategy in real life decision-making processes. Recent stated preference studies show that strategies used by respondents in hypothetical settings are consistent with how they normally process information in real markets. Of the different discrete choice models using cut-offs, the Swait (2001) model allows different heuristics to be captured. The few applications of this soft cut-off approach mainly focus on the effects of cut-offs on attribute estimates and willingness to pay, but scant attention has been paid to the consequences of cut-off elicitation. We focused our investigation on the effects of self-reported cut-offs on choice consistency. In line with studies on context and complexity effects in choice modelling, we parameterize the scale parameter on the basis of two alternative measures related to the stated cut-offs: (1) the number of potential violations included in the choice cards, and (2) the number of cut-offs stated at the most severe level. Moreover, we investigated whether different treatments (hypothetical vs. real-money) affect cut-off elicitations, violations and choice consistency. We tested our models using an empirical dataset generated by a field choice experiment on consumers' preferences for apples. The data show that most respondents stated they had cut-off values in mind when buying apples, but, as expected, in most cases they violated them in their actual choices. Estimation of heteroskedastic logit models shows that the proposed measures related to stated cut-offs have a systematic effect on the variance of the error term, but only in the hypothetical treatment. That is, respondents were found to be less consistent as the number of potential violations or the number of cut-offs stated at the most severe level increased. In the real-money treatment, neither of the coefficients of the two measures was statistically significant. This may indicate that when choices are binding, consumers put necessary effort into each choice card, regardless of the number of potential violations, and they are more consistent.