Regret is considered an important determinant of choice-behavior in a variety of disciplines including marketing, microeconomics, psychology and the management sciences. Simply put, regret is what you experience when a non-chosen alternative performs better than the chosen one on one or more relevant attributes. Regret-based theories and models are built around the notion that individuals minimize anticipated regret – rather than maximizing utility – when choosing.

            Recently, this notion of minimizing anticipated regret as a determinant of choice has been translated into a generic discrete-choice formulation. A Random Regret Minimization-approach (RRM) has been developed for the econometric analysis of risky and riskless choices in multinomial and multi-attribute contexts, using tractable logit-type probabilities. It allows for a straightforward estimation, based on observed choices, of parameters reflecting decision-makers’ valuation of alternatives and their attributes, within a regret-based framework. By integrating the behaviorally intuitive notion of regret-minimization in an econometrically tractable model form, the recently developed RRM-model paradigm is potentially very useful for choice analysis.

Motivated by RRM’s potential as shown in recent empirical studies (Chorus et al., 2008; Chorus et al., Forthcoming), this paper provides an elaborate, both theoretical and empirical, comparison of RRM with its natural counterpart: Random Utility Maximization. We first show how, in contrast with conventional RUM-models that postulate that the utility of an alternative only depends on its own performance, RRM assumes that regret associated with an alternative depends on its performance, at each attribute level, relative to other alternatives in the choice set. As a result, RRM does not suffer from the IIA-property, even when random errors are specified as white noise. We go on to show, using a formal proof and a numerical illustration, how RRM – in contrast with conventional RUM – is able to capture the so-called ‘compromise-effect’. This choice set composition-effect is widely established in empirical consumer choice-research: people are found to be inclined to choose alternatives with a mediocre performance on relevant attributes over alternatives with a more extreme performance (very good on one attribute, and poor on another). For our empirical analyses we use a revealed parking choice dataset, and a stated route choice-dataset that was designed to optimally discriminate between RUM- and RRM-driven choice behavior. We find that RRM performs better than RUM on the given datasets by capturing the popularity of ‘compromise’ alternatives. The paper concludes by deriving from the analyses a number of managerial and policy-implications.

Chorus, C.G., Arentze, T.A., Timmermans, H.J.P., Forthcoming. Spatial choice: A matter of utility or regret? Environment & Planning Part B

Chorus, C.G., Arentze, T.A., Timmermans, H.J.P., 2008. A Random Regret Minimization model of travel choice. Transportation Research Part B, 42(1), pp 1-18