Choice behaviour is affected by complex factors; in addition, when facing new situations people might undergo a process of experimentation and learning, through trial and error. These represent a challenge to improve realism in modelling work. Furthermore, the choice process may present restrictions on the perception and evaluation of attributes that can vary within the population according to certain characteristics that are referred to as thresholds. These thresholds of perception can be seen as "just noticeable differences". In this sense, perceptible changes would be values above a threshold whereas those below it would not cause a reaction in the individual because utilities do not change. It may also happen that utilities are so similar that are perceived as equal by the individual. The use of binary Stated Preference (SP) surveys in which the individual can choose between only two alternatives is very frequent in applied work. However a good design requires that utilities are balanced, and this may imply cases where utilities are so close to be undistinguishable for the respondent; for this reason, it may be important to incorporate an "indifference option". A discrete choice model to adequately treat this case is developed. To model the inclusion of an indifference option thresholds are considered that are dynamic and depend on the experiences and restrictions of the respondents. Thus, it is possible to postulate that these thresholds vary in the population, according with the concept of psychological threshold in the theory of consumer choice. A discrete choice model is proposed that describes the behaviour of individuals when they confront thresholds of indifference which let them obtain similar utilities from specific situations in a SP experiment. As an important contribution, our model introduces the indifference option so that the individual is not forced to choose one alternative at first view. It is postulated that if thresholds exist they could be stochastic, random, differ among population and even be a function of socio-economic characteristics and choice conditions. This formulation allows estimation of the parameters of the threshold probability distribution starting from information about choices. The model was applied to synthetic data and to real data coming from a SP survey with rating data and found that if there is evidence about the existence of perception thresholds in the population, the use of models without this capability should lead to errors in prediction and in estimation, although this only occurs when the effect of the indifference threshold in the utility function is strong. For the covering abstract see ITRD E137145.
Samenvatting