Respondent Heterogeneity and Alternative Interpretations of Scale Values in Destination Choice Models∗

Abstract

Alternative aggregation assumptions are discussed in the context of probabilistic spatial choice models for localized samples and single‐choice data. Three aggregation schemes are treated: sampling randomness, simply‐aggregated intrinsic randomness and randomly‐aggregated intrinsic randomness. The first and the last are consistent with heterogeneous populations, but simple aggregation imposes strong conditions on population homogeneity. Some implications for cognitive‐behavioral geography are considered in reference to the constant and random utility models.

The Concept of Choice and Cognitive Explanations of Spatial Behavior

“…Whether or not individual i visits site j is regarded as the outcome of a Bernoulli trial with probability pij. Elsewhere [48] it has been argued that the imputed randomness may be assigned to a priori ignorance about the individual's utility schedule (e.g., in a random utility model applied across a heterogeneous population with fixed but unknown and differing utilities [16]). This interpretation is termed sampling randomness.…”

“…The substantive interpretation of the intervening utility construct varies between the constant and random utility models, and according to whether sampling or intrinsic randomness is invoked [48]. In the former case, the scale portrays variability in individually meaningful and cognitively interpretable strict or random utilities.…”

“…For habitual choices, of course, imputed spatial choice probabilities only reflect a priori ignorance about a consumer's habits, rather than the irreducible unpredictability of individual choice implied in many psychological choice models (for example, in a Luce model with known strict utilities, choice remains unpredictable). This observation suggests that habitual choices may only be incorporated in the spatial choice paradigm under sampling randomness and not with intrinsic randomness [48].…”

“…An intrinsically random [48] interpretation of the random utility model does permit both utility density functions and their variance parameters to be interpreted at an individual level, although this is not commonly recognized in the geographic literature. Existing probabilistic formulations, insofar as they address variability rather than simply multinomial expected values, mainly stress variability due to individual differences in heterogeneous populations.…”

“…Nor does it ascribe to the former the importance implied by the cognitive perspective. I have suggested elsewhere that doubly-stochastic forms of the constant and random utility models permit intra-and inter-individual variation to be treated explicitly [48]. In the case of the random utility model this permits descriptions by distinct probability distributions of population variation in affective values and discriminal dispersions and momentary variations in response strength for particular individuals.…”