Dynamic Heterogeneous Choice Heuristics: A Bayesian Hidden Markov Mixture Model Approach

Abstract

Standard choice models implicitly assume that consumers, in order to maximize their expected utilities, compare each of the alternatives in their choice sets in terms of all available attributes. Consumer-level utility functions are frequently taken as linear, and overwhelmingly so as compensatory. However, due to limitations in information process capacity, characteristics of choice task environment and other internal or external constraints, consumers may search for satisfying alternatives rather than optimal ones by invoking other non-compensatory heuristics which free them from arduous attribute-by-attribute comparison. The question arises as to how often these non-compensatory rules are applied, and whether researchers can detect them using only standard data sources. This study aims to address two main issues regarding consumers use of decision-rules and heuristics in the real world: (1) whether they are heterogeneous across consumers and (2) whether they are changing for individual consumers over time. To these ends, we extend the standard linear compensatory rule assumption to more faithfully capture dynamic heuristic usage for each consumer. There are three reference heuristics studied in this paper, the well-known linear compensatory, disjunctive and conjunctive rules. Conditional on this known set of possible heuristics, a dynamic heterogeneous hidden Markov mixture choice model is developed to capture heuristic dynamics at the individual-level. When estimated on detergent scanner data, the proposed model offers strong evidence supporting both heterogeneity and dynamics in heuristics usage.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 2 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.