International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
That's good enough! Satisficing in stated choice experiments
Erlend Dancke Sandorf, Danny Campbell

Last modified: 28 March 2017


In discrete choice experiments (DCEs) individuals are often faced with a sequence of choice tasks containing several alternatives described by multiple attributes taking on a number of different levels. When analyzing such data, researchers typically assume that individuals consider all aspects of each alternative and choose the alternative that maximizes their utility (McFadden, 1974). However, it is now widely accepted that individuals are not fully rational, but tend to fall back on simplifying heuristics and use rules of thumb to better manage complex and difficult choices (Gigerenzer and Gaissmaier, 2011). Such behaviors represent deviations from random utility theory and is likely to lead to misguided inferences about individuals' preferences unless we can develop models to properly address the actual choice behavior.

One of the fundamental basics of microeconomic theory is the problem of choice and the assumption of homo economicus, which describes the infinite ability of an individual to make utility maximizing choices with a full information set and complete knowledge of their preferences. Simon (1955) questions this assumption and postulates that in real life situations individuals do not have full information about all alternatives. Instead, alternatives are presented sequentially and searching for information and additional alternatives is costly. Consequently, individuals will constantly evaluate the already seen alternatives against the potential benefit of finding one that will be better. As such, individuals might choose an alternative that meets their aspiration level (i.e. an acceptable level) rather than one that maximize utility. This type of boundedly rational behavior is known as satisficing. Despite the experimental evidence that shows that individuals do make choices that are (partly) consistent with the satisficing heuristic (Caplin et al., 2011; Reutskaja et al., 2011; Stüttgen et al., 2012), the issue has only been identified and dealt with indirectly in stated choice experiments.

To increase our model's capacity to accurately predict choice we hypothesize that when reading a choice card, from left to right, an individual will choose the first alternative meeting a certain satisfactory level; consequently disregarding all remaining options. Formally, an individual chooses the first alternative that meets their satisficing requirements:

Note that if none of the alternatives meet the individual's satisficing requirements, then the individual will choose `none':

When we think about the range of satisficing requirements that might be used by individuals, there is likely to be heterogeneity in the criteria. Indeed, what is considered satisfactory for one individual will be different from that of another (e.g., while one respondent chooses the first alternative priced less than $10, another might continue searching for an alternative priced less than $5, whereas others might focus on the levels of other attributes and/or combinations of a number of attribute levels). Since the satisficing criteria adopted by individuals are unknown to the researcher, we need to accommodate all possible satisficing criteria. We can formulate this range of behaviors using a latent class framework, whereby we make probabilistic statements about the adoption of each satisficing criterion, with the full probability per individual allocated across all possible satisficing conditions. Rather than estimate marginal utilities parameters, the objective in our model is to derive the estimates of the unconditional probabilities. To consider the possibility that only a subset of the sampled individuals use the satisficing heuristic we include an additional latent class based on the common assumption of utility maximizing behavior. This allows us to probabilistically determine the proportion of individuals who exhibit rational and boundedly rational behavior. Simon (1955) points out that moving from a single choice situation to a sequence of choices might lead individuals to revise their satisficing criterion, a revision that is likely linked to institutional and value learning, as well as fatigue (Campbell et al., 2015; Czajkowski et al., 2014). We allow for respondents to use different satisficing criteria throughout the sequence of choices by segmenting the choice sequence into shorter panels. As such, this paper is a first attempt to systematically explore the use of the satisficing heuristic in the context of a stated choice experiment. We consider 944 possible satisficing rules and allow respondents to revise the rules adopted throughout the choice sequence.

We use data from a stated choice experiment conducted in the Republic of Ireland aimed at eliciting willingness-to-pay for value-added services to chicken meat. Our results show that while the satisficing heuristic was indeed used by individuals in this dataset, only a minority exhibited this type of behavior throughout the entire sequence of choices. However, breaking the sequence of choices into early and late choice tasks, as well as early, middle and late choice tasks, reveals that the use of the heuristic follows a learning and fatigue pattern as an individual progresses through the sequence. However, we remark on the dilemma this creates, since detecting satisficing decision-making is much more difficult when fewer choice observations are used. This aside, we find convincing support that ``rational'' behavior is the dominant form of decision making, which reinforces the standard modeling assumption. Nevertheless, accommodating satisficing behavior significantly impacts model fit and marginal willingness-to-pay.

Campbell, D., Boeri, M., Doherty, E., Hutchinson, W.G., 2015. Learning, Fatigue and Preference Formation in Discrete Choice Experiments. Journal of Economic Behaviorand Organization 119, 345-363.

Caplin, A., Dean, M., Martin, D., 2011. Search and Satisficing. The American Economic Review , 2899-2922.

Czajkowski, M., Giergiczny, M., Greene, W.H., 2014. Learning and Fatigue Effects Revisited: Investigating the Eects of Accounting for Unobservable Preference and Scale Heterogeneity. Land Economics 90, 324-351.

Gigerenzer, G., Gaissmaier, W., 2011. Heuristic Decision Making. Annual Review of Psychology 62, 451-482.

McFadden, D., 1974. Conditional Logit Analysis of Qualitative Choice Behavior, in:Zarembka, P. (Ed.), Frontiers in Econometrics. Academic Press, New York, pp. 105-142.

Reutskaja, E., Nagel, R., Camerer, C.F., Rangel, A., 2011. Search Dynamics in Consumer Choice Under Time Pressure: An Eye-Tracking Study. The American Economic Review, 900-926.

Simon, H.A., 1955. A Behavioral Model of Rational Choice. The Quarterly Journal of Economics 69, 99-118.

Stuttgen, P., Boatwright, P., Monroe, R.T., 2012. A Satisficing Choice Model. Marketing Science 31, 878-899.


Conference registration is required in order to view papers.