HEALTH ECONOMICS

Health Econ. 11: 655–658 (2002)

Published online 31 May 2002 in Wiley InterScience (www.interscience.wiley.com). DOI:10.1002/hec.720

HEALTH ECONOMICS LETTERS

Decisionmaking heuristics and the elicitation of preferences:Being fast and frugal about the future

John Cairns*, Marjon van der Pol and Andrew LloydHealth Economics Research Unit, University of Aberdeen, Scotland, UK

Summary

It has been suggested that individuals employ simple decision heuristics when answering stated preference questions.Evidence from discrete choice experiments of individuals failing to trade may indicate that they employ simpledecision making heuristics. However, individuals might not trade because their preferences are not captured by therange of trade-offs they are offered. This is explored by offering a series of choices where the trade-offs implied bysubsequent choices depend on the subject’s responses to previous choices. The results suggest that individuals answerdiscrete choices without recourse to simplifying heuristics, and that information is generated on their preferencesrather than on how they make such choices. Copyright # 2002 John Wiley & Sons, Ltd.

Keywords decision making heuristics; preference elicitation; time preference

IntroductionStated preference methods are widely used in thearea of health economics. When employing statedpreference methods economists generally assumethat individuals have ‘fully formed and highlyarticulated preferences which they can quickly andaccurately access and apply to any form ofdecision problem’ [1]. It is also an implicitassumption that individuals employ compensatorydecision making – that is they ‘trade-off’ theattributes or variables in any decision. Thisposition has recently been questioned by workwhich suggests that people may use computationalshortcuts, or heuristics in their decision making.Heuristics are employed by people to simplifycognitive tasks such as judgement and decisionmaking. Gigerenzer et al. [2] have described aspecific group of heuristics which are based on theconcept of bounded rationality and are referred to

as fast and frugal heuristics. Such heuristics takeless time, and require less knowledge and lesscomputation. This work implies that individualsmay employ simple decision heuristics whenanswering stated preference questions.

Gigerenzer et al. [2] outline their view ofrationality as based largely upon fast and frugalheuristics that involve little computation and donot calculate probabilities or utilities. Fast andfrugal heuristics can guide decision making bysimple stopping rules which limit the search forinformation. This view stands in contrast to theconventional view in economics that differentvariables in a decision can be directly comparedon the basis of quantitative probabilities andutilities. Rather than weighing up the utilities ofdifferent choices, a fast and frugal decision makerwould compare the alternatives on the basis of asingle variable (the so-called one reason decisionmaking). If this variable can differentiate the

Copyright # 2002 John Wiley & Sons, Ltd.Received 18 June 2001

Accepted 17 August 2001

*Correspondence to: Health Economics Research Unit, University of Aberdeen, Foresterhill, Aberdeen AB25 2ZD, Scotland, UK.Tel.: 44(0)1224 553269; fax: 44(0)1224 662994; e-mail: j.cairns@abdn.ac.uk

alternatives then a decision is made straight away –if it cannot then the decision maker moves ontothe next variable. The evidence suggests that a fastand frugal heuristic can be as accurate as aregression model (which incorporates all thevariables in a decision) but is computationallymuch simpler. Much of the early work on fast andfrugal heuristics has concentrated on simulationstudies. These studies demonstrate how well suchheuristics can exploit the patterns of informationin real world environments when making decisionsor judgements [2]. A recent study details how GP’sappear to employ fast and frugal heuristics inprescribing decisions for lipid lowering drugs [3].

This literature on fast and frugal heuristicspotentially has two important implications for theelicitation of preferences. The first is that ratherthan measuring individual’s preferences with muchprecision the best that can be achieved is analtogether rougher approximation to these prefer-ences. This would be the case, at any rate, if theheuristic adopted by the individual results in theirmaking choices that do generally accord with theirunderlying preferences. The second potential im-plication is more damaging to the enterprise ofeliciting preferences. It may be the case, in someinstances, that little or nothing is being learntabout the individual’s preferences and rather moreis being learnt about how individuals answerquestions.

Recent studies using discrete choice experimentsprovide some evidence that is strongly suggestiveof the use of fast and frugal heuristics [4,5]. Indiscrete choice experiments individuals are pre-sented with several trade-offs between two or moreattributes with the implicit assumption thatindividuals will engage in compensatory decisionmaking. However several studies have found that alarge proportion of individuals fail to trade, thatis, they always choose the option that maximisesone particular attribute [4,5]. This pattern ofresponse is consistent with the application ofsimple decision heuristics, that is, to limit thesearch for information these individuals maycompare the two scenarios on the basis of a singleattribute only. However, it may also be the casethat individuals do compare the two scenarios onthe basis of all attributes but that the trade-offsoffered between the attributes are not largeenough. The range of trade-offs offered in adiscrete choice experiment is often relativelynarrow and individuals’ preferences may thereforenot be captured by this range.

Without further investigation it cannot bedetermined whether individuals employ simpledecision heuristics or whether the range of trade-offs offered in the experiment is too narrow. Oneway of testing this is by presenting those indivi-duals who appear to use fast and frugal heuristicswith progressively more extreme trade-offs. Ifindividuals fail to trade even when faced by veryextreme trade-offs this would suggest the use ofsuch heuristics. Continued failure to trade wouldimply either that the individual has a dominantpreference for a particular attribute or that theindividual is employing a simplifying heuristic.The dominant preference explanation can be ruledout if individuals’ preferences satisfy the axioms ofrandom utility theory on which discrete choiceexperiments are based, in particular, the axioms ofcontinuity and non-satiation. Otherwise it cannotbe determined whether people do not tradebecause they employ fast and frugal heuristics.

The context of the study is time preferences forhealth. In a previous study, Van der Pol andCairns [6] obtained responses to a series of eightdiscrete choices from 766 members of the generalpublic by means of a postal questionnaire. Eachoption specified the timing of onset and theduration of a period of ill-health, and individualswere asked to indicate which they would prefer.Forty eight per cent of respondents traded timingand duration. Whereas, 24% always chose theoption with the shortest delay, 21% always chosethe option with the longest delay and 7% alwayschose the option which involved the shortestduration of ill-health. Therefore the choices madeby just over half of the sample are consistent withthe application of simple decision heuristics, suchas pick the option offering the shortest duration ofill-health. But these responses are also consistentwith the respondent having a time preference rateoutwith the range offered. They might have tradedhad they been presented with more extreme trade-offs.

A simple test

The trade-offs faced by individuals in van der Poland Cairns [6] varied across the eight choices. Thechoices implied time preference rates ranging from�0.03–0.14. It is not possible to go back to theindividuals taking part in that study becauseparticipation was anonymous, quite apart from

Copyright # 2002 John Wiley & Sons, Ltd. Health Econ. 11: 655–658 (2002)

J. Cairns et al.656

numerous practical obstacles. However, a recentstudy by Cairns and van der Pol [7] offers anopportunity to explore further the issue of whethersimple decision making heuristics are being em-ployed in response to questions eliciting timepreferences. The central feature of the study designwas that following an initial choice between twooptions, the trade-offs implied by subsequentchoices depended on the subject’s responses. Willsubjects continue to choose so as to minimisedelay, maximise delay or minimise duration, orwill they start to trade duration and delay inresponse to the changing trade-offs they are beingoffered?

The discrete choices implied a trade-off betweenduration and timing of ill-health where the ill-health was described in terms of the chronicfatigue syndrome. Scenario A was held constantthroughout the experiment and consisted of 20days of ill-health in 2 years time. The duration ofill-health in 7 years time in scenario B was variedthroughout. An adaptive web-based questionnairewas developed. For a detailed account of thedesign of the experiment, the estimation methods,the sample, and the results, see Cairns and van derPol [7]. Responses were obtained from 203university students. Thirteen of whom expressedindifference between options A and B in the initialchoice and in two subsequent choices, indicatingthat they either did not understand the questions,or were making protest answers. Subjects werepresented with between 3 and 11 choices and onaverage received 5.5 choices. Of the 190 subjectsonly one subject failed to trade-off duration anddelay at some stage in response to the repeatedfollow-up questions. He always chose the optionwith the shortest delay before the onset of theperiod of ill-health. These findings are in markedcontrast with those of the previous study.

Discussion

One interpretation of these results is that indivi-duals answer discrete choices without recourse tosimplifying heuristics, and information is gener-ated on their preferences rather than on how theymake such choices. But a number of potentiallimitations of the analysis should be noted. Theinitial evidence of the potential use of fast andfrugal heuristics is derived from one group ofsubjects, while the ‘evidence’ against the use of

such heuristics comes from a quite different groupof subjects. Moreover, there are differences be-tween the two surveys. The first survey was apostal survey of the general public where it wasevident that there were eight choices to make andno incentive was offered to encourage participa-tion. The second survey involved universitystudents answering a web-based questionnairewhere the number of choices to be made wasunknown and there was a financial incentive toparticipate. One possibility is that the two groupsbehave differently. Several hypotheses can begenerated regarding the use of fast and frugalheuristics. It may be the case that some individualsalways employ simple decision heuristics whenfaced with a decision problem, some individualsmay never employ simple decision heuristics or itmay be the case that whether individuals employsimple decision heuristics depends on the complex-ity of the decision problem. It is also plausible thatthe use of simple decision heuristics depends onindividuals’ characteristics such as education, ageor occupation. There is evidence that decisionmaking strategies are dependent upon factorsrelated to the task, context and the individual [8].Regarding complexity, in both surveys respon-dents only had two attributes to consider – theduration and timing of a period of ill-health. Forsuch a task respondents may be able to use all ofthe available information and may be able to tradeoff attributes but had the task been more complexgreater use of fast and frugal heuristics might thenhave been observed.

Another consideration stems from differences inthe instruments. Some respondents to the firstsurvey may have adopted simple heuristics becausethey knew how many questions they were beingasked to answer and because of the lack of anexplicit incentive to complete the task. Also thequestions differed slightly between surveys. In theweb-based survey option A was held constant andoption B was varied, whereas there was nocommon option in the eight postal choices. If thismade the questions harder to answer it might haveencouraged greater recourse to fast and frugalheuristics. Finally, this paper only examinedspecific fast and frugal heuristics which followedlexicographic rules. However, individuals mayhave employed other decision making strategies.

The comparison of the results from these twostudies is suggestive rather than conclusive. Ithighlights the limited knowledge regarding what istaking place when individuals’ preferences (in this

DecisionMaking Heuristics 657

Copyright # 2002 John Wiley & Sons, Ltd. Health Econ. 11: 655–658 (2002)

case time preferences) are elicited, and should be astimulus for further research. The need for suchresearch has been highlighted by a number ofresearchers [1,10].

This paper examined the use of heuristics withrespect to one particular elicitation method. It canbe hypothesised that different elicitation methods(such as discrete choice and open-ended questions)encourage the use of different heuristics. Gigerenzeret al. [2] claim that people hold an adaptive toolboxof heuristics in their head for different tasks. Thereis indeed evidence of systematic differences betweendifferent elicitation methods [9]. Future researchshould explore whether the influence of differentdecision making heuristics could explain whyconsistent differences are reported to emerge fromthe use of different elicitation methods.

Acknowledgements

Financial support from the Chief Scientist Office of theScottish Executive Health Department and the MedicalResearch Council is gratefully acknowledged. The viewsexpressed in this paper are those of the authors and notnecessarily those of the funders.

References1. Loomes G. Probabilities versus money: a test of

some fundamental assumptions about rationaldecision making. The Economic Journal 1998; 108:477–489.

2. Gigerenzer G, Todd PM, and the ABC ResearchGroup. Simple Heuristics That Make Us Smart.Oxford University Press: Oxford, 1999.

3. Dhami M, Harries C. Fast and frugal versusregression models of human judgement. Thinkingand Reasoning 2001; 7: 5–27.

4. Bryan S, Buxton M, Sheldon R, Grant A. Magneticresonance imaging for the investigation of kneeinjuries: an investigation of preference. HealthEconomics 1998; 7: 595–604.

5. Ryan M, McIntosh E, Dean T, Old P. Trade-offsbetween location and waiting times in the provisionof health care: the case of elective surgery on the Isleof Wight. Journal of Public Health Medicine 2000;22: 202–210.

6. Pol M van der, Cairns J. Estimating time prefer-ences for health using discrete choice experiments.Social Science and Medicine 2001; 52: 1459–1470.

7. Cairns J, Pol M van der. Discrete Choice withRepeated Follow-up: A Web-based Experiment. HEBWorking Paper 30, University of Bergen: Bergen,2001. (http://orion.uib.no/sefos/HEB/)

8. Payne JW, Bettman JR, Johnson EJ. The AdaptiveDecision Maker. Cambridge University Press: Cam-bridge, 1993.

9. Bateman IJ, Langford IH, Rasbash J. Elicitationeffects in contingent valuation studies. In ValuingEnvironmental Preferences: Theory and Practice ofthe Contingent Valuation Method in US, EU, andDeveloping Countries, Bateman IJ, Willis KG (eds).Oxford University Press: Oxford, 1999.

10. Ryan M, Scott DA, Donaldson C. Valuing healthcare using willingness to pay: a comparison of thepayment card and dichotomous choice methods.Health Economics Research Unit: University ofAberdeen, 2001.

J. Cairns et al.658

Copyright # 2002 John Wiley & Sons, Ltd. Health Econ. 11: 655–658 (2002)