at SciVerse ScienceDirect

Journal of Air Transport Management 29 (2013) 39e45

Contents lists available

Journal of Air Transport Management

journal homepage: www.elsevier .com/locate/ ja ir t raman

Effects of the addition of simple and double decoys on the purchasingprocess of airline tickets

David Gonzalez-Prieto*, Jose M. Sallan, Pep Simo, Raimon CarrionUniversitat Politècnica de Catalunya-Barcelona Tech, C/Colom na11, Terrassa 08222, Spain

Keywords:Consumer behaviorDecoy effectCommercial aviationPurchasing process

* Corresponding author. ETSEIAT Department of Mana11, Terrassa 08222, Spain. Tel.: þ34 6303031389868

E-mail addresses: david.gonzalez-prieto@upc.egmail.com (D. Gonzalez-Prieto).

0969-6997/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.jairtraman.2013.02.002

a b s t r a c t

Air ticket purchases can be considered to be based on sequential decision-making processes with alimited number of alternatives with clearly identified product attributes. With this scenario, decoycontextual effects could potentially be useful in increasing the profitability of each choice set throughdriving the attention of the users to a particular alternative by changing their perceptions, such as theperceived attractiveness of particular options, to benefit one specific alternative. This study validates theefficiency of the addition of decoy options in increasing the proportion of users who select the targetoption in a choice set and introduces, theoretically and empirically, the use of double decoys. Threedistinct hypothetical choice sets are configured using two different types of decoy.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Currently considerable possibilities of customization are offeredto passengers in order to meet their different needs through mul-tiple services. Several combinations of these multiple attributes canbe chosen or added to a base product during the purchasing processwhich aims to provide the appropriate ticket for any occasion. Thusthe purchasing scenario is based on sequential decision-makingprocesses with a limited number of alternatives with clearly iden-tified product attributes. With this scenario, decoy contextual ef-fects could potentially be useful in increasing each choice setprofitability through driving the users’ attention to a particularalternative, and changing the users’ perceptions of the perceivedattractiveness of the multiple options, to benefit one specificalternative.

This paper aims to contribute to research on decoy options byexamining the main theories concerning the cognitive processesthat lead to a change in an individual’s perception, and by providingempirical data supporting the use of these techniques in the pur-chasing process of air tickets. Moreover, the effects of the additionof a double decoy within a choice set are theoretically and empir-ically evaluated, with implications that, to our knowledge, haveremained unexplored. The study has been developed by taking intoaccount the characteristics of air ticket purchasing processes and

nagement, Office 3.16, Colom9.du, david.gonzalez.prieto@

All rights reserved.

their specific context in order to provide guidance on possiblechanges that could enhance the profitability of the air ticket pur-chasing process. A further aim is to understand which mechanismsdrive the hypothetical benefits that stem from the introduction ofdecoy options in the choice set.

2. Theoretical background on decoy options

Within a set of alternatives, the decoy option is the alternativewhich is added to the choice set with the sole purpose of increasingthe perceived attractiveness of one of the original options. Thiseffect is referred to as asymmetric dominance and was first intro-duced by Huber et al. (1982). The decoy option does not stand as avalid alternative in the set because in most cases this option iscompletely dominated by one of the original options in the set,from now referred as the target option. However, its utility lies inthe capacity this alternative has in modifying consumer prefer-ences and the perceived attractiveness of one of the options in theoriginal set.

In Fig. 1 a particular situation is presented. For air ticket productcategories two attributes are considered: price and quality. Thereare two original options named as the competitor (option A) andthe target (option B). While the competitor presents higher levelswith the price attribute, (note that price is perceived as moreattractive as it decreases), the target presents higher levels with thequality attribute. Therefore, in the binary choice set integrated bythe option target and competitor, the dominant dimension for aparticular alternative is defined as the alternative that presents thehigher levels of the original choice set.

Fig. 1. Nearly dominated decoys in two attribute space. Asymmetrically dominatedzone and partial dominated zones are represented. Note. DR, nearly dominated rangedecoy; DF, nearly dominated frequency decoy; DD, double decoy (DR þ DF in the samechoice set); core, core set.

D. Gonzalez-Prieto et al. / Journal of Air Transport Management 29 (2013) 39e4540

In Fig. 1, two alternative decoys are introduced in order to in-crease the attractiveness of the target option. The DF decoy presentsa slightly better price than the targetoption but quite a reduced levelof quality, while the DR presents a slightly better value than thetarget in quality but quite a higher price. Both options are expectedto produce decoy contextual effects when presented in the choiceset. Note that these decoys are out of the asymmetrically dominatedzone, so they are defined as nearly dominated decoys. In this studynearly dominated decoys are used instead of the traditional asym-metrically dominated decoys. The reason is to prevent any biasderived from the presence of illogical alternatives, which, forexample, could provide more quality at the same price.

This contextual effect has important implications on a theoret-ical basis if one considers that it violates the regularity theorem(Luce,1977), which postulates that the addition of a newalternativein a set of options cannot increase the probability of choosing one ofthe original options in the set. Moreover, the axiom of the inde-pendence of irrelevant alternatives (Luce, 1959) is also violated asthe introduction of an irrelevant alternative as a decoy option in aset leads to a non-proportional reduction in the market sharecompared with that in the original option with the previousconfiguration.

It is important to note that although the effects of decoy optionshave been studied in terms of performance, defined as the capacityof a decoy option to increase the proportion of users who choosethe targeted alternative, no interaction between different decoyoptions in the same choice set has been previously reported(Gonzalez-Prieto and Lordan, 2013). The result of this interactioncould be either positive (the presence of an extra decoy wouldreinforce the already reinforced attractiveness of the target) ornegative (the extra decoy interacts with the other negativelycanceling the overall effect of the reinforcing target attractiveness).

Apart from the interaction effects related to the cognitive pro-cesses uniquely associated with decoy options, increasing thenumber of options should also be considered. It could be logicallyassumed that the more options available, the more possibilities ofbeing successful when searching for a particular product. Howeverthe introduction of an additional option in the choice set as a decoydoes not strictly increase the number of useful and valid optionswithin the choice set. The increase of available alternatives canincrease the complexity of evaluations since the complexity of thechoice task increases. Complexity can stem from two main facts:trade-offs and the number of comparisons. As the attractiveness of

alternatives rises, individuals experience conflict because of thecognitive stress associated with making trade-offs between op-tions.When this happens, the attractiveness of choosing the defaultoption or simply not choosing can increase (Dhar, 1997). Consumerresearch suggests that as both the number of options and the in-formation about options increases, people tend to consider fewerchoices and to process a smaller fraction of the overall informationavailable regarding their choices (Hauser and Wernerfelt, 1990).

Many researchers have tried to explainwhich cognitive processesexist with decoy contextual effects (Ariely andWallsten, 1995; Dharand Glazer, 1996; Pettibone and Wedell, 2000; Simonson andTversky, 1992). These studies describe different processes thatoccurwhen adding decoy options in a set: weight change, value shiftand value added (Pettibone and Wedell, 2007). The weight changemodel refers to variations in the importance assigned to the pre-sented attributes of the options when a decoy is introduced. Thevalue shift implies a change in the perceived attractiveness of theattributes in either the target or the competitor option because of theaddition of a decoy option in the choice set (Park and Kim, 2005;Pettibone andWedell, 2000).Value addedprocesses are related to theneed of providing justification, external or internal, for choosing aparticular option of the set. Adding a decoy option, that providesjustification for choosing the target over the competitor, also benefitsthe target bymaking it appear the less risky option in the choice set.

These processes are explained by different theories aiming toaccurately predict the shifts occurring when adding a decoy. It isimportant to note that there are two models which can provideexplanation for shifts in perceptions when adding decoy options ina choice set: the weight-change model assumes that when adding adecoy option within the decoy set the relative weight of the attri-butes changes. On the other hand, the value shift model postulatesthat relative weights assigned to attributes remain constant but thesubjective values, or perceived attractiveness, of each attribute shiftbecause of the presence of the decoy (Wedell, 1991). The later hasobtained more empirical support (Pechtl, 2009).

Nevertheless, a more general approach is required in this case inorder to account for both effects due to the specific nature of theseexperiments and to consider the value added processes that mayoccur. The general multiattribute utility (MAUT) framework (vonWinterfeldt and Edwards, 1986) includes both effects, as it con-siders that choice depends on a comparison of overall attractive-ness values for alternatives that result from weighted additiveintegration of dimensional values (Wedell, 1991). Moreover, anadditional component is included in order to account for value-added effects. It is represented as follows:

ATi ¼X

mWmVmi þ Ki (1)

ATi represents the overall attractiveness value of one particularalternative, Wm is the context dependent weighting of attribute m,Vmi represents the context dependent perceived attractiveness ofattributem of the alternative i, and Ki represents the value added toone particular alternative i in a particular context. This equationincorporates the three effects accounting for the weight changeeffects, the value shift in the attributes’ perceived attractivenessand the value added due to the relational properties of oneparticular alternative i.

2.1. Hypothesis

2.1.1. Weight-changeChanges in the dimensions’ weights are considered through a

preference vector whose slope is the weight of dimension 2(quality) divided by the weight on dimension 1 (price). It is

D. Gonzalez-Prieto et al. / Journal of Air Transport Management 29 (2013) 39e45 41

important to note that when aiming to favor the targeted alterna-tive the weight of its weakest attribute should decrease in order toenhance the weight of its strongest attribute or dimension (Wedell,1991). When adding a decoy alternative, an altered weightingscheme is presented resulting in an increase in the target’s likeli-hood of being chosen. With Fig. 1 focusing on the option target, onecould differentiate between the strong dimension or dominantdimension and the weak dimension for this option. Dimension 2 isstronger than dimension 1 for the target because of its higherrelative value. Regarding the interaction between the two decoyoptions, as the altered weighted scheme for each decoy choice setfavors the target by decreasing the relative weight of its weakdimension, it could be assumed that for the addition of effects, thesame would happen in the double-decoy choice set.

Hypothesis 1. The relative weight of dimension 2 is higher in eitherthe decoy set with DR or DF than in the original choice set. This effect ismaintained when the two decoy options are included in the samechoice set.

2.1.2. Value shiftWhen accounting for the change in the perceived attractiveness

or value for the different attributes there are three different the-ories providing explanation regarding the options presented inFig. 1: The range-frequency theory, the assimilation theory and thedistance-density theory (see Pechtl, 2009 for a complete review).However, only two of them are considered: the range-frequencytheory because of its importance and widespread use indesigning and analyzing decoy options, and the later, just consid-ering the density principle, because of its good adjustment toempirical results (Park and Kim, 2005; Pechtl, 2009).

2.1.2.1. Range-frequency theory. This theory, first introduced byParducci (1995) proposes that attribute values are a function of theircorresponding range and frequency values. When it is applied to thechoice set in Fig. 1, the DR decoy option extends the range ofdimension1 levelswhen comparedwith the core set,where only theoriginal options TandCare presented. This fact leads to an increase inthe range of values under DR presence because the target does notoccupy the last position in the set for this attribute.When referring todimension 2, as the range in this attribute is also extended, the tar-get’s range values for this attribute are lower in the decoy setwithDRthan in the original set. However, as the decoy is nearly dominatedand the increase in dimension 2 ismuch reduced comparedwith theincrease in dimension 1, the target option does not lose its strongposition for dimension 2, but shares it. This theory does not predictany changes in competitor attributes perceived attractiveness.

Hypothesis 2. The perceived attractiveness of dimension 1 for thetarget is higher in the decoy set with DR than in the original choice set.

Hypothesis 3. The perceived attractiveness of dimension 2 for thetarget is lower in the decoy set with DR than in the original choice set.

Hypothesis 4. The increase in the perceived attractiveness ofdimension 1 is significantly higher than the decrease in the perceivedattractiveness for dimension 2 in the decoy set with DR than in theoriginal choice set.

Regarding the inclusion of the DF option, assuming an analogousargument to that developed for the DR choice, where behavior of anearly dominated decoy was assimilated to an asymmetricallydominated decoy, the following assumptions result:

Hypothesis 5. The perceived attractiveness of dimension 2 for thetarget option is higher in the decoy set with DF than in the originalchoice set.

Hypothesis 6. The perceived attractiveness of dimension 1 for thetarget option does not change when adding a DF decoy to the originalchoice set.

Concerning the choice set where both decoys are included, theproperties of each of them are supposed to be additive. Focusing ondimension 1, the inclusion of the DF decoy option maintains thetarget’s perceived attractiveness for dimension 1 unchanged, as it isstill theworst in the choice set. However,whenDR is also included inthe choice set, the range of this dimension is increased. In this newcontext, the target option loses theworst position in this dimensionand therefore its range of values for this dimension increases.Referring to dimension 2, DF presence leads to a higher perceivedattractiveness for the target option. On the other hand, DR repre-sents a slightly better value than the target’s value for dimension 2.This fact leads to a decrease in the range value for this attribute,which compensates for the prior increase produced by the presenceof DF. Following these assumptions, two hypotheses are presented:

Hypothesis 7. The perceived attractiveness of dimension 1 for thetarget is higher in the decoy set with DF and DR than in the originalchoice set.

Hypothesis 8. The perceived attractiveness of dimension 2 for thetarget remains constant with respect to the original choice set.

Note that there is no need for a hypothesis concerning an overallbalance of the changes in the attribute values produced by this typeof decoy because its effect favoring the target is directly deducted.

2.1.2.2. Density principle. The density principle derived from thedistanceedensity theory has been applied in previous decoyresearch (Glazer et al., 1991; Pechtl, 2009). It postulates that twoobjects in a space region are perceived as less similar when a newthird object appears, located next to one of the original options. Inthis new context, there is a denser region formed by this pair ofobjects, which is perceived as a standard region in terms of attri-bute values. This standard constitutes a new subjective referencepoint from which evaluation of the isolated alternative is made. Inthis case, the changes in perceived attractiveness of particular di-mensions are only evaluated for the competitor option. Whenadapting this principle to the proposed scenarios in Fig. 1:

Hypothesis 9. The overall perceived attractiveness of the target re-mains unchanged for both types of decoy.

Hypothesis 10. The perceived attractiveness of dimension 1 for thecompetitor is higher in either the decoy set with DR or DF than in theoriginal choice set.

Hypothesis 11. The perceived attractiveness of dimension 2 for thecompetitor is lower in either the decoy set with DR or DF than in theoriginal choice set.

Regarding the double decoy choice set, the standard regionformed by the DR, the DF and the target is denser than in theprevious cases. This situation leads to stronger effects than in thesituation where only single decoy is included in the choice set.

Hypothesis 12. The overall perceived attractiveness of the targetremains unchanged for the double decoy choice set.

Hypothesis 13. The perceived attractiveness of dimension 1 forcompetitor in presence of bothDFandDR in the same choice set is higherthan in either the decoy set with DR or DF or the original choice set.

Hypothesis 14. The perceived attractiveness of dimension 2 forcompetitor in presence of both DF and DR in the same choice set islower than in either the decoy set with DR or DF or the originalchoice set.

D. Gonzalez-Prieto et al. / Journal of Air Transport Management 29 (2013) 39e4542

3. Choice set design in air tickets

The particular characteristics of the air ticket as a product orservice to be considered regarding the design of decoy options arethe following: first, a strong preference of the user for the price.There is an imbalance between the importance of the primaryattribute, the price, and the rest of attributes that could characterizethe product. Moreover, there is difficulty in identifying a repre-sentative and relevant second attribute, which could be associatedwith air tickets due to the variety of additional attributes in theform of add-on or extra services presented by different airlines.Usually the product is represented by two separate attributes.These must be represented by continuous variables or at least havesufficient values so they can be understood as such. Given itsimportance to airline tickets, price is one of the air ticket attributes.Regarding the second attribute, a continuous quality measure witha hundred points scale has been considered. In the survey used,quality is referred to as a continuous measure reported by some ofthe company’s frequent customers, where customer satisfaction isrepresented with different on-board services and features, such asthe quantity and quality of meals, seat pitch, general comfort andpersonal attention. This means that customers focus not just on onespecific attribute but more generally on service improvement dueto a higher price, which avoids particular multiple customer pref-erences that could vary across different groups of the populationwhich could bias the experiment.

Incorporating this information in Equation (1) the final model isobtained as follows:

ATi ¼ Wp$Vpi þWq$Vqi þ Ki (2)

ATi represents the overall attractiveness value of one particularalternative, Wp is the context dependent weighting of attributeprice and Vpi represents the context dependent perceived attrac-tiveness of this attribute for alternative i, either the target or thecompetitor. The same reasoning applies to the second term of theequation for the attribute quality. Ki represents the value added ofeach of these alternatives.

3.1. Justification for a no-choice option

In most experiments with this type of decoy options intended togenerate attraction effects, informants cannot opt for the no option,which in real life would be not to buy. However, in a situation ofdecision making in a real purchase, the user is not required topurchase one of the products offered within the range that is pre-sented and they cannot just buy the product or buy it elsewhere.This situation may occur when the user is not able to easily justifythe reason for choosing one alternative over the other (Dhar, 1997).The purchase is made only when there is a clear preference for aproduct or when the cost of delay or failure to complete the pur-chase is high, or the product is urgently needed (Dhar andSimonson, 2003). The probability that the user decides not tochoose is directly proportional to the difficulty of the decision-making process, usually linked to uncertainty and doubt, becauseno alternative is sufficiently relevant or attractive to the consumer.This situation leads to emotional stress (Luce, 1998) because whenthe choice is made it implies that the attractive attributes of thealternatives not selected are missed.

The inclusion of the option of not to choose depends very muchon the nature of the product. In the case of products considered aslow involvement products (e.g., eggs, coffee, bread) it makes nosense to include this option in the experiment, as the cost to theconsumer when not purchasing these products would be too highfor the option not to choose to be considered. In contrast, products

with high involvement are more likely to produce a process ofprocrastination with the user decision (e.g., video cameras, travel).With this research, due to the nature of the product as well as thepurchasing process, a no choice option is included in order toprevent any bias that could result from forced choice.

In the three option set either including DR or DF, the presence ofthe nearly asymmetrically dominated option is likely to enhancethe consumers’ confidence that the target alternative is a goodchoice (Huber and Klein, 1991; Simonson and Tversky, 1992). In theoriginal choice set, uncertainty and cognitive stress, due to the factthat each option has an advantage in one dimension and disad-vantage in the other, would probably lead to an increased popu-larity of the no-choice option (Dhar and Simonson,2003).Therefore, it is expected that the share of the no-choice op-tion is reduced when either DR or DF is included in the originalchoice set. Nevertheless the complexity of the choice process whenboth decoys are introduced at the same time could lead to cognitivestress as previouslymentioned. Comparisons between an increasednumber of options could attenuate the attraction effect generatedby these decoys in terms of reducing the proportion of the no-choice option. Therefore, in the double decoy choice set, theincreased complexity of the choice task could have no effect on theproportion of respondents who choose the no-choice option.

Hypothesis 15. The inclusion of a decoy option within the choice setleads to a reduction in the share of the no-choice option.

Hypothesis 16. The inclusion of the double-decoy within the choiceset has no effect in the share of the no-choice option.

4. Methodology

4.1. Overall description of the experiment

The participants were from a mixed sample of 466 graduatesand undergraduates from different universities (average age: 26.92,standard deviation: 11.83). This experiment had decoy type, DR, DF,or DD (double decoy) as the basic design variable, which wasmanipulated between the informants. A choice set was randomlyassigned to each of the informants. Three different choice sets wereconstructed for the same product category which was air ticketscovering the round trip between Barcelona-New York. The DRdecoy was constructed by using a slightly better value than thetarget alternative in its dominant dimension and a poorer valuethan the target in its weak dimension. The DF decoy was con-structed by using a slightly better value than the target in its weakdimension and a poorer value than the target in its strong dimen-sion. Nearly dominated decoys were introduced in order to preventcustomers from over-analyzing the choice set because of thepresence of an illogical optionwhich could be, for example, offeringthe same quality at lower prices. Any bias because of this waseliminated in order to enhance experiment reliability, especiallywhen considering that the product category in which the experi-ment was developed has suffered from consumer distrust towardthe sometimes confusing purchasing process with hidden chargesor extra fares.

The prices offered came from an early stage of the study inwhich a group of 20 undergraduates with international flightexperience identified the most common range of prices for thisroute. These ranges were reduced by a 1.5 factor in order to avoidextreme values. Decoy options were designed to fit this range ofcommon prices. All materials and instructions were presented in aformal paper survey which aimed to understand consumer pref-erences in international flights. Each choice set was presented as a3 � 2 matrix with rows corresponding to the alternatives and

D. Gonzalez-Prieto et al. / Journal of Air Transport Management 29 (2013) 39e45 43

columns corresponding to the attribute values. Participants wereasked to make their choice in the first page with no more infor-mation other than the alternatives and the instructions. Once thechoice was made, they were told to proceed to the next page wherethe attribute importance and attractiveness ratings for eachdimension and every alternative had to be rated in a 7-point Likertscales with 1 representing the lower value of attractiveness. Theuse of the 7-point Likert scales, instead of the more traditional 5-point scales or a good/neutral/bad option, allows the statisticalpower to be enhanced since it reduces scale coarseness. Coarsenessappears when a construct, that is continuous in nature, is measuredusing items that result in different true scores being collapsed intothe same category (Aguinis et al., 2009).

5. Results

5.1. Presentation and structure

The results are presented in two sections. The first examines theperformance of the different decoy options through two key points:the increase in the proportion of users who finally chose the targetoption and the variation in the proportion of users deciding not tobuy. This later point is represented by those choosing the no-choiceoption. The second examines the results of the ANOVAs conductedseparately for variations in the dimensions’ weights as well asvariations in the attractiveness ratings and in the overall attrac-tiveness inferred from Equation (2). These studies were conductedfor both the target and competitor options. Differences existing inthe relative weights between the attributes as a function of decoypresence were also examined.

5.2. Decoy performance

Results from separate chi-square analysis are reported in Table 1for both the target and the no choice option across groups. For eachtype of decoy the analysis is conducted for evaluating the change inchoice proportion as well as the change in the percentage of userswho choose not to buy.

The decoy effect should reflect a significant effect of the pres-ence of either DR or DF in the choice set on the proportion of userswho choose the target option. For the DR decoy, the predicted biasis obtained as the variation in the proportion of users who choosethe target significantly varies across groups. It is important to notethat under DR, the share distribution varies greatly from a situationin which the choice proportion is, approximately, 0.3 for the targetand 0.7 for the competitor in the core set, to a 0.65 for the targetand 0.35 for the competitor. The proportions are nearly invertedunder DR. The proportion of users who choose the decoy option inevery situation is assumed to be negligible. Referring to the inclu-sion of DF in the choice set, the effect is also significant regardingthe changes in target’s choice proportions. Concerning changes in

Table 1Choice proportions and chi-square analysis for the target and the no-choice optionin the three hypothetical choice sets.

Target option No-choice option

Choice proportion Chi-square Choice proportion Chi-square

Core 0.29 e 0.10 e

DR 0.65 25.58a 0.01 4.07c

DF 0.44 4.49c 0.01 6.01c

DD 0.48 7.12b 0.04 2.43

ap < 0.001; bp < 0.01; cp < 0.05; dp < 0.1.Note. DR, nearly dominated range decoy; DF, nearly dominated frequency decoy;DD, double decoy (DR þ DF in the same choice set); core, core set.

the selection of the no-choice option, both decoys are found to havea significant effect as expected in H15.

Regarding the double-decoy choice set, it can be observed that italso increases the proportion of users who select the target option.Nevertheless, this increase is lower than in the choice set whichonly includes DR as the decoy option and very similar to the effectobserved in the choice set which only includes DF as decoy option.Therefore, both decoys interact and the addition of DF diminishesthe overall effect of DR regarding the variation of the target’s choiceproportion. It is interesting to note that with the choice proportionof the no-choice option, no significant variation is observed in thechoice set including both decoys compared with the core set asstated in H16. The reason behind this phenomenon could be thecognitive stress or saturation which generates uncertainty in theuser and, even when both decoy options are reinforcing the targetin different ways, some users tend to direct their decision to the nochoice-option.

5.3. Shift in perceptions

Two main shifts are evaluated and presented in Table 2. Meanratings for the perceived quality of the different dimensions forboth the target and competitor alternative are presented as is thedimensional weight for both the price and quality attributes. Thefinal two columns present the mean value of overall attractivenessfor each option computed through Equation (2), not consideringthe inclusion of value added at this point, but only the interactionbetween the dimensional weights and particular attractivenessrating for each attribute for both the target and competitor. Resultsfrom the ANOVAs conducted on the full sample are reported inTable 2.

5.3.1. Range-frequency theoryFor the DR decoy, the predicted pattern of the range-frequency

theory in H2 regarding the shift in attribute values of the target isconfirmed, as an increase in the attribute value of the target’s weakdimension is observed. When a more expensive option is added,the perceived attribute value of the price for the rest of the optionsincreases. Concerning its strong dimension, quality, a decrease inthe perceived attractiveness is observed, thus confirming H3.Nevertheless, the increase in the attractiveness of its weak attributeis higher than the decrease in the dominant one as predicted in H4.For the DF decoy, an increase in the perceived quality of the target isobserved. The results confirm the logical assumption that when analternative with less quality is introduced it reinforces the alter-native with higher levels in this attribute. No significant change inthe perceived attractiveness of price is observed under DF presence.This confirms H5 and H6. No changes in the competitor attributes’perceived attractiveness is predicted by the range-frequency the-ory, although significant variations are found across groups. Eitherthe DR or DF decoy affect competitor attribute values in the samedirection: they increase the competitor’s perceived attractivenessin price and decrease its perceived attractiveness in quality. Bothdecoys have a higher price than the competitor; this fact reinforcesits position in this dimension. The opposite happens regardingquality, as the inclusion of decoy options makes the competitorattribute values in this dimension appear even lower than in thecore set.

Regarding the double-decoy choice set, an increase in theperceived attractiveness for the weak dimension of option targetis observed, confirming H7. Moreover, as hypothesized in H8 theinteraction between DR and DF among dimension 2 impliesthat the perceived attractiveness for this dimension remainsunchanged.

Table 2Mean and ANOVAs analysis for shifts in attribute values for every alternative and each dimension. Moreover, dimensional weight is also analyzed and total attractiveness foreach alternative is computed.

Mean of perceivedprice attractiveness. Vpi

Mean of perceived qualityattractiveness. Vqi

Mean of dimensionalweights. Wi

Mean of totalAttractiveness. ATi

Target Competitor Target Competitor Price Quality Target Competitor

Core 3.19 5.25 5.27 3.09 5.95 4.36 41.90 44.76DR 4.19 [30.56]a 5.56 [2.81]d 4.93 [2.98]d 2.21 [27.43]a 5.32 [16.20]a 3.88 [4.39]c 41.28 [0.09] 37.68 [19.01]a

DF 3.04 [0.63] 5.45 [1.16] 5.95 [11.83]a 2.44 [14.41]a 5.09 [26.21]a 4.13 [1.00] 39.36 [1.52] 38.09 [15.71]a

DD 4.00 [19.15]a 5.61 [24.78]a 5.23 [0.04] 2.28 [3.51]d 5.70 [2.40] 3.63 [9.49]b 42.28 [0.03] 40.11 [7.62]b

In brackets: F value.ap < 0.001; bp < 0.01; cp < 0.05; dp < 0.1.Note. DR, nearly dominated range decoy; DF, nearly dominated frequency decoy; DD, double decoy (DR þ DF in the same choice set); core, core set.

D. Gonzalez-Prieto et al. / Journal of Air Transport Management 29 (2013) 39e4544

5.3.2. Density principleThe denser space region is formed by the targetedecoy pair in

both three-option choice sets. As this region is taken as a referencepoint for judgment and alternative attractiveness valuation, thecompetitor attributes’ perceived attractiveness will change.Regarding quality, a decrease in its perceived attractiveness isobserved, as stated in H11. Moreover, H10 is also confirmed as theprice perceived attractiveness for the competitor increases when adecoy option is included in the choice set. No changes are assumedfor the target overall perceived attractiveness, as this theory con-siders no change in dimensional weights or changes in the targetattribute perceived attractiveness. H9 is confirmed because theoverall attractiveness remains unchanged for this option but, as thepreviously mentioned assumptions are not fulfilled, the hypothesiscould not be validated but just considered as a mathematicalcoincidence rather than accurate theory.

Concerning the double-decoy choice set, as the space region inwhich most of alternatives are located is denser than in the singledecoy sets, an increase in the above mentioned effect is assumed.For dimension 1, a higher increase in the competitor’s attractive-ness is observed as expected in H13. Nevertheless referring todimension 2, despite of a significant variation, H14 is not confirmedas this variation is not greater than for the single decoy choice sets.The double decoy choice set also involves no change in the overallperceived attractiveness of the target, as stated in H12. However adecrease in the overall perceived attractiveness for the competitoris found, as previously reported for either DR or DF choice sets. It isimportant to note that the value shift processes in attribute valuesoccur when either DF or DR decoys are included in the choice set.Nevertheless, none of the above theories can fully account for theeffects observed. Concerning the total attractiveness of an option, adecrease in competitor attractiveness is observed rather than anincrease in the target overall attractiveness. Therefore, the valueshift processes tend to modify and damage the competitor’sperception as the target’s perceived overall attractiveness remainsunchanged.

Table 3ANOVAs analysis for relative dimensional weight for every choice set. Differences in thedifferences in each attribute value between both alternatives are also reported.

Relative weightWprice/Wquality

Relative differences

Total perceived attractivenessdifference between target andcompetitor. ATt � ATc

Differenin priceVpt � V

Core 1.73 �2.86 �2.06DR 1.61 [0.46] 3.60 [11.02]a �1.37 [9DF 1.40 [3.39]d 1.27 [3.84]d �2.42 [2DD 1.96 [1.38] 2.18 [5.88]c �1.61 [3

In brackets: F value.ap < 0.001; bp < 0.01; cp < 0.05; dp < 0.1.Note. DR, nearly dominated range decoy; DF, nearly dominated frequency decoy; DD, do

5.3.3. Weight-changeWhen either a DF or DR decoy is included in the choice set, the

relative weight of the strong dimension over the weak dimension islikely to increase because of the reinforcement that the decoy op-tion introduces for the target’s dominant attribute. The relativeweight between both dimensions is presented in Table 3, obtainedfrom direct importance ratings for each attribute. Results from theANOVAs conducted across different choice sets are reported.

Significance in the attributes’ relative weight change is onlyfound for the inclusion of DF in the choice set. Nevertheless, nosignificance is found for the DR decoy and the double decoy. Thedirection of this relative weight change favors the target’s domi-nant dimension as predicted by the theory in the case of DF. Hy-pothesis H1 is confirmed for DF and rejected for DR and DD. Despitethis change in attribute weights for the inclusion of DF, the p-valueis relatively high. Therefore it is possible to conclude that weightchange processes do not always occur for every type of nearlydominated decoys nor for the double decoy and, for this particularcondition and experiment, the effect of weight change is, in spite ofbeing significant for a single case, reduced.

6. Discussion of the results

As reported in the results, the efficiency of the decoy options inincreasing the proportion of users choosing the target option iscompletely proven. Moreover, the inclusion of both options con-tributes to the decrease of the proportion of users who do notchoose any available option with the DR option generating thestrongest effect.

One of the aims of the study was to evaluate the processeswhich provide the explanation for the decoy effects for the specificproduct category of air tickets, and further to obtain empirical dataabout the interactions occurring among these two types of decoy.These results may not be generalizable for different product cate-gories, even when there are similarities either in purchasing pro-cesses or product structure. Although the two theories have been

overall attractiveness between target and competitor are also presented. Moreover,

ce in perceived attractivenessbetween target and competitor.pc

Difference in perceived attractiveness inquality between target and competitor.Vqt � Vqc

2.18.24]b 2.72 [3.29]d

.04] 3.51 [18.25]a

.41]d 2.95 [6.71]b

uble decoy (DR þ DF in the same choice set); core, core set.

D. Gonzalez-Prieto et al. / Journal of Air Transport Management 29 (2013) 39e45 45

adapted in this study to explain value shift, neither of them are ableto predict the observed changes in perceived attributes completely,as in previous studies (Pechtl, 2009). Both theories assume nochange in dimensional weight may occur. The density principleseems to be the most robust explanation to evaluate changes inoverall attractiveness for the different decoy options. It is importantto note that both price and quality are operationalized in a metricscale. The type of information found to be affecting the cognitiveprocesses is likely to be produced with the addition of decoy op-tions in the choice set (Sen, 1998). This fact should be consideredwhen including other dimensions, instead of a metric quality scale,to account for other particular services or add-ons.

The empirical results obtained show that the shift processesaffect the overall attractiveness of the competitor, reducing it,rather than increasing that of the target. These results are in linewith other studies, which also found shifts in perceptions for thecompetitor when introducing decoys in the choice set (Moran andMeyer, 2006; Pechtl, 2009).

Regarding the interactions between both decoys when includedin the same choice set, possible explanations for the observed dif-ferences can be suggested taking into account the results of theprevious studies. First, the lack of decrease in the proportion of theno-choice option under the presence of the double decoy could beexplained by the cognitive stress due to the addition of more op-tions rather than for trade-offs. One could assume that the no-choice proportion in the core set stems from uncertainty andstress derived from trade-offs. As one of the main effects of decoyoptions is to diminish the proportion of respondent uncertaintywhen comparing the option’s overall attractiveness, cognitivestress due to trade-offs between possible choice are assumed not tobe present for the double decoy choice set.

7. Conclusions

Although decoy effects have been reported as a stable androbust phenomenon in the literature for multiple products, theparticular characteristics of air tickets as a product categoryrequired empirical validation for evaluating its compatibility withthe use of this technique. The choice sets included in the experi-ment fit the current situation of commercial aviation market inwhich every day more passengers tend to buy the cheaper option(Mason, 2000). The contribution of this study has two main ele-ments. First, results concerning the empirical validation of both therange-frequency theory and the density principle highlight theneed for developing a more robust theory, which can account forchanges in the perceived attractiveness of the attributes for thetarget, and at the same time provide explanation about how thecompetitor is damaged by the inclusion of the decoy option morethan enhancing the perception of the target.

Moreover, new areas of research are suggested as interactionbetween different types of decoys is introduced. The aim of thisstudy is not to provide a robust assessment of the double decoyeffect but to explore it in order to provide alternative solutions forcommercial aviation purchasing processes. The study and devel-opment of synergistic decoy choice sets, in which the presence ofextra decoys reinforce the already reinforced target, would providenew lines of research, open new questions and potentially enrichthe theory.

Themanagerial implications are considerable. Airline marketingand revenue departments could increase the profitability of thesequential decisions of their purchasing processes as they imple-ment decoy options in order to drive users’ attention to those tar-geted alternative. This could be implemented without restrictioneither for the basic air tickets purchasing processes or those forextra services or add-ons airlines could offer. Moreover, the addi-tion of decoy options reduces the proportion of users who finallydecide to defer their decision so, even when they are not pur-chasing the targeted option, they are actually buying something.This fact is a better situation than the no-choice option.

References

Aguinis, H., Pierce, C.A., Culpepper, S.A., 2009. Scale coarseness as a methodologicalartifact: correcting correlation coefficients attenuated from using coarse scales.Organizational Research Methods 12, 623e652.

Ariely, D., Wallsten, T.S., 1995. Seeking subjective dominance in multidimensionalspace: an explanation of the asymmetric dominance effect. OrganizationalBehavior and Human Decision Processes 63, 223e232.

Dhar, R., 1997. Consumer reference for a no-choice option. Journal of ConsumerResearch 24, 215e231.

Dhar, R., Glazer, R., 1996. Similarity in context: cognitive representation andviolation of preference and perceptual invariance in consumer choice. Organi-zational Behavior and Human Decision Processes 67, 280e293.

Dhar, R., Simonson, I., 2003. The effect of forced choice on choice. Journal of Mar-keting Research 40, 146e160.

Glazer, R., Kahn, B.E., Moore, W.L., 1991. The influence of external constraints onbrand choice: the lone-alternative effect. Journal of Consumer Research 18,119e127.

Gonzalez-Prieto, D., Lordan, O., 2013. Contextual effects and psychological featuresinfluencing decoy options: a review and research agenda. Intangible Capital 9(1). http://dx.doi.org/10.3926/ic.398.

Hauser, J.R., Wernerfelt, B., 1990. An evaluation cost model of consideration sets.Journal of Consumer Research 16, 393e408.

Huber, J., Klein, N., 1991. Adapting cutoffs to the choice environment: the effects ofattribute correlation and reliability. Journal of Consumer Research 18, 346e357.

Huber, J., Payne, J.W., Puto, C., 1982. Adding asymmetrically dominated alternatives:violations of regularity and the similarity hypothesis. Journal of ConsumerResearch 9, 90e98.

Luce, R.D., 1959. Individual Choice Behavior. John Wiley, Oxford, England.Luce, R.D., 1977. The choice axiom after twenty years. Journal of Mathematical

Psychology 15, 215e233.Luce, M.L., 1998. Choosing to avoid: coping with negatively emotion-laden con-

sumer decisions. Journal of Consumer Research 24, 409e433.Mason, K., 2000. The propensity of business travellers to use low cost airlines.

Journal of Transport Geography 8, 107e119.Moran, S., Meyer, J., 2006. Using context effects to increase a leader’s advantage:

what set of alternatives should be included in the comparison set? Interna-tional Journal of Research in Marketing 23, 141e154.

Parducci, A., 1995. Happiness, Pleasure, and Judgment. Lawrence Erlbaum,Mahwah, NJ.

Park, J., Kim, J.K., 2005. The effects of decoy on preference shifts: the role ofattractiveness and providing justification. Journal of Consumer Psychology 15,94e107.

Pechtl, H., 2009. Value structures in a decoy and compromise effect experiment.Psychology & Marketing 26, 736e759.

Pettibone, J.C., Wedell, D.H., 2000. Examining models of nondominated decoy ef-fects across judgment and choice. Organizational Behavior and Human DecisionProcesses 81, 300e328.

Pettibone, J.C., Wedell, D.H., 2007. Testing alternative explanations of phantomdecoy effects. Journal of Behavioral Decision Making 20, 323e341.

Sen, S., 1998. Knowledge, information mode, and the attraction effect. Journal ofConsumer Research 25, 64e77.

Simonson, I., Tversky, A., 1992. Choice in context: trade-off contrast and extreme-ness aversion. Journal of Marketing Research 29, 281e295.

von Winterfeldt, D., Edwards, W., 1986. Decision Analysis and Behavioral Research.Cambridge University Press, Cambridge, UK.

Wedell, D.H., 1991. Distinguishing among models of contextually induced prefer-ence reversals. Journal of Experimental Psychology: Learning, Memory, andCognition 17, 767e778.