Good things come to those who don’t wait : an experiment on loss aversion

Good things come to those who don’t wait: An experiment on loss aversion

Student Yoeri Min Student number 10599088

University University of Amsterdam Faculty Economics and Business Study program Business Administration (MSc)

Specialization Marketing

First supervisor Mr. A. el Haji, MSc Second examiner Mr. J. Labadie, MSc Date of submission 29th of August 2015

STATEMENT OF ORIGINALITY

This document is written by Yoeri Min who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

ACKNOWLEDGEMENT

This thesis is submitted in partial fulfilment of the Master’s degree in Business Administration, specialization Marketing of the University of Amsterdam.

First, I want to thank Anouar El Haji for providing all the guidance I needed to finish my Master’s thesis in Business Administration. The mutual enthusiasm in waiting lines and irrational behavior resulted in valuable discussions that encouraged and motivated me to deliver the best result possible. A quick thank you goes out to Jorge Labadie who offered me the possibility to change from primary supervisor and subject, so that I could pursue my ambitions and interests. Last, I want to thank the program manager of the Aviation Department of the Amsterdam University of Applied Sciences, Geert Boosten, for all the resources that were made available for my Master’s thesis.

Have fun reading the thesis! Kind regards,

ABSTRACT

Waiting for a service is classified as an undesirable activity, yet not all consumers take full advantage of the benefits of Self Service Technology (SST) to reduce their waiting time. This study examined the moderating role of message framing in the relationship between communicated time savings and SST usage. An experimental vignette study was conducted (N = 292). As hypothesized, communicating large time savings yielded a higher percentage of SST usage than communicating small time savings. However, this positive relationship was not found to be moderated by message framing. Further analyses show that perceived usefulness mediates the relationship between communicated time savings and SST usage. The perceived usefulness of SST increases when larger time savings are communicated, and a consumer perceiving SST to be highly useful is more likely to use SST. Furthermore, message framing was found to moderate the relationship between perceived usefulness of the SST and SST usage. Respondents perceiving SST as useless were found to be more affected by framing time savings as a loss than respondents perceiving SST to be highly useful. Thus, framing time savings as a loss yields higher percentages of SST use, but only for respondents perceiving those time savings as futile benefits. The theoretical and managerial implications of the study are discussed, after which suggestions for further research are provided.

TABLE OF CONTENT STATEMENT OF ORIGINALITY ... i ACKNOWLEDGEMENT ... ii ABSTRACT ... iii TABLE OF CONTENT ... iv LIST OF TABLES ... vi

LIST OF FIGURES ... vii

1. INTRODUCTION ... 1

1.1 Waiting lines ... 1

1.2 Self Service Technology ... 1

1.3 Encouraging the use of Self Service Technology ... 2

2. LITERATURE REVIEW ... 3

2.1 Waiting line psychology... 3

2.1.1 Waiting line satisfaction ... 3

2.1.2 Waiting line experience ... 4

2.1.3 Individual differences... 6

2.2 Self-service economy and technology ... 6

2.2.1 Attitude and behavioral intentions towards SST ... 7

2.2.2 SST characteristics ... 8

2.2.3 Situational influences ... 8

2.3 Prospect theory ... 9

2.3.1 Prospect theory and applications ... 11

2.3.2 Prospect theory and field experiments ... 12

2.4 Problem definition ... 13

2.3.1 Problem statement ... 13

2.3.2 Delimitations of the study ... 13

3. THEORETICAL FRAMEWORK ... 15

3.1 Conceptual framework ... 15

3.2 Hypotheses ... 15

4. METHOD ... 17

4.1 Pre-test time savings ... 17

4.1.1 Participants ... 17 4.1.2 Procedure ... 17 4.1.3 Results ... 18 4.2 Pre-test framing ... 19 4.2.1 Participants ... 19 4.2.2 Design... 19 4.2.3 Treatments ... 19 4.2.4 Procedure ... 20 4.2.5 Data ... 21 4.2.6 Results ... 21

4.3 Experimental vignette study ... 22

4.3.1 Participants ... 22

4.3.2 Design... 22

5. RESULTS ... 26

5.1 Data ... 26

5.2 Analyses ... 27

5.2.1 Testing the hypotheses ... 28

5.2.2 The mediating role of perceived usefulness ... 29

5.2.3 The moderating role of framing ... 30

6. DISCUSSION ... 32

6.1 Framing time savings ... 32

6.2 Mediating effect ... 33

6.3 Interaction effect ... 34

7. DISCUSSION ... 35

7.1 Academic relevance ... 35

7.2 Managerial implications ... 35

7.3 Limitations and further research ... 36

REFERENCES ... 38

APPENDICES ... 43

Appendix I Interview script Amsterdam Airport Schiphol ... 43

Appendix II Pre-test framing ... 46

Appendix III Experimental stimuli ... 48

LIST OF TABLES

Table 1 – Means, standard deviations, and correlations pre-test time savings ... 18

Table 2 – Factorial design pre-test framing (2x1) ... 19

Table 3 – Means, standard deviations, and correlations pre-test framing ... 21

Table 4 – Factorial design experimental vignette study (2x3) ... 23

Table 5 – Means, standard deviations, and correlations experimental vignette study ... 27

Table 6 – Results experimental vignette study ... 28

Table 7 – Logistic regression results of framing and time savings, N = 292 ... 29

LIST OF FIGURES

Figure 1 – The value function, proposed by Kahneman and Tversky (1979) ... 10

Figure 2 – The probability weighting function, proposed by Kahneman and Tversky (1992) ... 11

Figure 3 – Conceptual framework ... 15

Figure 4 – Pre-test treatment 1: Save time ... 20

Figure 5 – Pre-test treatment 2: Don’t lose time ... 20

Figure 6 – Results pre-test framing ... 22

Figure 7 – Results experimental vignette study I ... 27

Figure 8 – Results experimental vignette study II ... 28

Figure 9 – Interaction effect of framing and perceived usefulness ... 31

1. INTRODUCTION

Waiting is a ubiquitous activity. People wait for busses, people wait at the supermarket, people wait at airports, and so forth. Furthermore, people dislike waiting. Because of psychological factors such as delay gratification (e.g. Mischel, Shoda and Rodriguez, 1989) and consciousness of time (e.g. Maister, 1985), people become impatient, bored, anxious and even angry while waiting (Schwartz, 2011). Managing waiting lines is therefore classified as an important marketing challenge (Kostecki, 1996).

1.1 Waiting lines

The importance of tackling the waiting line challenge is illustrated by a number of factors. First, referring to a quote of Benjamin Franklin (1748): “… time is money”. Time spent waiting could have been used more

effectively, introducing a cost of waiting (DeVoe, Pfeffer and Kozlowski, 2011). Second, objective waiting time, subjective waiting time and serving time affect customer satisfaction (Tom and Lucey, 1997). In fact, some customers dislike waiting so much, they hire people to wait in line for them (ABC News, 2014). Third,

according to Kostecki (1996), dealing with waiting lines is an integral part of time management skills in service organizations. Consequently, lacking proper (waiting) time management could result in a competitive

disadvantage. Last, waiting lines remain an unresolved issue in most organizations, while being an inevitable result of the business model in others (Ford, Sturman and Heaton, 2011).

1.2 Self Service Technology

Self Service Technology (SST) is used to decrease wait time before being served. SST allows consumers to “engage in all or a portion of the provision of a service or product” (Castro, Atkinson and Ezell, 2010, p. 4). The introduction of SST provides benefits for consumers. Making use of SST reduces the temporal, psychological, or economic customer costs of a transaction (Sneath, Kennett and Megehee, 2002; Pujari, 2004; Curran and Meuter, 2007; Cunningham, Young and Gerlach, 2008). For example, SST often allows business to be available 24 hours a day, rather than being bound to traditional working hours (Castro et al., 2010). SST provides significant benefits for businesses as well. The introduction of SST improves productivity, efficiency and capacity by reducing labor costs, optimally using floor space, faster responses to consumer enquiries, and increasing throughput capacity (Dabholkar, 1996; Walker et al., 2002; Meuter, Ostrom, Bitner and Roundtree, 2003; Liljander, Gillberg, Gummerus and Van Riel, 2006; Gelderman, Ghijsen and Van Diemen, 2011). For example,

of advantages for both businesses and consumers, SST grows rapidly, and will likely continue to do so. The total spending on self-service kiosks increased from $436 million in 2006 to an estimated $1.3 trillion in 2011 (Datatrend, 2010).

SST is not always beneficial, as it has not been fully embraced by society yet. For example, after increasing the number of self-checkout systems at the expense of staffed checkouts, Wal-Mart discovered that longer lines began forming at its staffed checkouts to deal with consumers using complicated and tedious methods of payment, such as coupons and price matching (Wall Street Journal, 2014). In general, new technology such as self-check-in desks at airports and self-checkout systems at retailers require significant product and behavioral changes, keeping consumers from using SST (Gourville, 2006). Although it seems that consumers initially avoid SST, Chen, Chen and Chen (2009) show that once consumers try SST and perceive it as satisfactory, they are eager to use it in the future.

1.3 Encouraging the use of Self Service Technology

By embracing SST designed to fulfil customer needs, strategically managing floor plans and introduction, organizations can take full advantage of the rapidly growing self-service economy. It would therefore be logical to encourage the use of SST. However, some organizations assume waiting as an inevitable part of their business model, and focus on improving waiting line satisfaction and experience using external stimuli. This is best illustrated in the amusement park industry and sporting events (e.g., Heger, Offermans and Frens, 2009; Baker and Jones, 2011; Alexander, MacLaren, O’Gorman and White, 2012).

The bulk of SST related studies focus on customers’ static attitudes toward the technology (e.g., Sneath et al., 2002; Beatson, Lee and Coote, 2007; Curran and Meuter, 2007; Cunningham et al., 2008). To date, no research has examined whether external stimuli that are used to improve waiting line satisfaction and experience could be used to influence consumers’ choice between SST and personal service. More specifically, no research has been conducted in which the effects of loss aversion by framing time savings, could encourage the use of SST. This leads to the following research question: “To what extent can framing of saved time encourage the use of SST?”

2. LITERATURE REVIEW

This thesis explores whether framing of time savings affect consumer preference for SST. The choice for SST depends on three elements: waiting line psychology, SST characteristics, and the effects of prospect theory. Therefore, relevant literature on these three subjects is reviewed. The literature review is followed by a conceptual framework and hypotheses. Both qualitative and quantitative methods are used to test the effect of framing time savings on consumer preference for SST. The research methodology is presented along with the data analysis and results. Last, research findings, implications and limitations are discussed, providing suggestions for further research.

2.1 Waiting line psychology

A vast amount of scholars have studied the mathematical theory of waiting lines (Eiselt and Sandblom, 2012). These studies focused on the objective reality of queue management techniques such as adding servers, altering queueing order, and increasing serving times (Maister, 2005). However, the experience of waiting had been relatively neglected: waiting line psychology (Maister, 2005). Maister (1985) was the first to provide seven propositions related to subjective waiting time. Although Jones and Peppiatt (1996) question the generalizability and interrelationships between the propositions of Maister (1985), the original propositions provide a clear starting point of waiting line psychology.

2.1.1 Waiting line satisfaction

As defined by Tse and Wilton (1988), customer satisfaction is generally associated with meeting or surpassing customer expectation of products and services supplied by a company. In contrast to objective waiting line length, cost and efficiency, subjective waiting experience is often not part of this evaluation (Norman, 2008). However, waits are often inevitable, thereby forming an important factor in the overall service experience (Van Riel, Allard, Semeijn, Ribbink and Bomert-Peters, 2012).

Empirical studies have shown that objective waiting time, subjective waiting time and serving time affect customer satisfaction (Tom and Lucey, 1997). With regard to the first, Taylor (1994) shows that long objective waits increase uncertainty and anger, particularly when the delay can be accounted to the service provider. With regard to the second, research shows that satisfaction increases monotonically when consumers perceive waiting time to be less than expected (Kumar, Kalwani and Dada, 1997; Davis and Heineke, 1998).

the wait, but increases toward the end of the wait (Kumar et al., 1997; Davis and Heineke, 1998). A more recent study related to serving time by Lin, Xia and Bei (2014) found that consumers prefer saving an extra two minutes on one occasion to saving one minute on two occasions. “As consumers find it more difficult to

conceptualize instances of saving time”, consumers prefer immediate time savings over more distant ones (Lin et al., 2014, p. 10). Furthermore, consumers prefer waiting an extra minute two times to waiting an extra two minutes one time.

2.1.2 Waiting line experience

In exploring paradigm shifts in the field of marketing, Achrol and Kotler (2012) argue that consumer satisfaction is achieved by having satisfactory experiences, filtered through our senses. The marketing paradigm shift introduced a stronger focus on customer waiting experience, through managing the perceptions of waiting (e.g. Dawes and Rowley, 1996; Friedman and Friedman, 1997; Flaherty, Freidin, and Sautu, 2005; Gnoth, Bigné and Andreu, 2006). To manage consumer waiting perceptions, Norman (2008) developed seven design principles for waiting lines.

First, emotions strongly influence the waiting experience. Emotions influence the judgement of consumers (Angie, Connelly, Waples and Kligyte, 2011). It is therefore important to manage emotions that can influence the waiting experience. Consumers’ emotional experience is affected by the service environment, including physical ambient elements such as lighting, temperature and music, even as employee interaction and crowding (Andreu et al., 2006; Chien and Lin, 2014). Positive perceptions of the service environment have a positive effect on positive emotions. Following Norman (2008), waiting line environments therefore need to be “bright and cheery, attractive and inviting”, both in terms of physical appearance and employee behavior.

Second, it is essential to provide clear and unambiguous communication concerning waiting line purpose, length, and requirements (Norman, 2008). Osuna (1985) argues that consumers dislike waiting, especially when waiting time is uncertain. Anxiety and stress increase due to a sense of waste and uncertainty in terms of remaining waiting time (Osuna, 1985). It is therefore important to manage anxiety, stress and

uncertainty through real time, either explicit or implicit, information concerning waiting time (Osuna, 1985). Furthermore, Conte, Scarsini and Sürücü (2014) show that many consumers react irrationally to waiting lines and information displays. Many consumers join the shorter queue when faced with two waiting lines; even though information is displayed indicating the longer queue has a faster service speed (Conte et al., 2014).

Third, waiting should be perceived as appropriate, both in its cause and duration. A clear reason for the wait can be tolerated as long as its duration is appropriate to the reason (Norman, 2008). Furthermore, waiting might be tolerated when the service provider clearly shows it is trying to minimize waiting time, for example by staffing all service positions with employees. Buell and Norton (2011) show that appropriateness of the wait can be increased by introducing labor illusion. Labor illusion is defined as the mere appearance of effort. Their results imply that the mere signaling of effort would be enough for consumers to prefer longer waits to those with shorter waits, even when the results are the same (Buell and Norton, 2011).

Fourth, by setting accurate expectations, perceptions of fairness are maintained (Maister, 1985). Besides, exceeding expectations increases consumer satisfaction (Norman, 2008). It is better to pleasantly surprise consumers with a shorter waiting time than expected, than angering them with a promised time that was not met (Norman, 2008). Furthermore, one could manage expectations through increasing the feeling of progress (Soman and Shi, 2003). Soman and Shi (2003) show that consumer choice is driven by the perception of

progress towards the goal. In a waiting line context, the faster the line moves, the more favorable the expectations.

Fifth, occupied time passes more quickly than unoccupied time (Kostecki, 1996). Occupying idle time is therefore one way to manage consumers’ waiting perceptions (Carmon, Shanthikumar, and Carmon, 1995). To illustrate, Borges, Herter and Chebat (2015) show that, depending on the type of content, TV screens reduce the perceived waiting time and increase the consumers’ waiting satisfaction, even when the objective waiting time remains constant. The shape of the waiting line, its visibility and rapidity of movement also influence the consumer’s perception of waiting time (Lovelock, 1988). For example, waiting lines in Disney’s theme parks curve to let them appear shorter. Moreover, these curves often hide parts of the line ahead (Kostecki, 1996).

Sixth, similar to Maister (1985) waiting should be fair (Norman, 2008). Negative emotions are triggered when people wrongfully and unjustly join a queue (Sasser, Olsen and Wyckoff, 1979). Therefore, many

organizations provide each consumer with a number, initiating an ordering process (Maister, 2005). This ordering process could either be based on ‘first in, first out’, or based on the importance of consumers. Whatever priority rules apply, rules must “match with the consumer’s sense of equity, either by adjusting the rules or by actively convincing the client that the rules are indeed appropriate” (Maister, 2005, p. 7).

Last, waiting experiences can differ from the memory of waiting experiences. Redelmeier and

duration of the experience varied, lengthy procedures were not remembered as particularly aversive. Instead, painful medical procedures are determined by “the intensity of pain at the worst part and the final part of the experience” (Redelmeier and Kahneman, 1996). Supported by follow-up studies, the results show that the memory of an experience is more important than the experience itself. Effective strategies have been developed for influencing these memories, including finishing strong, segmenting the positive while combining the negative, get bad experiences over with as soon as possible, and building commitment (Chase and Dasu, 2001). All can be operationalized in waiting lines by starting strong, ending strong, and burying unavoidable unpleasant aspects in the middle (Norman, 2009).

2.1.3 Individual differences

A cross-national study of Flaherty et al. (2005), states that differences in experienced time are not caused because of a difference in peoples’ personalities, but because of a difference in circumstances. In other words, within the same environment, under the same circumstances, all people should perceive time to pass equally fast. However, a longitudinal examination of Sweeny, Andrews and King (2014) provides the first empirical evidence in experienced time differences, caused by peoples’ personalities. Because dispositional optimists1 reflect a positive outlook on the future, while defensive pessimists2 tend to embrace pessimism as a motivating and reassuring mindset, dispositional optimists were less anxious with waiting than defensive pessimists and people uncomfortable with uncertainty (Sweeny et al., 2014). Still, the results suggest that waiting is most difficult at the start and end of a waiting period, irrespective of the personality. This corresponds with the ‘rosy’

retrospection theory of Mitchell, Thompson, Peterson and Cronk (1997), which states that any unpleasantness suffered along the way is minimized as long as the overall outcome is sufficiently pleasurable.

2.2 Self-service economy and technology

In advanced capitalist societies, technical innovation resulting in cheaper and simpler machinery, and rising labor costs encouraged the emergence of the self-service economy and technology (Gershuny, 1978). SST is defined as “technological interfaces allowing customers to produce services independent of involvement of a service employee” (Zeithaml, 2009, p. 56). Self-service involves every phase of the customer journey: pre-sales,

1

Dispositional optimism is defined as a global expectation that more good (desirable) things than bad (undesirable) will happen in the future (Scheier, Carver and Bridges, 1985).

2

Defensive pessimism “captures a tendency to embrace pessimism as a motivating and ultimately reassuring mindset” (Sweeny et al., 2014, p. 1016).

sales and post-sales. Consumers are becoming more and more demanding in all three phases in terms of speed, friendliness, interaction and transparency. Fortunately, the introduction of SST improves productivity and efficiency by reducing waiting lines, optimally using available floor space, and increasing the available throughput capacity (Dabholkar, 1996; Meuter et al. 2003; Liljander et al. 2006; Gelderman et al. 2011).

Self-service is available in various forms. Examples include ATMs, airport self-service check-in kiosks, food-ordering kiosks, self-checkout machines in supermarkets, self-service gasoline stations, ticket vending machines, and various internet and/or telephone-based SST such as online hotel booking and telephone banking (Castro et al., 2010). According to Van Belleghem (2013), 70% of consumers expect a company website to include a self-service application. Furthermore, 40% of consumers prefer self-service to human contact for their future contact with companies (Van Belleghem, 2013).

Some companies even force consumers to use SST. For example, departing passengers from Exeter airport of the low-cost regional airline Flybe are required to either attend the self-service kiosk, or print their boarding pass using e-mail, mobile phone or website. Then, passengers are required to use the automated baggage check-in for hold baggage (Exeter airport, 2015). Mandatory self-service check-in does not only apply to low-cost airlines. In fact, as of July 2014, national carrier Malaysia Airlines introduced a mandatory self-check-in requirement for all economy class passengers flying out of Kuala Lumpur International Airport (Malaysia Airlines, 2014). Although companies increasingly replace full-service for SST, Reinders, Dabholkar and Frambach (2008) found that forcing people to use SST negatively affects attitudes toward using SST and negatively affects attitudes toward the service provider.

2.2.1 Attitude and behavioral intentions towards SST

The introduction of SST altered consumer service satisfaction and experience dramatically (Bitner, Brown, and Meuter, 2000). According to Dixon and Walsman (2014), one must design and manage services that are trusted, emotionally positive and memorable through altering the sequences, duration, choices and attributions of the provided service. Giving consumers a sense of control improves service satisfaction (Dixon and Walsman, 2014). SST provides additional control and reduces the waiting time perception, but not in all cases. For example, results from Kokkinou and Cranage (2013) show that longer SST processing times and higher failure rates led to longer waiting times, especially when customers demand for the product or service was high. Furthermore, although SST reduces labor costs and attract new customers (Bitner, Ostrom and Meuter, 2002),

Selnes and Hansen (2001) show that the lack of interpersonal contact, and the associated weakening of the social bond between the consumer and the company, might reduce customer loyalty.

In 1996, Dabholkar was one of the first to investigate how consumers evaluate SST. He proposed two models: an attribute model based specifically on what consumers would expect from SST, and an overall affect model based on the feelings toward the use of SST. Since then, many scholars have researched the attitudes towards, and behavioral intentions to use SST in a variety of contexts ranging from airlines (e.g. Liljander et al., 2006), personal banking (e.g. Snellman and Vihtkari, 2003), and retail (e.g. Weijters, Rangarajan, Falk, and Schillewaert, 2007) to hotels (e.g. Beatson, Coote, and Rudd, 2006). All studies focused on the attitudes and/or behavioral intentions toward SST, to examine the determinants of those attitudes and intentions. Although a meta-analysis of Meuter, Bitner, Ostrom and Brown (2005) show that SST characteristics and individual differences form the basis of those attitude and intention determinants, Wang, Harris and Patterson (2012) argue situational influences form an important determinant as well.

2.2.2 SST characteristics

The static attitude toward SST is formed through the characteristics perceived usefulness, ease of use, risk, control, and fun (Wang et al., 2012). Elliot, Meng and Hall (2012) confirm that perceived usefulness has a direct influence on consumer’s attitude toward SST (Walker and Johnson, 2006), even as ease of use and risk (Curran and Meuter, 2005). Furthermore, the more sense of control, the more positive the attitude towards SST (Lee and Allaway, 2002). Fun also has a direct influence on consumers' attitude toward using SST (Weijters et al. 2007). Curran and Meuter (2007) recognize the importance of fun in the adoption process of SST, as fun was found to be more important than utility. Concluding, when SST is perceived as useful, easy to use, safe, controllable and enjoyable, consumers are more likely to have positive attitudes and associated behavioral intentions toward the respective SST.

2.2.3 Situational influences

Little research recognizes that external stimuli also affect consumers’ decisions to use or not use SST. Wang et al. (2012) argue that perceived waiting time, perceived task complexity, and companion influence are three situational factors that affect a consumer’s decision between using SST and personal service. First, irrespective of the prior attitudes toward SST, “consumers always compare queues, looking for the shorter one to minimize waiting time” (Wang et al., 2012, p. 66). Second, the type and number of products/services purchased influences

the perceived task complexity. The higher the number of items, the higher the perceived task complexity, the less likely the consumer is tempted to use SST (Wang et al., 2012). Last, companions can influence the choice for SST. For example, older customers are more likely to use an SST when they are accompanied by children. In addition, young people are often eager to use SST to impress their peers (Wang et al., 2012). Collier, Moore, Horky and Moore (2015) provide partial support for the study of Wang et al. (2012) as their study found four situational variables that strongly influence consumers’ SST decisions. These situational variables include order size, wait-time tolerance, location convenience, and employee presence, but did not consider the presence of store employees, the mood of the consumer, and the level of personal assistance during a self-service transaction (Collier et al., 2015).

2.3 Prospect theory

In a response to the psychological inadequacies of expected utility theory, at that time the pillar of standard decision theory, Kahneman and Tversky developed prospect theory (1979). Prospect theory refers to the behavioral economic theory in which “decision making under risk can be viewed as a choice between prospects or gambles” (Kahneman and Tversky, 1979, p. 263).

Prospect theory includes four elements: reference dependence, loss aversion, diminishing sensitivity and probability weighting (Barberis, 2013). First, instead of being defined by total wealth, value is defined by gains and losses, measured relative to a reference point.

Second, loss aversion is introduced: the tendency to strongly prefer avoiding losses to acquiring gains (Kahneman and Tversky, 1984). The theory assumes that losses have greater impact on preferences than gains (Tversky and Kahneman, 1991). Psychologically, losses are twice as powerful as gains. Formally, this effect is represented by having the value function steeper for losses than for gains, resulting in an overall decision-theoretic picture represented by an S-shaped curve in figure 1.

Figure 1 – The value function, proposed by Kahneman and Tversky (1979), adapted from Barberis (2013).

Third, the function is concave in terms of gains, while it is being convex in terms of losses, known as diminishing sensitivity. The S-shaped asymmetry implies a bigger impact of losses than of gains, in case of the same variation in absolute value. For example, replacing a $100 dollar gain with a $200 dollar gain has a significant utility impact, while replacing a $1,000 dollar gain with a $1,100 dollar gain has a much smaller impact (Barberis, 2013). Because of the concavity over gains, people prefer a certain gain of $500 dollar to a 50% chance of gaining $1,000 dollar. On the other hand, people prefer a 50% chance of losing $1,000 dollar to a certain loss of $500 dollar (Barberis, 2013).

The fourth component of prospect theory is called probability weighting. Prospect theory assumes people “do not weight outcomes by their objective probabilities, but rather by transformed probabilities or decision weights” (Barbaris, 2013, p. 176). The solid line in figure 2 represents the weighting function proposed by Kahneman and Tversky (1991). The dotted line in figure 2 represents the expected utility benchmark. As can be seen in the figure, probability weighting causes low probabilities to be overweighed, while underweighting high probabilities. For example, people prefer a 0.1% chance of winning $5,000 dollar to a certain gain of $5 dollar. Furthermore, people prefer a certain loss of 5$ dollar to a 0.1% chance of losing $5,000 dollar (Barberis, 2013).

Figure 2 – The probability weighting function, proposed by Kahneman and Tversky (1992), adapted from Barberis (2013).

Blavatskyy extends prospect theory by including “decision problems where outcomes may not be measurable in monetary terms” (2011, p. 128). In other words, the effects of prospect theory do not only hold for decision problems involving money, but do also hold for nonmonetary outcomes such as loss of a relative, loss of faith, time, reputation, prestige, and entitlements (Blavatskyy, 2011). For example, applying prospect theory to waiting lines, people should prefer a 1% chance of saving 100 minutes over a certain gain of 1 minute. Furthermore, people should prefer a certain loss of 1 minute over a 1% chance of losing 100 minutes.

2.3.1 Prospect theory and applications

Prospect theory has broad applications ranging from psychology and economics to politics and sociology (Mayer, 1992). In economics, prospect theory is well reflected by research on labor-management negotiations, consumer choice and mental accounting (Mayer, 1992). Prospect theory is closely related to alternative

abnormalities in behavioral economics, such as the endowment effect (Thaler 1980), status quo bias (Samuelson and Zeckhauser, 1988), and observed discrepancies in willingness to pay (Hanneman, 1991). Although

significant status quo bias effects have been found in field experiments (Samuelson and Zeckhauser 1988), little empirical field research found framing manipulations to have economically significant effects (Blavatskyy, 2011).

In terms of sociology, framing and loss aversion can be used to enhance cooperation in social dilemmas. First illustrated by Fleishman (1988), van Assen (1998) shows that leaders can frame a negotiation as a situation in which hindrance leads to a loss, while collective cooperation leads to a lesser loss or a bigger gain. As people evaluate outcomes in terms of losses and gains, comparing it to a previously set subjective reference point, avoiding losses, the conditions of cooperation are enhanced.

2.3.2 Prospect theory and field experiments

Because of “difficulties associated with executing a clean empirical test” of loss aversion in the field, discoveries in behavioral economics and framing manipulations have primarily been studied in controlled laboratory settings (Hossain and List, 2012, p. 2151). Recently, studies have tried to find evidence of loss aversion in in the field.

Dimmock and Kouwenberg (2010) study whether loss aversion affects a household’s portfolio choice. Household survey data was used to examine the relationship between a loss-aversion coefficient identifying the level of loss aversion and household participation in equity markets, allocations to equity, and allocations between mutual funds and individual stocks. Results show that households scoring high on loss aversion are less likely to participate in equity markets and direct stockholding (Dimmock and Kouwenberg, 2010). In an

alternative context, Fehr and Goette (2007) argue the relationship between employee work effort and employee wage depends on the level of loss aversion. Results from a randomized field experiment involving employees of two Swiss messenger services indeed showed that loss averse individuals were more likely to put less effort into their work when their wages were increased (Fehr and Goette, 2007).

Finding support for the existence of loss aversion in the field is one thing, using prospect theory to manipulate behavior is another. Hossain and List (2012) study the effect of framing manipulations, i.e.

conditional incentives framed as both losses and gains, on productivity of individuals and teams. They found that total team productivity of a Chinese high-tech manufacturing facility was enhanced by 1%, purely due to the framing manipulations. Moreover, this effect was stable over the entire duration of the experiment (Hossain and List, 2012).

Dholakia and Simonson argue that loss aversion is more conspicuous when “comparisons are explicit rather than implicit” (2005). A field experiment involving real online auctions exploited this effect by

encouraging bidders to make explicit comparisons between the presented product and a list of all upcoming products at the auction, thereby significantly influencing bidding behavior (Dholakia and Simonson, 2005).

Bidding behavior tended to be more cautious, represented by fewer bids and bidding outbursts, increased sniping, lower bids and bids on multiple items, and later bids (Dholakia and Simonson, 2005).

Not all field experiments concerning loss aversion and framing manipulations yield consistent results. In contrast to earlier laboratory experiments, Levitt, List, Neckermann and Sadoff (2012) did not find an increase in student effort when rewards were framed as losses. Although framing manipulations did not seem to apply to students, it did to teachers. Fryer, Levitt, List and Sadoff (2012) show that by paying teachers their expected bonus in advance and taking the money back if students did not improve sufficiently, student performance was significantly improved. Not finding support for alternative explanations, paying teachers their expected bonus in advance triggers loss aversion: “giving back the money is more painful for the teachers than falling short of a bonus in the traditional ex post payment model” (Fryer et al., 2012). As a consequence, teachers will put more effort into teaching, to minimize the risk of having to pay the bonus back.

Ho, Lim, and Camerer (2006) provide insights into the solutions and opportunities behavioral economics provide for an applied discipline such as marketing. In terms of prospect theory, framing and loss aversion, companies might consider deliberately communicating in terms of losses instead of gains. In other words, companies could focus on time and value lost if consumers do not use a product or service. When operationalized in a waiting line environment, loss aversion could stimulate the choice between various waiting lines by communicating time savings as a loss instead of a gain. However, the granularity of meaning, the relation of semantic versus conceptual structure, and the status of reference points (Mayer, 1992) need to be carefully examined when conducting an experiment examining the differences between time savings framed as a loss versus communicating time savings a gain.

2.4 Problem definition

2.4.1 Problem statement

After reviewing the existing literature on waiting time psychology, SST and loss aversion, the following research question is formulated: ‘To what extent can framing of saved time encourage the use of SST?

2.4.2 Delimitations of the study

Not all SST is the same. Attitudes, satisfaction and choice between SST and personal service mainly depend on the characteristics of a specific self-service desk or kiosk (Dabholkar, 1996). This research does not examine the

effects of framing of saved time between various types of SST. The research is limited to one type of SST: Self Service Check In (SSCI) at Amsterdam Schiphol Airport (AAS).

Behavioral, normative and control beliefs affect the choice between SST and personal service. Although, theory of planned behavior is used as a theoretical foundation to conceptualize the motives of

consumers to use or not use SSCI, the aim of this research is not to provide additional insight into the conceptual framework of SST satisfaction, experience, evaluation and re-use intentions. Therefore, although intention to use SST in the future mainly depends on customer experience and satisfaction with the SST in the past (Chen et al., 2009), this study will not measure customer satisfaction after using SST. It is merely interested in the percentage of consumers using the SST compared to personal service.

3. THEORETICAL FRAMEWORK

3.1 Conceptual framework

The literature review has led to the development of a conceptual framework, displayed in figure 3. Communicating time savings affect the percentage of consumers using SST, and is moderated by framing the amount of time savings as a gain or a loss.

Figure 3 – Conceptual framework

3.2 Hypotheses

Communicating time savings is hypothesized to influence the percentage of consumers using SST. First, waiting is one of the most frustrating, boring and annoying experiences there is. While waiting, people become impatient, bored, anxious, and even angry (Schwartz, 2011). Nobody likes to wait, especially when reminded of it. This is clearly illustrated by the common practice of consumers to search and join the shortest waiting line or, after processing available information, join the line with the shortest waiting time (Whitt, 1986). Therefore, it is likely that consumers will prefer to use SST once they are informed that it would save them time. The higher the communicated time savings, the higher the percentage of consumers using SST should be.

Second, uncertain waits are longer than certain waits (Maister, 1985). Consumers will therefore prefer and choose the line that provides most certainty, especially when time-pressure is high. By stating that SST is at least X minutes faster than a staffed service desk, SST has become the more certain choice in terms of waiting time. Therefore, when providing information regarding SST time savings, consumers are more eager to use SST. The higher the communicated time savings, the higher the certainty, the more eager to use SST.

Last, communicating time savings for using SST influences the static attitude towards SST. Although SST ease of use, and fun is not affected by communicating time savings, it does increases the perceived usefulness of the SST. If a SST is perceived as more useful, it is more likely for consumers to use the respective SST (Wang et al., 2012). Moreover, by communicating time savings, the risk of using SST is reduced. Even

Time savings

Framing

desk. Based on that information, consumers feel more in control, increasing the use of SST. The higher the communicated time savings, the higher the perceived usefulness and control, the lower the perceived risk, and the higher the percentage of consumers using SST should be. Overall, the following is hypothesized.

H1: The greater the communicated time savings, the greater the percentage of consumers using SST.

Previous studies have shown that, when directly compared or weighted against each other, losses loom larger than gains (Kahneman and Tversky, 1984). This asymmetry between the power of positive and negative expectations has an evolutionary history. Organisms that treat threats as more urgent have a better chance to survive and reproduce (Bruer, 2012). A study from Hjorth and Fosgerau (2011) show that this effect can be extended to loss aversion of an individual in the time dimension. Their study showed that an individual valued time losses as 3.7 times higher than equally-sized time savings. This degree of loss aversion is even higher than with respect to costs, similar to Horowitz and McConnell (2002). Loss aversion can be operationalized by framing time savings as a gain or a loss. Therefore, the following is hypothesized.

H2: Framing time savings as a loss results in a higher percentage of consumers using SST than framing time savings as a gain.

Based on the previous two main effects, the following moderation is hypothesized.

H3: The positive relationship between the communicated time savings and percentage of consumers using SST is moderated by message framing, so that this relationship is stronger for communicating time savings as a loss.

Qualitative interviews and quantitative vignette studies were conducted to validate the stimuli and treatments of an experimental vignette study testing the hypotheses above. As pointed out by many scholars (e.g. Falk and Heckman, 2009), the use of multiple methods helps to better understand the mechanisms observed, increasing external and internal validity.

4. METHOD

Structured interviews were conducted to develop experimental stimuli that communicate small, medium and large time savings. Moreover, an experimental vignette study was conducted to test whether framing of time savings indeed triggers loss aversion. These pre-tests are the basis for the experimental vignette study design and treatments. All studies refer to the use of Self Service Check In (SSCI). SSCI allows departing passengers at airports to check in without the help of an airline agent, and is a representative type of SST (Castro et al., 2010). Most international hub airports have introduced SSCI in addition to their regular check-in and transfer desks.

4.1 Pre-test time savings

Although precise processing times differ depending on airport size and complexity, passengers can wait up to 30 minutes in line for checking in with their flight (Skidmore, 2007). Contrary, it generally takes 8 minutes to use SSCI (Statistic Brain, 2013). Therefore, objective time savings could easily be calculated. However, consumers evaluate waiting time subjectively (Hornik, 1984). To avoid skepticism of first-time users, the amount of time consumers think they could save when using SSCI is communicated. In order to identify these subjective small, medium, and large time savings, structured interviews were conducted.

4.1.1 Participants

Data for validating the experiment time saving treatments was collected from May 1st 2015 till May 8th 2015 by means of structured interviews with students, colleagues and acquaintances from the personal network of the Master student. A purposive heterogeneous sample of SSCI users and non-users was selected.

4.1.2 Procedure

The telephone administered structured interviews were scheduled to last 10 minutes. The limited length of the interviews, even as the needlessness to identify non-verbal communication, justifies the used interview method (Van Waveren, 2004). Each respondent was undisturbed by people and/or the environment (Evers, 2007). Beforehand, interviewees were thanked for their participation, introduced to the research, and reminded of the non-existence of right and wrong answers. The interview script (Appendix I) covered questions concerning demographics, flying behavior, last trip, check-in waiting times, and SSCI use. Interviews were not recorded. Responses were treated confidentially and anonymously, and are intended for research purposes only.

4.1.3 Results

Table 1 presents the means, standard deviations, and zero-order correlations among the measured variables. Twenty interviews were conducted. All interviews lasted 5 to 10 minutes (M = 8 minutes). The sample consisted of 7 males and 13 females. Respondent’s age varied between 18 and 60 years (M = 25.65, SD = 13.70). Two respondents were married, one was divorced, while 17 were unmarried. All respondents were aware that SSCI is available at Amsterdam Airport Schiphol; 30 percent of the respondents even tried SSCI. On average,

respondents flew 3.2 times a year. The average perceived waiting time for checking-in at Amsterdam Airport Schiphol was 20.4 minutes. Besides, 10.1 minutes was classified as a small waiting time, while 30.5 minutes was classified as a large waiting time for checking in. Similarly, on average, respondents thought they could at least save 5.3 minutes, with a maximum of 13.1 minutes by using SSCI.

Variable Mean S.D. 1 2 3 4 5 6 7 8 9

Descriptives

1 Gender (0 = male, 1 = female) 0.65 0.49 X

2 Age 25.65 13.70 -.48* X

3 Marital status (0 = unmarried, 1 = married) 0.20 0.52 -.33 .94** X 4 Flight frequency (p. year) 3.20 1.51 -.26 .06 -.05 X 5 Avg. check-in waiting time (min.) 20.38 7.96 .04 .04 .01 -.11 X 6 Small check-in waiting time (min.) 10.13 3.49 .26 -.15 -.09 .02 .84** X 7 Large check-in waiting time (min.) 30.50 9.30 .16 -.04 -.02 -.12 .80** .73** X 8 Min. SSCI time savings (min.) 5.30 2.45 .27 -.38 -.25 -.19 .13 .23 -.10 X 9 Max. SSCI time savings (min.) 13.13 6.53 .15 -.41 -.27 -.19 -.14 -.02 -.29 .68** X * p < .05, ** p < .01

N = 20

Table 1 – Means, standard deviations, and correlations pre-test time savings

The increase of available information led to more sophisticated, educated and skeptical consumers (Labrecque, vor dem Esche, Mathwick Novak and Hofacker, 2013). The experimental stimuli for large time savings must therefore be high enough to be considered as large, but small enough to be considered credible for all consumers. Therefore, large time savings are represented by 10 minutes. Second, as respondents indicate they think they could at least save 5.3 minutes when using SSCI, communicating time savings below that value would still be considered low. Therefore, small time savings are represented by 2 minutes. Last, medium time savings are represented by 5 minutes.

4.2 Pre-test framing

An experimental vignette study was conducted to test whether the framing of time savings indeed triggers loss aversion, making consumers prefer SSCI over a staffed check-in desk.

4.2.1 Participants

Data for validating the experiment framing treatments was collected from March 6th 2015 till March 18th 2015 by means of an online survey, build with Qualtrics. The survey was distributed among the personal and social media network of the Master student, whereby potential respondents were asked to participate in a short two minute questionnaire, without mentioning the goal of the experiment.

4.2.2 Design

Table 2 presents the pre-test 2x1 between subjects design of the experiment. The independent variable was framing, while the dependent variable was check-in procedure choice, either SSCI or staffed check-in desk.

Choice check-in procedure

SSCI or staffed check-in desk

Framing Gain (Save time) G

Loss (Don’t lose time) L Table 2 – Factorial design pre-test framing (2x1)

4.2.3 Treatments

On average, 18% of worldwide departing passengers use SSCI, while 59% of departing passengers use a staffed check-in desk (Statistic Brain, 2013). The remaining passengers check in online. As consumers can wait up to 30 minutes in line for checking in with their flight (Skidmore, 2007), ten minute time savings are assumed to be enough to overcome preferences for staffed check-in desks for some consumers, but not for all.

Keysar, Hayakawa and An (2012) found that using a foreign language reduces decision-making biases. Consequently, effects of loss aversion could be less dramatic when articulated in a foreign language compared to the respondents’ native language. Therefore, saved time was framed in Dutch and English: “Bespaar tijd! Save Time! At least 10 minutes faster with Self Service Check-In” and “Verlies geen tijd! Don’t lose time! At least 10 minutes faster with Self Service Check-In”.

and positive in nature, showing smiling people, enjoying the check-in procedure. It is therefore unlikely that a difference in direct affect transfer explains a difference in check-in procedure choice. To conclude, randomizing respondents between the two treatments, a difference in check-in procedure choice is highly likely to be entirely caused by the framing of saved time.

4.2.4 Procedure

After being shown an introduction thanking the respondent for their participation, unveiling the research topic, and ensuring the anonymity and confidentiality of the responses, respondents were asked to fill in demographic information (Appendix II). Demographics included gender, age, highest degree or level of education completed, flight frequency per year, and flight purpose. To avoid bias and learning effects, each respondent was randomly assigned to one of the two treatments presented in figure 4 and 5. Respondents were asked to imagine they are flying to a destination abroad. They are in the departure terminal of the airport, ready to check in. They are faced with the situation as presented in figure 4 or 5.

Figure 4 – Pre-test framing treatment 1: Save time

Figure 5 – Pre-test framing treatment 2: Don’t lose time

After examining either figure 4 or 5, respondents were asked what check-in procedure they would use for checking in with their flight: Self Service Check-In Kiosk or Staffed Check-In Desk. Afterwards, respondents were thanked for their participation.

4.2.5 Data

It took respondents two minutes to fill in the survey. A total of 66 questionnaires were started and 63 were completed, resulting in a completion rate of 95%. Three respondents did not fill in their preference for SSCI or staffed check-in desk after completing the demographic survey questions. Although reducing statistical power, listwise deletion was used to delete these three responses (Enders, 2010). Two respondents filled in the survey twice. The second administration of the two individuals was excluded from analysis. Furthermore, two

respondents did not fill in their age. Respondents are equally likely to report age, and age is equally likely to be reported than other variables. The data is therefore missing at random (Howell, 2007). Even though variability is reduced and covariance and correlations are weakened, missing age values were replaced with the sample mean (Enders, 2010), resulting in a dataset of 61 responses.

4.2.6 Results

Table 3 presents the means, standard deviations, and zero-order correlations among the measured variables. The sample consisted of 40 males and 21 females. The sample as a whole was relatively young, ranging from 14 to 50 years old (M=24.17, SD=5.98). Education levels were all, although not equally, represented, as most participants completed their Bachelor’s Degree (39%). Nearly half (49%) of the sample flies one to two times a year, primarily because of leisure purposes (75%). Furthermore, similar to Elliot & Hall (2005), men and women were equally likely to use SSCI, χ2(1, N = 61) = .01, p = .91.

Variable Mean S.D. 1 2 3 4 5 6 7 Descriptives 1 Gender 1.34 0.48 X 2 Age 24.17 5.98 .09 X 3 Education 4.98 1.83 -.11 .28* X 4 Flight frequency 2.62 0.82 -.09 -.03 .14 X 5 Flight purpose 2.21 0.71 -.12 -.12 -.02 -.20 X 6 Check-in procedure 1.30 0.46 -.02 .16 -.01 -.01 -.20 X 7 Framing 1.54 0.50 .04 -.11 -.17 -.02 .23* -.27* X * p < .05 N = 61

Table 3 – Means, standard deviations, and correlations pre-test framing

Figure 6 presents the treatments and associated check-in choice of the respondents. Sixteen out of 28 (55%) of the respondents indicate they would use SSCI when confronted with 10 minute time savings framed as

test is used to validate the results, showing a significant difference between the choice for check-in procedure when framing the amount of saved time differently (p = .05, Fisher’s exact test). To conclude, the results of the pre-test show that the designed framing of time savings when using SSCI indeed triggers loss aversion, providing sufficient support to use this framing in a follow-up experiment.

Figure 6 – Results pre-test framing

4.3 Experimental vignette study

To minimize the possibility of contamination because of a decrease in control (Davis University of California, 2006), an experimental online vignette study was conducted, built with Qualtrics, acquiring a proportional distribution between realism, difficulty and level of control.

4.3.1 Participants

The experiment sample consisted of 328 Aviation Bachelor students from the Amsterdam University of Applied Sciences attending lectures and seminars from Monday 11th of May 2015 till Thursday 21st of May 2015, during which students were asked to participate in a short five minute questionnaire, without mentioning the goal of the experiment.

4.3.2 Design

Table 4 presents the factorial 2x3 between subjects design of the experiment. The independent variables were framing and communicated time savings. SSCI preference was the dependent variable. In line with El Haji

Save time Don't lose time

Self Service Check-In Kiosk 16 27

Staffed Check-In Desk 12 6

0 5 10 15 20 25 30 C o u n t Framing

level is determined by conditions beyond the control of the experimentalist (Saunders, Lewis and Thornhill, 2012). The experiment is therefore qualified as a quasi-experiment.

Communicated time savings

Low (2 min) Medium (5 min) High (10 min)

Framing Gain (Save time) GL GM GH

Loss (Don’t lose time) LL LM LH

Table 4 – Factorial design experimental vignette study (2x3)

4.3.3 Stimuli and control variables

The independent variable time savings was manipulated through communicating small, medium and large time savings. Corresponding to the successful time savings pretest (paragraph 4.1), small time savings were represented by 2 minutes, medium time savings were represented by 5 minutes, while large time savings were represented by 10 minutes (Appendix III).

The independent variable framing was manipulated through communicating time savings as a gain or a loss. Corresponding to the successful framing pre-test (paragraph 4.2), time savings were framed as follows: “Bespaar tijd! Save time! At least … minutes faster with Self Service Check In” and “Verlies geen tijd! Don’t lose time! At least … minutes faster with Self Service Check In (Appendix III).

The dependent variable SSCI usage was measured using the percentage of respondents within a treatment preferring SSCI over a staffed check-in desk. This dichotomous measure (i.e. any individual respondent either chooses to use SSCI or a staffed check-in desk) was chosen because it matches a real life situation. Typically, consumers either use SSCI or a staffed check-in desk at an airport.

As time pressure has no effect on risk attitudes for gains, but increases risk aversion for losses (Kocher, Pahlke, and Trautmann, 2013) consumers in the ‘loss’ treatments are more likely to prefer SSCI over a staffed check-in desk in this study. Therefore, the control variable perceived time pressure was included. The control variable was measured using 5 items on a 5 point Likert scale ranging from never to always, adopted from Putrevu and Ratchford (1998). For the purpose of this study, the items were slightly modified to include the phrase “check-in for my flight” or “checking-in for my flight”.

As consumers with more SST experience are more eager to use SST whenever the previous encounter was classified as satisfactory (Chen et al. 2009), the control variable SST experience was included. The control variable was measured using 2 items asking whether the respondent ever used SSCI before, and whether that

Communicating larger amounts of time savings is likely to influence the perceived usefulness of SST (Wang et al., 2012), while the perceived reliability, perceived ease of use, and perceived fun of SST ought to remain constant across all treatments and respondents. Therefore, the control variables perceived usefulness, perceived ease of use, perceived reliability and perceived fun were included. The control variables were measured using a total of 10 items on a 5 point Likert scale ranging from strongly disagree to strongly agree, adopted from Elliot et al. (2012). For the purpose of this study, the items were slightly modified to include the term “Self Service Check In”.

4.3.4 Procedure

Subjects must not be aware of other treatments, nor enter other treatments (El Haji, 2015). Therefore, students were asked to fill in the questionnaire silently and individually before, after or during one of their Aviation Engineering, Aviation Operations, or Aviation Logistics lectures or seminars. Questions concerning the questionnaire were answered one-on-one after students raised their hand. Furthermore, students were asked not to share and discuss their responses with others students until the 22nd of May 2015. By checking for duplicate IP addresses afterwards, subjects who filled in the questionnaire multiple times were removed from the dataset; only the first response was included in the analysis.

With regard to the questionnaire, respondents were shown an introduction thanking the respondent for their participation, unveiling the research topic, and ensuring the anonymity and confidentiality of the responses (Appendix IV). Afterwards, each respondent was automatically randomly assigned to one of the six treatments. As SSCI choice is affected by companion influence (Wang et al., 2012), respondents were asked to imagine they are flying to a destination abroad, not accompanied by friends, relatives or acquaintances. By mentioning the respondent is travelling individually, the variable companion influence is kept constant. Next, respondents were asked to imagine they are in the departure terminal of the airport, ready to check in. Respondents were being notified that the average waiting time at a staffed check-in desk is 30 minutes. Consumers expecting a large waiting time are more eager to use SSCI (Wang et al., 2012). By mentioning an average waiting time of 30 minutes, the variable perceived waiting time is kept constant.

Next, respondents were shown an image of SSCI, a sign communicating time savings framed as a gain or a loss, and an image of a staffed check-in desk. Similar to the framing pre-test, SSCI and the staffed check-in desk were visually represented by the same set of airlines: Air France, KLM, Delta Airlines, and Skyteam partners. Airline preference could therefore not explain a difference in check-in procedure choice. Both check-in

visualizations were the same size and positive in nature, showing smiling people, enjoying the check-in procedure. It is therefore unlikely that a difference in direct affect transfer explains a difference in check-in procedure choice. Finally, respondents were asked to indicate their check-in procedure preference.

Thereafter, respondents were asked to fill in whether they agree with statements regarding SSCI characteristics on a five point Likert scale, ranging from strongly disagree to strongly agree. Furthermore, respondents were asked whether they ever used SSCI before, and if so, if that experience was satisfying. Next, respondents’ perceived time pressure when checking in for their flight was measured using 5 items on a 5 point Likert scale, ranging from never to always. Finally, the respondents were asked to fill in demographics, including gender, age, education, marital status, household income, employment status, flight frequency and flight purpose, after which the respondents were thanked for their participation.

5. RESULTS

5.1 Data

The sample consisted of 328 Aviation Bachelor students from the Amsterdam University of Applied Sciences. 297 questionnaires were completed, resulting in a completion rate of 91%. Students that failed to complete the questionnaire were excluded from the dataset. The data was checked for duplicate IP addresses; only the first response from a particular IP address was considered valid. Although reducing statistical power, listwise deletion was used to delete three responses (Enders, 2010). Frequencies were checked to search for errors in the data. Two people failed to enter their age. Even though variability is reduced and covariance and correlations are weakened, missing age values were replaced with the sample mean (Enders, 2010). Two students filled in unrealistically high values of flight frequency per year (i.e. 103 and 192), and were excluded from the dataset, resulting in a sample size of 292 responses.

The skewness of all variable items varied between -1.3 and 0.1 indicating a moderately negative distribution. Considering the large sample size, skewness will not make a substantive difference in the analysis (Tabachnik and Fidell, 2001). The kurtosis of all variable items varied between -0.5 and 2.3, indicating a variance in flatness and pointiness of the item distributions. Kurtosis can result in an underestimate of the variance. However, this risk is reduced because of the large sample size (Tabachnick and Fidell, 2001).

The perceived time pressure item rPTP5 was recoded into a different variable called PTP5 to assure that no counter indicative items were included in the analysis. Following, a reliability check was run. Reliability examines the consistency of the measurement (Saunders, Lewis and Thornhill, 2012). The Cronbach’s alpha of perceived usefulness, perceived fun, perceived reliability, and perceived time pressure, has a value above .70, which is considered good. The Cronbach’s alpha values do not increase significantly when an item is deleted. Cronbach’s alpha of perceived ease of use is just below 0.7. However, the smaller number of items, the greater the likelihood of the reliability analysis to be inaccurate (de Vaus, 2002). Therefore, as the perceived ease of use scale consists of only two items, the reliability score is considered sufficient for further analysis. Afterwards, scale means have been computed for all variables.

Collinearity between variables tends to “inflate the standard errors of their regression coefficients, making it more difficult to obtain significant values” (Mehra, Kilduff and Brass, 2011, p135). To check for multicollinearity, the tolerance and VIF collinearity statistics associated with each variable. Multicollinearity poses a problem when tolerance is lower than 0.20, and VIF is higher than 5. Fortunately, none of the independent variables violated these criteria; multicollinearity does not limit the validity of the analyses.

5.2 Analyses

Table 5 presents the means, standard deviations and zero-order correlations among the measured variables. The sample consisted of 238 males and 54 females (age M=20.26, SD= .39). 190 respondents have used SST before.

Variable M S.D. 1 2 3 4 5 6 7 8 9 10 11 12

Descriptives

1 Gender (0 = male, 1 = female) 0.18 0.39 X

2 Age 20.26 1.80 .05 X

3 Flight frequency 3.08 3.75 .02 .04 X 4 Flight purpose (0 = business,

1 = leisure, 2 = VFR, 3 = other) 1.18 0.47 .16** .12 .22** X 5 Time savings 5.61 3.30 -.04 .01 -.08 .05 X 6 Framing (0 = gain, 1 = loss) 0.50 0.50 .10 -.09 .06 .01 .00 X 7 Perceived usefulness 3.88 0.74 -.04 -.09 .01 .08 .11 .06 (.82) 8 Perceived ease of use 3.27 0.79 -.04 -.07 .12* .09 .09 .04 .53** (.66) 9 Perceived reliability 3.62 0.68 -.07 -.09 .07 .04 .03 .02 .54** .56** (.70) 10 Perceived fun 2.89 0.88 .08 -.03 .02 .14* .09 .10 .32** .37** .30** (.85) 11 SSCI experience (0 = no, 1 = yes) 0.65 0.48 -.06 .14* .17** .03 .01 -.10 .20** .30** .27** .12* X 12 SSCI choice (0 = SSD, 1 = SSCI) 0.80 0.40 -.03 -.01 -.02 -.02 .14* .03 .45** .32** .34** .24** .26** X * p < .05 (2-tailed), ** p < .01 (2-tailed)

N = 292

Cronbach’s alpha of computed scale means are displayed between brackets Table 5 – Means, standard deviations, and correlations experimental vignette study

Table 6 and figure 7 show that 113 out of 145 (78%) of the respondents indicate they would use SSCI when confronted with time savings framed as a gain. 122 out of 147 (83%) of the respondents indicate they would use SSCI when confronted with time savings framed as a loss. Thus, framing time savings as a loss suggest to yield higher percentages of SSCI usage. However, a two-tailed Chi-square analysis without Yates correction showed that the difference in percentage of respondents choosing SSCI was statistically insignificant, χ2 (1, N=292) = 1.19, p = .28. 78% 83% 22% 17% 0% 20% 40% 60% 80% 100%

Save time Don't lose time

P

ercen

ta

g

e

Staffed Check-In Desk Self Service Check-In Kiosk

Chosen check-in procedure

Self Service Check-In Kiosk Staffed Check-In Desk

Framing Gain (Save time) 78% 22%

Loss (Don’t lose time) 83% 17%

Table 6 – Results experimental vignette study I

5.2.1 Testing the hypotheses

In total, 80% of all 292 responses preferred SSCI over a staffed service desk. However, as presented in figure 8, the percentage of SSCI choice differed for communicating 2, 5 and 10 minutes of time savings framed either as a gain or a loss.

Figure 8 – Results experimental vignette study II

To test hypothesis 1 and 2, a logistic model was fitted to the data (N = 292). In table 7, the parameters of the model are presented. A logistic regression analysis was chosen because of the dichotomous nature of the dependent variable and the multiple independent (but possibly interdependent) variables. The Omnibus test of model coefficients indicated that the overall model fit the data, χ2 (2) = 6.802, p < .05. This is confirmed by the non-significant value of the Hosmer and Lemeshow goodness-of-fit test, χ2 (4) = 3.064, p = .547. The logistic regression analysis showed that time savings had a significant influence on SSCI usage. With every minute of increase in communicated time savings, the odds of a respondent choosing SSCI increased with 11.7%, Wald’s χ2

(1) = 5.285, p < .05. Hypothesis 1 is thereby supported.

Planned contrasts revealed that communicating medium (i.e. 5 minute) time savings significantly increased the percentage of SSCI use compared to communicating small (i.e. 2 minute) time savings, Wald’s χ2 (1) = 8.845, p < .05. However, communicating large (i.e. 10 minute) time savings did not significantly increase the percentage of SSCI use compared to communicating medium (i.e. 5 minute) time savings, Wald’s χ2 (1) =

60% 65% 70% 75% 80% 85% 90%

2 minutes 5 minutes 10 minutes

%

SSC

I

u

se

Communicated time savings

Gain (Save time) Loss (Don't lose time)

.009, p > .05. This latter effect is likely to reflect an equilibrium concerning the amount of people willing to use SSCI. Communicating even larger time savings when using SSCI would not result in higher percentages of SSCI use, as it would not convince people of the temporal, psychological, or economic benefits of SSCI.

Although framing time savings as a loss consistently yielded a higher SSCI usage, no statistically significant effect was found, suggesting that the probability of a respondent choosing SSCI was the same for framing the time savings as a gain and a loss, Wald’s χ2(1) = 1.205, p > .05. Hypothesis 2 is thereby not supported. Step two of the logistic regression analysis showed that framing does not moderate the relationship between time savings and SSCI usage, Wald’s χ2(1) = .377, p > .05. The positive relationship between the communicated time savings and percentage of consumers using SST was not found to be moderated by message framing, so that this relationship is stronger for communicating time savings as a loss. Hypothesis 3 is thereby not supported.

Step 1 B SE β Wald’s χ2 df p eβ (odds ratio)

Constant .676 .313 4.664 1 .031 1.966

Framing (0 = gain, 1 = loss) .329 .300 1.205 1 .272 1.390

Time savings .111 .048 5.285 1 .022 1.117

Test χ2 df p

Omnibus overall model evaluation 6.802 2 .033

Hosmer and Lemeshow goodness-of-fit test 3.064 4 .547

Step 2 B SE β Wald’s χ2

df p eβ (odds ratio)

Constant .813 .353 5.300 1 .021 2.255

Framing (0 = gain, 1 = loss) -.405 .877 .213 1 .644 1.087

Time savings .084 .057 2.170 1 .141 .667

Framing by time savings .094 .106 .780 1 .377 1.099

Test χ2 df p

Omnibus overall model evaluation .783 1 .376

Hosmer and Lemeshow goodness-of-fit test 2.361 4 .670

Table 7 – Logistic regression results of framing and time savings, N = 292

5.2.2 The mediating role of perceived usefulness

A model was fitted to the data to examine the mediating role of perceived usefulness. In case of a mediating relationship, large communicated time savings are causing an increase in SSCI choice due to an increase in the perceived usefulness of SSCI. Testing the mediating relationship followed standard statistical procedures (Baron and Kenny, 1986). To test for mediation, four statistical tests were conducted to check whether a significant