Assessing the impact of categorization and categorical
ambiguity on consumers’ willingness-to-pay and the
moderator effect of ambiguity tolerance in a real-life
experiment
Bachelor Thesis
Maarten van den Berg
Amsterdam, July 17, 2014
Student number: 10264426 Thesis Seminar Business studies Supervisor: Bram Kuijken Semester 2, block 3
Abstract
This study contributes to the literature in the field of consumers’ behavior while these consumers are being exposed to category labels given by organizations. Using a real-life experiment, we aimed to find out whether or not Categorization and Categorical Ambiguity affect consumers’ Willingness-To-Pay (WTP) and how Ambiguity Tolerance (AT) moderates this relationship. A total of 518 consumers participated in an online second-price sealed bid auction and were randomly assigned to different treatments. Results of the study indicate that Categorization and Categorical ambiguity have a positive relationship with consumers’ WTP compared to the circumstances in which only general features are mentioned. However, no differences has been found between the consumers’ WTP under circumstances of
Categorization and Categorical Ambiguity. Unexpectedly, measurements of AT turned out to be internally inconsistent. Therefore, no new insights have been found on the moderating effect of AT.
Table of Content
Foreword ………..5
1. Introduction ……….6
2. Literature review………..8
2.1 Consumers’ Willingness-to-pay………8
2.2 Consumers’ Uncertainty in the adoption process……….8
2.3 Product categorization………...10
2.4 Categorical ambiguity………...11
2.5Ambiguity Tolerance………...11
3. Theoretical Framework………..14
3.1 Influences of categorization on WTP and categorical ambiguity………..14
3.2 The moderating effect of ambiguity tolerance on consumers’ willingness-to-pay….15 3.3 The conceptual framework………...16
4. Methodology……….17
4.1 Research method ………17
4.2 Data collection ………17
4.3 Experiment design………18
4.3.1 Vickrey´s auction model………..18
4.3.2 Auctioned product………... 19
4.3.3 Treatments ………...20
4.4 Variables and Measurements ...20
4.4.1 Dependent variable Willingness-to-pay ………..20
4.4.2 Moderator Ambiguity Tolerance ………21
4.5 Sample characteristics ……….21
5. Results………....22
5.1 Descriptive statistics………....22
5.2 Testing the assumptions………24
5.2.1 Testing for normality ………..……….24
5.2.2 Testing for homogeneity of variance……….25
5.3 Normalizing the data ………26
5.5 Correlations……….29
5.6 Hypothesis Testing ……….29
5.6.1 Testing Hypothesis 1 and 2………29
5.6.2 Testing Hypothesis 3 and 4……….31
5.7 Additional Findings ………32
6. Discussion ………34
6.1Findings ………...34
6.2 Theoretical implications ……….35
6.3 Managerial implications ……….36
6.4 Limitations and suggestions for future research ……….37
7. Conclusion ………....39
References ………40
Appendix A: Treatment Design ……….44
Appendix B: Ambiguity Tolerance Measurements ………..45
Appendix C: Correlation Matrix ………...46
Foreword
This thesis was written as completion to the Bachelor of Economic and Business of the University of Amsterdam. I would like to thank some people for their help and support. First of all, I want to thank my supervisor Mr. B. Kuijken for his support and guidance. During the whole process, he was helpful by giving advice, feedback and was there to motivate you to remain positive when things did not turn out as expected. Secondly, I want to thank all the people who registered for Veylinx’s database. Particularly my family, friends and colleagues.
1. Introduction
Nowadays, new products and services are constantly launched by many organizations
throughout the world. Some will be incremental others will be radical. The organizations face the challenge of bringing these products to the attention of the customer and to make them understand what the product is about. Especially with innovative products, product adoption is affected by customer uncertainty about the need for or benefit of the product (Veryzer, 1998). For example with the introduction of the Apple iPad in 2010, it was the question what the product really was about . The Apple iPad is neither a smartphone, a netbook nor a tablet PC, but includes some elements of all of them (Henderson & Yeow, 2012). But how do you position a novel product in such a way that people understand what the benefits of it are? Especially for managers it is from importance to understand when and why consumers see their new products as beneficial to them. While formulating and implementing marketing strategies, this knowledge can be helpful to achieve the goals set by the organization in the best and most effective way.
In the literature, much research has been done on how consumers try to understand these novel product and how they respond to certain variables. More specifically, many studies focused on the concept of categorization. The categorization theory says that
consumers try to understand a product by placing it within an existing category (Bloch, 1995). The concept of categorization is useful to organization while positioning their products
(Viswanathan and Childers, 1999) . As much as it can be useful to organizations, it also has its dangers. A phenomenon that has been studied extensively in literature is about
categorization ambiguity. Categorization ambiguity occurs when information of new products makes it difficult or impossible to place the novel offering in a single, existing category (Gregan-Paxton et al.,2005). This problem is getting more and more relevant nowadays. Currently, many products that are launched are ambiguous with regard to the product categories, which means that the features of the products belong to several categories (Moreau et all.,2001).
Current studies already showed what effect categorization (Rajagopal & Burnkrant, 2004) and categorical ambiguity (Zukerman,1991) can have on what consumers are willing to pay for a product. Furthermore, consumers can differ in the way they react to these ambiguous stimuli because they differ in the level of ambiguous tolerance (Furnham and
7 Ribchester, 1995; Blake, Zenhausen, Perloff and Heslin, 1973; Budner, 1962).However, little attention has been given to how this affects the relationships described above.
Therefore, this study aims to give more insight into the effects of categorization and categorical ambiguity on consumers’ willingness-to-pay and how the level of consumers’ ambiguity tolerance influences this. It will expand the literature by taking a first step in studying the moderating effect of ambiguity tolerance and how this affects consumers’ willingness-to-pay (WTP) in a real-life experiment. The question central in this study is: To what extent do categorization and categorical ambiguity affect consumers’ willingness to pay and how does ambiguity tolerance moderate these effects?
In an attempt to answer the central research question, an second-price sealed bid auction experiment has been conducted. The design is based on the auction model of Vickrey (1961). In the second section, the relevant literature on this topic will be reviewed. In the third section, the theoretical framework will be proposed. In the fourth section, the methodology will be discussed which will be followed up by the study results in section five. In the last section, a conclusion to the central research question will be given together with the general discussion.
8
2. Literature Review
In this section, a brief overview of the existing literature concerning the defined research topic will be given. Firstly, the literature on consumer’s willing-to-pay and its relationship with uncertainty will be discussed briefly. Secondly, customer uncertainty in the adoption of innovative product will be outlined. Thirdly, the concept of categorization with regard to how consumer try to understand novel products will be discussed. Finally, an overview of the literature on the downsides of categorization is given and how the different types of
consumers respond to this. This will be done with discussing categorical ambiguity and ambiguity tolerance.
2.1 Consumers’ Willingness-to-pay
In this study, willingness-to-pay (WTP) is one of the crucial variables that is investigated. Consumers willingness-to-pay is the maximum price below or at which she will demand and buy one unit of a given product (Mankiw, 2012). This correspondents to the consumers floor reservation price, which is the maximum price at or below which a consumer will definitely buy one unit of the product (Wang et al., 2007). According to Simonson and Drolet (2004), the consumers’ willingness-to-pay price reflects both the sacrificed monetary equivalent in acquiring the product or service and the perceived value of it. Moreover, customers first try to judge what value a product is to them, before making a buying decision (Monroe, 2003). Some studies showed that uncertainty can influence the consumers’ willingness-to-pay. The study of Wang et al. (2007) showed that customers can be uncertain about their preferences and the utility of a product. Therefore, it is better to conceptualize willingness-to-pay as a range of values of price points, instead of a single point. Furthermore, it is shown that uncertainty and price range are positively related (Dost and Wilken, 2012; Wang et al., 2007). Accordingly, the findings of Wang et al. (2007) implicates that an increase in uncertainty will lead to a lower consumer’s floor reservation price, and vice versa. So uncertainty is a factor to keep in mind while studying willingness-to-pay.
2.2 Consumers’ Uncertainty in the adoption process
The concept of uncertainty is conceptualized extensively in the literature (Lipshitz & Strauss, 1997). According to Shiu et al. (2011) , consumer uncertainty is about the difference between
9 the consumer’s ideal state of information and the actual available information. Even in the adoption process, the phenomenon consumer uncertainty is relevant. While the attention on new product development and the activities related, such as marketing and research &
development is increasing, the success rate of new products shows little improvements (Wind and Mahajan, 1997). According to Urban, Weinberg and Hauser (1996) novel products lead to consumers needing to change their behavior and require learning. While projecting the
success and benefits of the new products in the development phase, you can expect that organizations did this well. However, when the adopting phase comes closer, it seems that the decision process of the customer changes due to uncertainty regarding the benefits it will gain from the new product (Castaño et al.,2008). The uncertainty perceived of new discontinuous products is greater than with incremental changes, which results in different adoption
processes (Hoeffler, 2003). While adopting new products, the new product should result in symbolic and functional value for the customer. According to Hoeffler (2003), functional value is mainly about how functional the product will be. Symbolic value is about the added value to someone’s self-esteem and status (Fournier, 1998).
Furthermore, Shiu et al. (2011) propose a multidimensional conceptualization of consumer uncertainty. This conceptualization consist out of three types of consumer uncertainties. These are knowledge, choice and evaluation uncertainty. In the literature, consumer knowledge and the consequences on the evaluation of products and the processing of information has been examined extensively (Cordell, 1997; Bettman & Park, 1980). Knowledge uncertainty can been seen as ‘the degree of confidence individuals have in their understanding of salient information, features, functionality, and utilities regarding the product under consideration’ (Shiu et al., 2011, p. 587). However, consumer knowledge can also have an effect on the purchase process of the consumer, the ultimate buying decision (Joseph & Hutchinson, 1987). When consumer try to make a buying decision, they want to make rational choices (Tversky & Kahneman, 1986) . However, when there is choice uncertainty, this is becoming more and more difficult. Choice uncertainty is about having trouble choosing the right option out of a set of options (Urbany, Dickson,& Wilkie, 1989). The last part of the multidimensional conceptualization is about evaluation uncertainty which means that consumers struggle to process and integrate information which are needed to form judgments (Shiu et al., 2011).
10
2.3 Product categorization
Uncertainty about the potential benefits of an innovative product can have negative effects on the adoption and revenues of the product (Rajagopal & Burnkrant, 2004; Zuckerman, 1991). In the literature, research has been done on how consumers try to understand products and its potential benefits. According to Mervis and Rosch (1981) people distinguish objects from the qualities, attributes and properties it has. Furthermore, humans understand objects by
assigning meaning to it and by comparing it to other objects in the class (Locke, 1690). In the literature, assigning a product to a class developed in form of the concept of categorization (Mervis and Rosch,1981). The concept of categorization says that consumers try to
understand a product by placing it within an existing category (Bloch, 1995). Consequently, this understanding takes place in the form of comparing a product to the knowledge about that certain category (Cohen and Basu, 1987). The categorization can take place at several product levels such as product type, product class and at the brand level (Sujan and Dekleva, 1987). As described above, much research has been done on how people use knowledge to classify novel product. Additionally, much research has been done on how categories can be used to make inferences about these novel products (Moreau, Markman & Lehmann, 2001). According to Waldmann, Holyoak, and Fratianne (1995), when a product is categorized, the information of that category is transferred to the novel product. This makes clear that while positioning a product in the market, categorization is an aspect that can be useful and should be taken into account (Viswanathan and Childers, 1999). Organizations that are positioning a novel product can have it placed in a certain category. But it has to take a proactive approach to make this happen (Bloch, 1995). A way of doing this for an organization is by giving a reasonable category label to the novel product, which is a good suggestion for the product’s category (Moreau, Markman & Lehmann, 2001). Consequently, this label will be a reason for consumers to make more extensive inferences from the product’s category than with the absence of such a label (Gregan-Paxton, 1999).
In the literature, several reasons are given for when and why consumers make more extensive inferences. First of all, when there is missing information about a product,
consumers are likely to base their inferences on basis of the stated category (Gelman & Markman, 1986). Also, giving a category label makes consumers focus more on the features that belong to that category and distracts them from the ones that belong to other categories (Ross & Murphy, 1996). Furthermore, a label encourages consumers to think of the product as
11 a whole. Consequently, this consumers are more likely to transfer the information from the given category to the new product, in order to maximize the perceived similarity of the novel product and the given category (Gregan-Paxton, 1999). All in all, giving a category label to the novel product is likely to lead to consumers placing the product in the given category.
2.4 Categorical ambiguity
Besides that categorization can be useful for organizations, it can also have it downsides. A phenomenon studied in the literature is about categorical ambiguity. Categorization ambiguity occurs when information of new products makes it difficult or impossible to place the novel offering in a single, existing category (Gregan-Paxton et al.,2005). Currently, many products that are launched are ambiguous with regard to the product categories, which means that the features of the products belong to several categories (Moreau et all.,2001). The literature showed in many cases, even when products where hard to place in a certain category, that consumers place their inferences on basis of single categories ((Murphy & Ross, 1994, 1999; Ross & Murphy, 1996). However, some results showed a multiple inference strategy. Murphy & Ross (1994) study showed signs of a multiple category inference stategy but extreme measurements were needed to come to this result (Gregan-Paxton et al., (2005). Another study of Ross & Murphy (1999) showed more supporting evidence of a multiple category inference strategy by consumers. It showed that that with products that can be clearly placed in more categories, consumers base their inferences in multiple categories. The study of Gregan-Paxton et al. (2005) did further research on whether consumers make product
inferences on basis of a single or multiple category strategy under the condition of categorical ambiguity. It showed that two factors, category familiarity and the nature of the category cue, are responsible for inferences made on basis of either a single or a multiple category inference strategy. A single category inference strategy is likely to occur when the perceptually cued category is more familiar than the conceptually cued category. A multiple inference strategy is more likely to occur when the perceptually cued category is less than or equal familiar (Gregan-Paxton et al., (2005), p.127).
2.5 Ambiguity Tolerance
It is shown that ambiguous situations are relevant in the context of categorization. However, it remains the question who are affected by ambiguous situations and how they respond to this. A concept relevant to this question is the concept of ambiguity tolerance, which is
12 studied extensively in the literature. This concept is studied in various branches of psychology for more than 60 years ( Frenkel-Brunswik, 1948). According to Furnham and Ribchester (1995) ambiguity tolerance is about “the way an individual perceived and processes
information about ambiguous situations or stimuli when confronted by an array of unfamiliar, complex, or incongruent clues” (p. 179), On the one hand, there are the persons with low tolerance that are likely to react prematurely, avoids ambiguous stimuli and experiences stress. On the other hand, persons with high tolerance of ambiguity see ambiguous stimuli as desirable, challenging and doesn’t denies their complexity of incongruity (Furnham and Ribchester, 1995;Blake, Zenhausen, Perloff and Heslin, 1973; Budner, 1962).
In literature, ambiguity tolerance has been seen as a personality variable ( Budner, 1962) and a property of both national cultures (Hofstede, 1984) and organizations (Furnham & Gunter, 1993). While most studies based their definition of ambiguity tolerance on the study findings of Frenkel-Brunswik’s (1949; 1951), the characteristics of ambiguity tolerance can be categorized into primary and secondary ones (Bochner, 1965). The primary ones are for example about the need for categorization, need for certainty and the early selection and maintenance of just one solution in situations that are perceived as ambiguous. The secondary characteristics are all the traits such as authoritarian, closed minded, anxious and aggressive (Furnham and Ribchester, 1995, p. 181).
The concept of ambiguity tolerance is also being seen as relevant for organizations. According to Furnham and Ribchester (1995), ambiguity tolerance have been studied in two ways with regard to organizations. First, it can been seen as a property of individuals which can influence organizations in different ways. Secondly, it can be seen as a property of the organization itself as part of the corporate culture. Most research has been focused on
ambiguity tolerance as a property of individuals (Furnham and Ribchester, 1995). An example is the study of Lysonski and Andrews (1990). The study was to look at the moderating role of ambiguity tolerance role conflict/ambiguity purpose of this and job satisfaction. It turned out that there is a negative main effect between these variables.
Whereas ambiguity tolerance is relevant and important for internal purposes, the ambiguity tolerance of organizations’ customers should also be taken into consideration. A study of Richardson et al. (1996) has been focused on the variables around private brand proneness. This study included the effect of intolerance of ambiguity on perceived value of money and reliance on extrinsic cues. It has been shown that intolerance for ambiguity is negatively related with peoples perceived value of money. Furthermore, it has also been shown that that people with great intolerance for ambiguity are more likely to rely on extrinsic
13 cues to value the quality of a product. All in all, literature showed that the concept of
ambiguity tolerance is from importance for both internal (e.g. individual and organization property) and external ( e.g. perceived value of money) purposes.
14
3. Theoretical Framework
3.1 Influences of categorization on WTP and categorical ambiguity
Currently, many new products being launched are difficult to place within an existing product category. This is a relevant problem for organizations because literature showed that
consumers try to understand and assign meaning to objects by comparing it to other objects in the product class (Locke, 1690; Bloch, 1995). Consumers being unable to place a product into an existing category can be having trouble in indentifying what the product is about. Consequently, this will lead to consumers being unable to identify the symbolic and
functional value of the product. When consumers are uncertain about what value the product is to them, this can negatively influence the adoption process of a new product. Literature showed that uncertainty about the potential benefits of an innovative product can have negative effects on the adoption of the product and the willingness-to-pay of consumers (Rajagopal & Burnkrant, 2004; Zuckerman, 1991).
To overcome this problem, organizations can take a proactive approach in having their product being places in an existing category (Bloch, 1995). A way of doing this for an
organization is by giving a reasonable category label to the novel product, which is a good suggestion for the product’s category (Moreau, Markman & Lehmann, 2001). Concluding, to avoid the negative effects on the consumers’ willingness-to-pay due to uncertainty about the potential benefits, organizations should have their product being placed within an existing category. Therefore, hypothesis one is set as the following (H1):
H1: Clearly communicating a product´s category membership has a positive effect on consumers’ willingness-to- pay for a product.
Moreau et al. (2001) states that many novel products posses features that belong to several product categories. So placing a product into a single category is becoming difficult
nowadays. When an organization fails to have their novel product being placed within a single category, this will impact the market position of the product. More specifically, Zuckerman (1991) stated that if a product tries to compete in a market, it should be seen as a relevant player in that specific product category. Otherwise, this can lead to a illegitimacy discount. Therefore, the presence of category ambiguity can lead to consumers paying less for a certain product. In more specific, this will be compared to the willingness-to-pay under conditions of categorization. Hence, hypothesis two is set as the following (H2):
15
H2: Consumers willingness-to- pay for a certain product will be lower if they receive multiple suggestions for the product’s category compared to when they are given just one suggestion.
3.2 The moderating effect of ambiguity tolerance on consumers’ willingness-to-pay
In the first two hypothesis no distinction has been made with regard to the different
consumers organizations have to deal with. One could argue that not all the consumers react in the same way on specific events and situations. With respect to the concept of
categorization and categorical ambiguity, a clear distinction can be made on how consumers react to ambiguous situations. In the literature, this concept is called ambiguity tolerance. A distinction has been made with regard to the level of tolerance a person has for ambiguous stimuli and how it reacts to this. Persons with low tolerance are likely to react prematurely, avoids ambiguous stimuli and experiences stress. On the other hand, persons with high tolerance of ambiguity see ambiguous stimuli as desirable, challenging and doesn’t denies their complexity of incongruity (Furnham and Ribchester, 1995;Blake, Zenhausen, Perloff and Heslin, 1973; Budner, 1962).
As hypothesized earlier, categorization of a product will lead to an increase of consumers’ willingness-to-pay. However, it can be argued that for persons with low ambiguity tolerance the effect will be different than for persons with high ambiguity tolerance. Persons with low ambiguity tolerance are likely to avoid ambiguous stimuli. Therefore, categorization is expected to be received as beneficial to persons with low
ambiguity tolerance because it will decrease the level of ambiguous stimuli received. On the other hand, persons with high ambiguity tolerance see ambiguous stimuli as challenging and desirable. Consequently, the categorization of a product is still expected to have a positive effect on their willingness-to-pay because the potential benefits received by buying the product is becoming more clear. However, the decrease in ambiguous stimuli will also make the product less challenging. Therefore, the hypothesized positive effect of categorization on consumers’ willingness-to-pay is expected to be stronger for consumers with low ambiguity tolerance. Hence, hypothesis three is set as the following (H3):
H3: The positive effect of categorization on consumers’ willingness-to-pay will be
stronger for consumers with low ambiguity tolerance than for consumers with high ambiguity tolerance.
16 Also with regard to the hypothesized negative effect of categorical ambiguity on consumers’ willingness-to-pay, it can be argued that persons with low ambiguity tolerance will react differently than persons with high ambiguity tolerance. When categorical ambiguity is
present, logically ambiguous stimuli will be there. Therefore, the persons with low ambiguity tolerance ,which prefer avoiding ambiguous stimuli, are expected to be affected more
negatively than persons with high ambiguity tolerance. After all, persons with high ambiguity tolerance see these ambiguous stimuli as more desirable than persons with low ambiguity tolerance. Hence, hypothesis four is set as the following (H4):
H4: The negative relationship between categorical ambiguity and consumers’ willingness-to-pay will be stronger for consumers with low
ambiguity tolerance than for consumers with high ambiguity tolerance.
17
4. Methodology
In this section the overall method of this study will be described. The general research design will be discussed and is followed up by how the used data has been collected. Furthermore, it will be explained what methodological choices have been made and what measurements have been used.
4.1 Research method
This study is focused on getting more insights into how several independent variables (categorization, categorical ambiguity and ambiguity tolerance) are affecting a consumers’ willingness-to-pay (WTP). Therefore, this is an explanatory study because it tries to establish causal relationships between variables (Saunder et al., 2013, 9. 172). An experiment will conducted because this is most suitable to study causal links between variables (Saunders, Lewis & Thornhill, 2009). Additionally, this experimental approach has often been seen as the golden standard against which the rigour of other strategies is assessed (Saunder et al., 2013, p. 174).
To conduct this research, the platform that will be used is the website of ‘Veylinx’ (www.veylinx.nl). This website is designed by two PhD candidates at the University of Amsterdam to conduct their own research and to give other students the opportunity to gather data through this platform. The website makes use of online auctions to gather its data. More specifically, the second-price sealed bid auction has been used which is based on the method of Vickrey (1961). In section 4.3 it will be explained why this specific method is used. It should be noted, that this platform’s auction are not virtually and the products will actually be sold to the winner. Besides, Veylinx does not strive to make any profit and this is stated clearly on their website. Every person can subscribe for free and participate in the auctions.
4.2 Data collection
The platform of Veylinx heavily relies on the amount of subscribers that actively participate in the auctions. For this reason, students will not receive free access to the website to run their experiments due to the limited subscribers. Especially because more than ten students were using the platform in a short period of time. To get access, an amount of 100 new subscribers had to be acquired per student. At the beginning of June 2014, the database of Veylinx consisted of more than 3200 subscribers.
18 email with an invitation to participate into an auction. Participating in an auction is voluntary at all times and subscribers can ignore an invitation for any reason. Veylinx tries to stimulate subscribers to participate in auctions by giving them a so-called ‘V-score’. This score will increase when you participate in the online auctions and decreases when you do not. With a high V-score you will have more chance of winning free prices that are given away once a month.
To acquire the compulsory 100 new subscribers, each student was free to reach that amount in its own way. In this case, this amount has been reached by the use of two strategies. First of all, close friends and family were contacted personally. This has been done with both word-of-mouth and electronic messages (SMS, e-mail and social media). However, even with the use of the internet, the 100 new subscribers has not been reached because not everyone is interested in subscribing to Veylinx. The second strategy was to get in contact with unknown people. This has been done by going into the centre of Amsterdam and ask people to
subscribe at Veylinx after carefully explaining what the platform is about. With the use of a tablet, people were given the opportunity to immediately complete their subscription.
4.3 Experiment design
4.3.1 Vickrey´s auction model
As mentioned in section 4.1, Veylinx makes use of Vickrey’s auction model. This auction model is an second-price sealed bid auction. This means that bidders cannot see what other participants are bidding for that specific product. The winner is the one that places the highest bid and he or she will pay the second-highest bid. According to Vickrey (1961),
“the optimal strategy for each bidder will be to make his bid equal to the full value of the article or contract to himself, i.e., to the highest amount he could afford to pay without incurring a net loss or to that price at which he would be on the margin of indifference as to whether he obtains the article or not” (p. 20). Furthermore, the Vickrey’s auction model is incentive-compatible. Therefore, truthful bidding dominates over- and underbidding which leads to the highest expected outcome (Riley, 1981).
Besides the advantages the Vickrey’s auction has, the literature has also shown some disadvantages. Vickrey’s auction does not reveal someone’s true willingness-to-pay from the beginning (Coursey, Hovis and Schulze, 1987) because participants first need to have some trial runs to learn what the best bidding strategy is to them (Roberson and Smith, 1982). However, the most bidders are likely to already have some experience with the bidding. Additionally, in this study participants will be randomly put into three different conditions. So
19 possible deviations, due to a lack of trial runs, are likely to affect all conditions equally. According to a study of Kagel et al. (1987), the Vickrey’s auction can still lead to overbidding because of competitive bidding. Again, the possible overbidding is expected to equally affect all the conditions.
In this experiment, three treatments will be used that will differ only on the variable that is being studied (section 4.3.3). By the use of this experimental set-up, in which there is great control over the aspects of the research process such as the context, the internal validity is ensured (Saunders et al., 2013, p. 176). In many experiments it is hard to generalize to the real world and measure actual behavior of consumers. The platform of Veylinx makes use of a actual and realistic context in which real consumer behavior is measured. Therefore, such as context will increase the external validity of the study’s results.
At final, participants were asked to fill in a small questionnaire after the auction to measure the level of ambiguity tolerance.
4.3.2 Auctioned product
The product that has been auctioned in this experiment is the ‘Sony Smartband’ (appendix A). The smartband is described by Sony as a smart accessory which is compatible with many Android smart phones. It is able to track your daily activities, can be used as a smart alarm clock, to control your music player, alarms you when you receive notifications on your
mobile phone and can fit every daily outfit. The smartband is available in many colors and the wristbands can be changed easily. The product is available since April 2014 and the advice price given by Sony was 99,95 euros. However, actual retail prices lay around 69,95 euros at the time the auction was held. The smartband is standard sold with a black wristband which is also the one used in the auction.
The Sony smartband was chosen for two reasons. Firstly, to be able to test the effect of categorization and categorical ambiguity on consumers’ willingness-pay, the product that is auctioned have to posses features of several product categories. The participant should be having trouble with indentifying in what category the product should be placed. The more different features it possesses, the more likely it will be that it is not straightforward to place it into a single category. As mentioned in the first paragraph, the smartband can been seen as a alarm clock, fashionable accessory, activity tracker and notification detector. Therefore, this product fits this requirement.
Secondly, to decrease the chance that people already have significant knowledge about the product and its features, a relatively recently launched product should be used. The
20 Sony smartband was launched in the month April this year and little advertising has been done to promote this product. Besides, the market of smart accessories is in its search for what consumers really want. Therefore, it is expected that most bidders have little knowledge about this product. Finally, the participants have only five minutes to complete the auction. Due to this limited time, there is limited time for doing research on this product and getting more knowledge about the product.
4.3.3 Treatments
A classical experiment has been conducted with this auction. Participants in the auction were randomly assigned to a treatment. These treatments differed on the cues given to which product category the Sony smartband belongs (manipulation/planned intervention). All other information and variables remained the same. In total, there were three treatments (appendix A). The first treatment, the control group, contained basic information about the product such as features, specifications and a picture. In the second treatment, the first experimental group, a single cue for a product category was given. This should help the participants to categorize the product and to better understand what the product really is about (categorization). The cue added was ‘activity tracker’. In the third treatment, the second experimental group, two additional cues were given for a product category. This treatment belongs to ‘categorical ambiguity’ because the presence of three category cues should make it more difficult for participants to place it within a single category. The cues given were ‘activity tracker, music remote, alarm clock’.
4.4 Variables and Measurements
In collecting the data, the focus in this experiment is on two variables. These are consumers´ willingness-to-pay and the level of ambiguity tolerance. Besides, participants were asked to fill in some general information about themselves (e.g. age and gender) while signing up for the platform of Veylinx.
4.4.1 Dependent variable Willingness-to-pay
The willingness-to-pay of the participants will be measured by using their auction bid value. Consequently, this correspondents to the consumers floor reservation price, which is the maximum price at or below which a consumer will definitely buy one unit of the product (Wang et al., 2007).
21 4.4.2 Moderator Ambiguity Tolerance
The moderator ambiguity tolerance was measured by four questions (Appendix B) that are subtracted from the a 50 item scale called MAT-50 developed by (Norton, 1975). An example item is: ‘ I prefer the certainty of always being in control of myself’. The scale used ranged from (1) very strongly to (5) very slightly. A high score means that the participant is tolerant for ambiguous stimuli.
4.5 Sample characteristics
In this experiment, the panel of Veylinx has been used which consist out of approximately 3200 people. From this panel, 518 participants actually placed a bid on the Sony Smartwatch. The final sample consisted of 282 male (54,5%) and a mean age of 44,87 years (SD=16,102). Of the participants, 79% were non-students and most of them had a graduate degree (29,8%) or followed a vocational education program (29,3%) degree. This relatively high presence of participant graduate could be explained by the fact that the students rely heavily on their personal network in recruiting the 100 persons. More important, Veylinx is an initiative supported by the University of Amsterdam, which could attract more people with a graduate degree because they care more about science and would like to attribute to this.
It should be noted that 66,2 % of the participants possesses an android smart phone. This is relevant because the auctioned product is only compatible with android. At final, the respondents that did not filled answers were eliminated for that specific variable but remained in the sample for the other variables.
22
5. Results
5.1 Descriptive statistics
Firstly, the descriptive statistics of this experiment will be presented. In table 1 and chart 1, a general overview of auction’s bidding amounts are given. This includes the mean, standard deviation, standard error, minimum and maximum of the bids in this auction. This will give a first impression in the possible differences among the treatments.
Table 1: descriptive statistics of different treatments
Treatment N Mean in € Std. Deviation Std. Error Minimum Maximum
1 177 9,22 15,90 1,20 0 100,00
2 162 10,37 15,89 1,25 0 80,00
3 178 10,16 17,75 1,33 0 150,00
Total 517 9,91 16,53 0,73 0 150,00
In table 1, it is shown that treatment 2, which involved the concept of categorization, has the highest mean bid. Treatment 3, which was related to categorical ambiguity, has the second highest mean bid. What can be noticed is that for all the treatments the standard deviation is high compared to the mean. This indicated that the bids are spread over a large range of
8,6 8,8 9 9,2 9,4 9,6 9,8 10 10,2 10,4 10,6
Treatment 1 Treatment 2 Treatment 3
23 values. Additionally, the difference between the minimum and maximum bids are also high. The maximum bids seems to be high compared to the mean bids. Furthermore, the minimum bids are in all the treatments equal to zero. This can be explained by the fact that participants are given the opportunity to bid zero if they are not interested in the product. These zero bids can possibly be the explanation of the relatively high standard deviation and maximum bids compared to the mean.
In table 2 and chart 2, the same descriptive statistics are shown excluding the zero bids. This has been done to check if the zero bids are the main explanation for the high standard deviation of the bids compared to the mean.
Table 2: descriptive statistics of different treatments excluding zero bids
Treatments N Mean in € Std. Deviation Std. Error Minimum Maximum
1 92 17,73 18.35 1,91 1 10000
2 102 16,47 17,34 1,72 50 8000
3 104 17,39 20,36 2,00 1 15000
Total 298 17,18, 18,69 1,08 1 15000
By excluding the zero bids, the descriptive statistics changed at some points. First of all, the mean bids of all treatments increased significantly which is not surprisingly. However, the first treatment has now the highest mean followed by the third. More importantly, the
standard deviation even increased which suggests that there are still many low values that are
15,8 16 16,2 16,4 16,6 16,8 17 17,2 17,4 17,6 17,8 18
Treatment 1 Treatment 2 Treatment 3
Chart 2: Mean auction bids for different treatments excluding zero bids
24 unequal to zero. Consequently, it seems the case that excluding the zero bids has not changed the fact that the bids are spread over a wide range of values and that the difference between the minimum and maximum bids is still high. Therefore, the zero bids will not me removed from the data.
5.2 Testing the assumptions
Before parametric tests can be used to analyze the data, some assumptions should be checked. Parametric tests are based on made assumptions..Therefore, it is important that you check the assumptions before deciding which statistical test is appropriate. Otherwise, your results are likely to be inaccurate (Field, 2009). The data should be normally distributed, variances should be homogeneous and the data has to be independent from the different participants. The data in this study is independent because participants received a personal invitation for the specific auction and were unlikely to confer with others about the treatment. Especially because the bids of others were sealed and there was a time limit to complete the auction. Therefore, someone’s bid is unlikely to depend on other bids because participant do not receive information about this. The two other assumptions will be tested in this section.
5.2.1 Testing for normality
In this section, normality of the data will be tested both visually and numerically. Through the use of histograms the normality will be tested visually. The normality will be tested
numerically by using the Kolmogorov-Smirnov test and Shapiro-Wilk test. The results can be found below.
25
Testing visually for normality with the histograms, it can be concluded that for all treatments the bids are not normally distributed and look positively skewed. This suggests that the data is more clustered around the lower end (Field, 2009). While testing numerically, these tests seem to confirm what was concluded by visually analyzing the data. Firstly, for all treatments the kurtosis is positive. This indicates a pointy and heavy-tailed distribution (Field, 2009; Levene,1960). Secondly, the kurtosis of all treatments are positive which indicates a pile up on the left of the distribution . At final, both the Kolmogorov-Smirnov and Shapiro-Wilk tests confirm that the data is not normally distributed (p < 0.05).
5.2.2 Testing homogeneity of variance
Another assumption that has to be tested is homogeneity of variance. This can been done by using Levene’s test (Field, 2009). The result is showed in table 4. Levene’s test showed that variances are not significantly different, F(2, 514) = 0,38, p>0.05. Therefore, the assumption of homogeinety of variances looks tenable.
Table 3: Tests for normality
Shapiro-Wilk Kolmogorov-Smirnov
Skewness Kurtosis Statistic P-value Statistic P-value
Treatment 1 (N=177) 2,645 8,806 .642 .000 .281 .000
Treatment 2 (N=162) 2,470 6,929 .680 .000 .257 .000
Treatment 3 (N=178) 4,088 24,858 .591 .000 .283 .000
26 Table 4: Tests for homogeneity of Variance
Levene’s statistic dfl df2
Auction bids (N=517) 0,38 2 514
5.3 Normalizing the data
When dealing with the problems of data that is not normally distributed, there are certain things that can been done according to Field (2009). Firstly, you could change scores but this not preferable because this will biases you statistical model. Secondly, you can transform the data to reduce the skew of the distribution. In section 5.2.1, it was shown that the distribution is positively skewed. Taking the logarithm of a set of number can correct for positive skew (Field, 2009). Furthermore, to overcome the problem of outliers, you could exclude some percentiles at the bottom and top of the data set.
In trying to normalize our data set, this is transformed by the natural logarithm.
Additionally, only the bids in the top 50 percentile were used. The results can be found below
Treatment 2 Treatment 1
27 Table 5: Tests for normality after transformation
Shapiro-Wilk Kolmogorov-Smirnov
Skewness Kurtosis Statistic P-value Statistic P-value
Treatment 1 (N=86) -2.687 6,887 .748 .000 .225 .000
Treatment 2 (N=78) .238 .141 .963 .022 .157 .000
Treatment 3 (N=87) .285 -.183 .962 .012 .105 .019
For all the treatments, it looks visually that the data follows a more normal distribution than before the transformation. Especially treatment 3 looks the most normally distributed now. However, both the Kolmogorov-Smirnov and Shapiro-Wilk tests show significant results (p < .05). So the assumption of normality is still violated. Noticeable is that the skewness and kurtosis of all treatments improved significantly. Especially, treatment 2 and 3 have a skewness and kurtosis that are close to zero.
As stated above, the Kolmogorov-Smirnov and Shapiro-Wilk showed that the
assumption of normality is still violated. However, some parametric techniques are ‘robust’ to this violation.ANOVA is generally known to be a robust test even when the data is skewed (Field, 2009). Consequently, a one-way independent ANOVA on the transformed data was carried out. Besides, the multiple regression method is also robust to normality violation if there is not highly skewed or kurtotic variables (Osborne & Waters, 2002). As shown in table 5, the skewness and kurtosis improved after the transformation of the data. Especially these of treatment 2 and 3 are very close to zero. Consequently, also regressions will be done to test the hypothesizes 3 and 4.
5.4 Reliability
In this experiment, the level of ambiguity tolerance was measured with four questions that are subtracted from the a 50 item scale called MAT-50 developed by (Norton, 1975). While the analysis of the reliability of the scale was originall based upon the complete scale which consisted out more than 50 questions, in this study only 4 has been used. Besides, these had to be translated to Dutch. Therefore, it is necessary to assess the scale reliability. According to Field (2009), the Cronbach’s Alpha is most commonly used to test for scale reliability. Of the total of 517 participants, 51 of them did not complete the questionnaire. These participants were removed from the computation of the Cronbach’s Alpha. Both the results of the reliability tests of the complete sample and the bid top 50 percentile are shown, which can be found below.
28
Table 6a Reliability Statistic Ambiguity Tolerance complete sample
Cronbach’s Alpha Number of items
.193 4
Table 6b Ambiguity Tolerance Scale improvements complete sample
Scale Item Cronbach's Alpha if Item Deleted
Item 1 -.112
Item 2 .124
Item 3 .480
Item 4 -.038
The Cronbach’s Alpha is said to be internally consistent and reliable if r> .70 (Field, 2009) . The Cronbach’s Alpha turned out to be lower than this, r=.193.. This indicates that the used scale is internally inconsistent and is not reliable. This was unexpected because the questions were substracted from the MAT-50 developed by Norton (1975) which is proven to be internally consistent by many studies If item 3 would be deleted, the Cronbach’s Alpha will increase to r= .480 which is a significant increase but still not reliable. As mentioned before, the bid top 50 percentile will be used in the analyses. The reliability test for this part of the sample is shown below in table 7ab.
Table 7a Reliability Statistic Ambiguity Tolerance top 50 percentile
Cronbach’s Alpha Number of items
.084 4
Table 7b Ambiguity Tolerance Scale improvements complete sample
Scale Item Cronbach's Alpha if Item Deleted
Item 1 -.342
Item 2 -.022
Item 3 .483
Item 4 -.180
Also in the bid top 50 percentile, the scale turned out be not reliable, r= .084. Again, if item 3 would be, the Cronbach’s Alpha will increase to r= .483 which is a significant increase but still not reliable. According to George and Mallery (2003) the following rules of thumb
29 regarding Cronbach’s Alpha are: “_ > .9 – Excellent, _ > .8 – Good, _ > .7 – Acceptable, _ > .6 – Questionable, _ > .5 – Poor, and _ < .5 – Unacceptable” (p. 231). Therefore, if item 3 would be deleted from the sample, the Cronbach’s Alpha will be at its highest unacceptable. Therefore, the information of the ambiguity tolerance scale is not reliable and cannot be used for further analyses.
5.5 Correlations
In Appendix C, the correlation matrixes can be found. To look if there are other interesting results of the experiment, besides the hypothesizes that will be tested, the most important findings will be discussed shortly.
Participants’ bid amount does have some interesting significant correlations. First of all, there is a significant correlation between bid amount and gender (r=-.143, p<.01). This suggests that men are willing to pay more for the Sony Smartband than women. Secondly, also a significant correlation has been found with students and age (r=.193, p<.01 and r=-.181, p<.01 respectively). This indicates that students and younger people in general are willing to pay more for such a product. This seems logical because students are likely to be young and vice versa. This correlations are not directly from importance for the rest of the study. However, this can be important information for managers which are selling such products. At final, participants’ bid amount shows us also a significant correlation with auction duration (r=.220, p<.01). This suggests that participants which placed high bids were longer in the auction. This can be explained by the fact that these people were interested in the product and therefore were carefully reading the advertisement and took longer to decide what bid to place. On the other hand, participants that were short in the auction placed a lower bid. This is likely to be explained by the fact that they were not interested in the product at all.
5.6 Hypothesis Testing
5.6.1 Testing Hypothesis 1and 2
Hypothesis one (H1) predicted that categorization of a product will lead to higher
pay. Furthermore, hypothesis two (H2) predicted a decrease in willingness-to-pay as categorical ambiguity would be present. To test for this, a one-way independent ANOVA test has been conducted.
30 the treatments (F(2,248)=5.365, p< .01). Additionally, there is a significant linear trend (F(1,248)=7.000, p< .01).
The F-value of the ANOVA test does only tell us that the experimental manipulation was generally successful but does not say what groups are affected (Field, 2009). To get more insights into what group means are different from each other , both planned contrast and post hoc tests have been conducted. The planned contrast test shows that there is significant difference between the mean value of treatment 1 and the combined mean value of treatment 2 and 3 (t(248)= 2.608, p<.05 (1-tailed)). However, no significant difference
between the means of treatment 2 and 3 have been revealed (t(248)= -.660, p>.05 (1-tailed)).
Table 8b Planned Contrast
Contrast Value of Contrast Std. Error t df
1 1.07220 .4111 2.608 248
2 -.07831 .1187 -.660 248
Post Hoc tests using the LSD test, confirmed what the planned contrast method showed. Treatment 1 is significant different from treatment 2 (p<.05) and treatment 3 (p<.05). However, treatment 2 is not significantly different from treatment 3(p>.05).
Table 8a ANOVA of Transformed Auction Bid Data Sum of Squares
df Mean Square F
Between groups (Combined) 16.371 2 8.186 5.365**
Linear term Unweighted 10.680 1 10.680 7.000**
Weighted 10.630 1 10.630 6.967**
Within Groups 378.415 248 .1525
31 Table 8c Multiple Comparisons, Turkey’s test
LSD Treatment (I) Treatment (J) Mean-Difference (I-J) Standard error 1 2 -.57525* .1931 3 -.49695* .1878 2 1 -.57525* .1931 3 .07831 .1926 3 1 49695* 1878 2 .07831 .1926
Consequently, hypothesis one (H1) is supported. This means that the suggestion of product category (categorization) leads to a higher consumers’ willingness-to-pay. However, hypothesis two (H2) is not supported. This means that giving more suggestions of the
product’s category does not lead to a decrease in consumer’s willingness-to-pay compared to a single suggestion of a product’s category. An additional finding is that the suggestions of multiple categories lead to a significant increase in the consumer’s willingness-to-pay than when no suggestions were given at all.
5.6.2 Testing Hypothesis 3 and 4
Hypothesis three (H3) and four (H4) were about the moderation effect of ambiguity tolerance on the relationships of categorization and categorical ambiguity on consumers’ willingness-to-pay. Unexpectedly, in section 5.4 has been shown that the Cronbach’s Alpha of the scale that measured ambiguity tolerance was, after deleting an item, r=.483. This can be defined as ‘unacceptable’ (George and Mallery, 2003). Therefore, the hypothesis three (H3) and four (H4) cannot be tested by conducting regressions. Consequently, the two hypothesis can neither be supported nor rejected.
32
5.7 Additional findings
After participants completed the auction, they were asked if they were able to place the
product within an existing category (Appendix B). The descriptives belonging to this question can be found in table 9a. These results suggest that there is little difference among the three treatments. The second treatment seems to have the highest mean, which means that in this treatment the participants had the most trouble placing the product within an existing
category. However, because the means are very close to each other, an one-way ANOVA has been conducted to test whether these small differences are significant. The results of the ANOVA (Table 9b) shows no significant difference between the mean values of the treatments (F(2,220)=1.363, p> .01).
Table 9a Descriptives Placing product within existing category
N Mean Std. Deviation Std. Error 1 77 2.52 1.021 .116 2 67 2.73 1.149 .140 3 79 2.44 1.071 .121
Table 9b ANOVA of Means
Sum of Squares
df Mean Square F
Between groups (Combined) 3.171 2 1.585 1.363
Within Groups 255.879 220 1.163
Total 394.786 222
Furthermore, it can be interesting to test for any relation between category certainty and the willingness-to-pay. To test for this, a linear regression has been conducted. The regression, with an explained variance of 4%, shows a main effect for category certainty (β=-.204, p<.01, R²=.04) on ln bid amount. This means that when the consumer is certain about the product’s category, the willingness-to-pay for the product increases.
33 Table 9c Linear Regression category certainty and WTP
Dependent variable: LN Bid Amount Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta
Can place product within existing category
34
6. Discussion
This study aimed to contribute to the literature of categorization and categorical ambiguity, and especially how the level of ambiguity tolerance of consumers affect these relationships. In this section, firstly the general findings will be discussed. Secondly, the theoretical and managerial implication will be proposed. At final, the limitations will be discussed together with suggestions for future research.
6.1Findings
In the first hypothesis was proposed that the categorization of a product would have a positive effect on consumers’ willingness-to-pay. The experiment was split up in different treatments to test if suggestions for one or more product categories would help participants to get a better understanding of the product. This study supported the first hypothesis. The suggestion of a product category leaded to an increase in the consumers’ willingness-to-pay for the product being auctioned. This result is in line with previous research. Categorization helps people to get better understanding of what the potential benefits of the product are to them (Mervis and Rosch,1981; Bloch, 1995; Cohen and Basu, 1987). So it helps to decrease the consumers uncertainty which leads to an increase in a consumers’ floor reservation price. (Wang et al., 2007).
The second hypothesis was about the how categorical ambiguity affects the
willingness-to-pay of consumers. It was proposed that categorical ambiguity would have a negative effect on consumers’ willingness-to-pay compared to the situation in which there is a single suggestion for the product’s category (categorization). Unexpectedly, the hypothesis was rejected. In fact, little difference was present in participants’ willingness-to-pay under condition of categorization (treatment 2) and under condition of categorical ambiguity
(treatment 3). The question is if the participants in treatment 3 actually made inferences about the product on basis of multiple categories. The literature showed in many cases, even when products where hard to place in a certain category, that consumers place their inferences on basis of single categories ((Murphy & Ross, 1994, 1999; Ross & Murphy, 1996).
Furthermore, the study of Gregan-Paxton (2005) showed that the variables familiarity and nature of the cue can influence if consumers place their inference on basis of a single or multiple inference strategy. A single category inference strategy is likely to occur when the perceptually cued category is more familiar than the conceptually cued category
(Gregan-35 Paxton et al., (2005), p.127). Therefore, an explanation of this unexpected result is that
participants experienced conflicting conceptual and perceptual category cues. Consequently, inferences made by the participants could have been equal in both treatments.
Unfortunately, hypothesizes 3 and 4 could neither be accepted nor rejected. It was proposed that the level of consumers’ ambiguity tolerance would affect the relationships tested in the first two hypothesis. The questions used, which were subtracted from the MAT-50 ambiguity tolerance questionnaire developed by Norton (1975), turned out to be unreliable and internally inconsistent. Therefore, there were no reliable measures available to test these hypothesizes.
An additional finding was that participants’ willingness-to-pay was higher in treatment 3 compared to treatment 1. This indicates that giving suggestions for multiple product
categories is better than giving no suggestions at all. From this you could argue that instead of carefully deciding which single suggestion for the product’s product category you should give to prevent categorical ambiguity, you just give multiple suggestions and this will have a positive effect. However, it is questionable if this is really the case in reality, Especially because there was little difference between treatment 2 and 3. As mentioned earlier, due to participants possibly experiencing conflicting conceptual and perceptual category cues in treatment 3, they may have base their inferences on basis of a single category. Therefore, it can be the case that the last treatment is not representative for the categorical ambiguity that was intended create. Also, it has been shown that category certainty leads to a higher
willingness-to-pay which is in line with previous research such as Wang et al. (2007) which showed that uncertainty leads to a lower willingness-to-pay.
6.2 Theoretical implications
It was intended to expand the literature by taking a first step in studying the moderating effect of ambiguity tolerance and how this affects consumers’ willingness-to-pay (WTP) in a real-life experiment. First of all, this experiment proves that categorization has a positive effect on consumers’ willingness-to-pay in a real-life experiment, which is in line with current
literature. Furthermore, the results of this study strengthens the line of research which is focusing on what influences consumers to take either a single or multiple inference strategy under circumstances of categorical ambiguity. It makes clear that simply giving more suggestions for the product’s category does not lead to consumers changing their bid behavior. Consequently, this study shows that research should continue on how consumers react on such suggestions in certain contexts. At final, this experiment showed that simply
36 subtracting a few questions from a well tested questionnaire does not result in reliable and internal consistent results. Future researchers that will make the use of questions from this questionnaire, should pre-test these questions and preferably include as many questions as possible from this questionnaire.
6.3 Managerial implications
The findings of this study suggests that more suggestions for a products´ category does not automatically lead to consumers´ paying less for a certain product than when just a single suggestion is given.
These findings have some implications for organizations and its managers that plan to bring a product to a market which possesses features of multiple product categories. This study proved that managers should be aware that taking a proactive approach in categorizing their product into a single category, can have positive effects on what consumers are willing to pay for it. However, categorization can draw the attention away from features that can be from importance to customers. Especially, because giving a category label makes consumers focus more on the features that belong to that category and distracts them from the ones that belong to other categories (Ross & Murphy, 1996). Besides, choosing and formulating one
suggestion for a category can be difficult because it is the question what the best suggestion is for a product that has features of multiple categories. By putting the focus on a single
category, managers could exclude a potential group of customers. Therefore, manager should be aware of what their customers seek in a product and which group is the most interesting to target when trying to get the product categorized.
On the other hand, this study showed no difference in willingness-to-pay between a single or multiple suggestion for the product’s category. Additionally, giving multiple suggestions leads to a higher willingness-to-pay compared to when no suggestions are given at al. On basis of this, it can be argued that managers can give multiple suggestions to prevent that customer become focused on just a small part of the features without this having any negative effect on the willingness-to-pay. However, it is recommended that managers will not do this without carefully considering that this could still turn out negative. Especially, because literature showed that categorical ambiguity can lead to customer uncertainty (Murphy & Ross, 1994). Additionally, variables such as familiarity and the nature of the category cue can affect how consumers react to multiple suggestions (Gregan-Paxton et al., (2005), p.127). These variables were not included in this experiment design. Therefore, it cannot be stated
37 with certainty that multiple suggestion for a product’s category will positively influences consumer’s willingness-to-pay in all circumstances.
At final, this study showed that organizations that are planning to bring products to the market, similar to the Sony Smartwatch , should target students and people younger than 30 first. Those groups are willing-to-pay more for such product and pricing strategies can be helpful to get the maximum revenues out of the different groups. A market skimming pricing strategy could be helpful to get the maximum revenues out of the mentioned groups.
Afterwards, other market segments can be addressed by lowering the prices of the product.
6.4 Limitations and suggestions for future research
When assessing the outcomes of this research, several limitations of the study need to be addressed. First of all, the collected data was not normally distributed. Many zero bids were placed which resulted in positively skewed data. Even with the use of natural logarithm and only using the bid top 50 amount, the data violated the assumption of normality which could have influenced the results of this study. Additionally, due to these transformations the remaining sample consisted only out of around 80 participants per treatment. Possible information could have been lost in this process. Therefore, future research should focus on getting a bigger sample which is preferably normally distributed to ensure more reliable results.
Secondly, the platform of Veylinx is argued to be useful in revealing customers’ willingness-to-pay but the online aspect can also have some drawbacks. In a online setting, it is likely that participants do some quick research on the auctioned product. Veylinx tries to minimize this by giving a five minute limit to complete the auction. However, price and product information is found quickly nowadays and this could have influenced the bidding strategy of participants. Especially the found information of the product features can have affect the influence of a category suggestion. To prevent this, experiments should be designed in such a way that little or no research can been done by the participants. This would be more ideal in establishing certain relationships between the variables. Of course, these studies do have to recognize that in real-life, research by consumers is likely to happen and will affect the strength of the study’s outcomes.
Thirdly, this study intended to get better understanding of the effect of consumer’s level of ambiguity tolerance on their willingness-to-pay. Unexpectedly, the measures of the participants level of ambiguity tolerance turned out to be internally inconsistent and
38 unreliable. The time limit in conducting this experiment made it not possible to conduct a extensive pre-test to test for consistency and reliability. In the future, researchers should try to incorporate the level of ambiguity tolerance in their studies. Especially because it is still believed that this variable is likely to affect consumers’ willingness-to-pay in some way. It is brought to the attention that subtracting questions from a reliable and internally consistent questionnaire as Norton (1975), still have to be pre-tested. Additionally, a significant amount of items should be used in studying this variable. A greater amount of questions is also needed to get a better measurement of consumers’ certainty of a product’s category. As mentioned earlier, an additional finding showed a positive effect of category certainty on willingness-to-pay. However, this was measured with a single item which is not preferably.
Lastly, to get better understanding on how the participants actually reacted on the suggested category labels, more measurements should have been needed. Due to the limited space for measurements after the auction, little insights were given about the participants behavior. Therefore, little information has been revealed about the process of customers in placing their final bid. Future research should aim to identify the variables that influence the behavior of consumers. Especially because some studies showed that under the circumstances of categorical ambiguity, consumers bases their inferences on basis of different kind of strategies (multiple vs. single). The study of Gregan-Paxton et al. (2005) made a first step in identifying some variables which can be from importance. Qualitative research can also be helpful to get deeper understanding of this process.
