Adoption of Informative Apps: What defines who wants to buy?

Adoption of Informative Apps: What defines who wants to

buy?

Adoption of Informative Apps: What defines who wants to buy? by Henriëtte Kip De Blokken 61 7894 CJ Zwartemeer 0651767167 henriettekip@live.nl Student number 1920901 Supervisor: Dr. J. van Doorn Second Supervisor: Dr. M.C. Non University of Groningen Faculty of Economics and Business

MSc Business Administration – Marketing Management & Research

Management Summary

Mobile communications are gaining importance. They enable people to reach and be reached at any time but also to request any type of information, at anytime and anywhere they want. Together with these new needs of consumers, new possibilities for companies have risen. Brand awareness and attitudes may be improved by launching a mobile phone application. Before starting to build an application, it is important to know what it should contain in order to be successfully adopted by your customers.

This research gives knowledge about the attributes that influence the adoption of mobile phone applications by consumers and professionals. Extensive research has been done on the topic of new product adoption and also in the area of mobile phone applications but this research focuses on informative apps specifically and tests the moderating effect of market segments. Since this research is conducted on behalf of het Kadaster, products and customers of het Kadaster are subject in this study.

After extensive literature research, a conceptual model is designed with five attributes; risk, enjoyment, ease of use, price and personalization. As a moderator is used the market segment, the dependent variable is the adoption of the application. With good argumentation, two hypothesis per attribute have been made. The first hypothesis is aimed to test if the attribute has an influence on the adoption of the application, the second hypothesis aims at testing if the effect of the attribute is stronger for consumers or professionals.

In order to test this hypotheses, a Choice Bases Conjoint Analysis has been conducted. A questionnaire has been designed and sent to about 11,000 customers of het Kadaster. Respondents were shown a set of three stimuli at a time and had to choose the most preferred. In these stimuli, the attributes have been operationalized by using mock ups. This is a pictorial of mobile phone which includes a prototype of the application. By operationalizing the attributes in different levels (high/low) the preference of respondents could be measured.

The results indicate that price, in general, has the largest influence on adoption of applications, followed by ease of use and enjoyment. Respondents rate price the cheaper the better. However, the effects of price and enjoyment are negative whereas the effect of ease of use is positive. Probably because this research focuses on an informative application, respondents do not rate enjoyment as important. On the other hand, they want a high level of easiness of use. Differences between consumers and professionals are found in personalization and ease of use. Consumers react positively, and professionals do not, on a high level of personalization and professionals tend to have a strong preference for a high level of ease of use.

Preface

After almost a year of hard work, my master thesis is finished. Without the help and support of several people, this research would not have existed in the current form.

First, I want to thank het Kadaster for giving me the opportunity to write my thesis about this topic. I gained new experiences and I learned a lot in a nice working environment.

I want to thank a few people personally. First I want to thank Bettine Baas, who was of great support during this process. She was very helpful with her feedback, put me in touch with other colleagues and had a never-ending enthusiasm. Furthermore, I want to thank Gaby Anink who also was very helpful with her critical feedback and ideas and Annemieke Poutsma, who helped me with MWM2, designing the questionnaire and ensured that it was send to all customers. Moreover, I want to thank all employees of het Kadaster who, in any way, contributed to this thesis.

I also want to thank Jenny van Doorn, for supervising me and giving me feedback on all my proposals and answering all my questions. Marielle Non for being of great support while designing this research and analyzing the results.

Finally, I want to thank family and friends who always supported me and always showed how proud they were.

Content

2.2 Adoption of mobile phone services and applications ... 10

2.2.1 Consumer based drivers and barriers ... 10

2.2.2 Social based drivers and barriers ... 12

2.2.3 Technology based drivers and barriers ... 13

2.2.4 Price... 13 3 Conceptual Framework ... 14 4 Research Design ... 19 4.1 Research method ... 19 4.2 Data collection... 21 4.3 Plan of Analysis... 23 5 Results ... 24 5.1 Descriptives ... 24 5.2 Model specification ... 26 5.3 Model fit... 26

5.4 Model estimation general model ... 26

5.4.1 Hypotheses... 28

5.5 Model estimation Latent Class model... 30

5.5.1 Defining number of classes ... 30

5.5.2 Model fit... 31

5.5.3 Model interpretation... 31

5.5.4 Hypotheses... 35

5.5.5 Covariates... 36

5.6 Hit rate... 37

5.7 Equalization prices and Willingness to pay ... 38

6.1 Summary of results... 40

6.2 Discussion... 40

6.3 Managerial implications ... 41

6.4 Scientific implications... 42

6.5 Limitations and future research ... 42

References... 43

1 Introduction

In today´s modern society, it is hard to imagine living without any form of mobile communications. Smartphones, tablets and notebooks belong to our daily routines in being connected to friends and being able to get any information at anytime and anywhere we want. Besides the ability of constant access to Internet, an innumerable amount of paid or free applications are available to serve any kind of need and are right away downloadable on the particular device. Games are long known for their popularity but informational applications are upcoming. For the mobile device user, it is of great importance to stay up to date when it comes to social networks, but also on daily news, weather, music or financial services. During the last few years we have seen a considerable development in services available through mobile devices, with new opportunities for users that are creating new motives for use and new channels for marketing communication and distribution. (Nysveen, Pedersen, Thorbjørnsen, 2005b)

Because of the large variety in the supply of mobile phone services, it is necessary to define what is actually meant by these services. A research of Kuo, Wu, and Deng (2009) gives direction by stating that "Mobile value-added services are digital services added to mobile phone networks other than voice services, including short message service, games, entertainments, web surfing, software applications and functions for achieving specific purposes". It is clear that all applications and services with increasing popularity, like Whatsapp and Wordfeud, belong to this category. In line with this, Anckar and D'Incau (2002) recognize that these mobile phone services have a promising future, due to the fact that certain customer values, like time critical, spontaneous, entertainment and efficiency needs, can be met by using these services.

Companies make good use of these "new" needs of consumers. Serious possibilities in marketing have risen with the emergence of mobile applications. Supermarkets provide online shopping lists, banks make it possible to securely engage in banking transactions and fashion stores offer you to stay up to date about the latest collection and sale promotions.

From business perspective, mobile phone apps are also gaining importance. This new marketing channel creates new opportunities for companies. Research has shown that consumers' brand attitude can be improved. Bellman et al. (2011) state that the high level of user engagement has a positive impact on the attitudes towards the sponsoring brand. An important difference in this case is the contrast with other forms of advertising. In comparison with a billboard, television commercial of direct email, a mobile application can be seen as "useful". This is probably due to the high level of user engagement, consumers have the feeling that they gain benefits by using the application. This contributes to the power of mobile phone applications as a form of advertising. (Bellman et al. 2011) Besides the fact that these so called "branded mobile phone apps" cause an increasing interest in the brand, they also increase the overall interest in the brands' product category. In that way, it would even benefit competitors.

developing and promoting the applications. By doing so, the application will meet needs and wishes of the target group and consequently, will be more successful. This research will contribute to the understanding of consumers' adoption of mobile phone applications by taking consumers' preferences into account.

Investigating determinants of new product adoption is important because product development is costly, life cycles are short and competition is fierce. Competitive advantage has to be clear from the consumers' perception, in such a way that he will buy your product instead of the competitors'. Another important issue is the return on the investments. In order to get return on a new (technological) product or service, it should be successfully introduced. Therefore, an introduction strategy, both targeted at new and existing customers, is necessary. Existing customers may be the most important target group for newly introduced products or services because they may be more likely to adopt the innovation as a result of their positive attitude toward the firm. (Prins and Verhoef, 2007)

The actual adoption of products by new or existing consumers can depend on a large range of factors; from product characteristics to consumer behavioural influences. Research has been done on several topics in the area of mobile phone applications but not specifically on the use of informational applications. Because it is not obvious which ingredients of mobile device applications contribute to a successful adoption of the application by the consumers, it will be investigated in this research. In order to do this, the following research question has been composed:

Which attributes influence the successful adoption of informational mobile device applications by consumers and professionals?

This research will look specifically at mobile device applications. A mobile device is any tool that allows access to a ubiquitous network beyond one specific access gate. The most common example of a mobile device is a mobile phone. For mobile marketing to be effective, this mobile device needs to be personal; that is, not shared with anyone else. This requires that each household member has their own device and that each device can be identified uniquely, as through a built-in SIM card. (Kaplan, 2012)

2 Literature Review

In the past years, extensive research has been conducted about the adoption or diffusion and the resistance of new (innovative) products. Sometimes, the line between adoption and resistance can be very small. Especially when the new product contains, from the perspective of the consumer, a certain amount of risk and the "old" product is less complex to use and already familiar to the consumer. (Hoyer and Macinnis, 2008)

Furthermore, Herzenstein, Posavac and Brakus (2007) find that the decision to purchase a product may depend on the needs and goals the product satisfies, as well as on price and other costs. These needs depend on the individual consumers and their price- and product perception.

Well known is the framework of Rogers, in which five different types of adopter groups are defined. These groups of consumers differ on certain characteristics and consequently on how and when they adopt new products. The Innovators are the first ones to adopt new products, when these are launched on the market, mainly because they are keen on gadgets and want to be the first. The Early Adopters are the next and very interested in new, advanced products that will make life easier. The next group consists of the Early Majority, consumers who are seeking for innovations but try to decrease any chance of risk and carefully consider a purchase. The fourth group exists of the Late Majority. These consumers are more conservative consumers who rely on tradition; they will not adopt a product very fast. The last group, the Laggards, are very sceptical and often resist adopting a new product. While the Late Majority and Laggards do not seem very interesting for companies to target, it may be useful to know why these groups are so resistant. In other words, which drivers and barriers are important to these groups. (Hoyer and Macinnis, 2008).

Important groups of people in an adoption process of new products are the so called innovators and early adopters, they are usually the first ones to try or buy. But also in these groups, barriers can cause the adoption time to grow. The research of Rogers (1995) tries to explain these barriers, which all are related to access specific issues of the customer. In addition, McCreadie and Rice (1999) distinguish six types of access: physical, cognitive, affective, economic, social, and political. Physical accessibility defines if the product or service is actually physical accessible. Cognitive accessibility refers to people’s ability to understand how systems work (technically) and the extent to which they master new technologies. Affective accessibility relates to attitudes and motivation with regard to the use of systems, such as confidence, efficiency and trust. Economic accessibility relates to benefits and costs, while social accessibility relates to cultural norms and political accessibility to power and knowledge gaps. Depending on to which extent the potential customer has all these accessibilities; he will buy or not buy the product.

will become the standard in the industry or the length of the product life cycle may be important factors. Furthermore, Hoyer and Macinnis (2008) state that also consumers' learning requirements do matter. Innovations need to be compatible in order to increase adoption and diffusion. Last but not least, the social relevance of the innovation influences the adoption. This relevance can be defined in observability, whether or not consumers can see others using the innovations, and social value, which reflects if the product is socially appropriate or desirable.

2.2 Adoption of mobile phone services and applications

Because this research is specifically about the adoption and diffusion of mobile phone applications, this will be discussed further in this literature review. The main difference with the usual marketing channels is that with the use of mobile phone applications, consumers decide for themselves to what extent they want to be involved. On the other hand, when consumers choose to download an application, they indirectly choose involvement with the brand on a high level. This high level of user engagement also causes increasing popularity of mobile phone apps. One can define them as “pull” rather than “push” advertising. The consumer talks to the brand, not the other way around. Second, consumers are exposed to the apps they opt into by downloading and they control how much information they reveal when customizing the app. (Bellman, Potter, Treleaven-Hassard, Robinson & Varan, 2011)

In line with the findings of Hoyer and Macinnis (2008), a distinction can be made within the factors that tend to have an influence on the adoption of mobile phone applications. The first one can be defined as consumer based drivers in which the consumers’ perception and behaviour defines if a product will be adopted. Next, the social environment of the consumer does influence their decision-making process and thereby the potential adoption of products. Finally, technical drivers and the price do define if the product will benefit the consumer in such a way, that he decides to buy the product. These four types of drivers and/or barriers will be discussed in the following subparagraphs. 2.2.1 Consumer based drivers and barriers

A very important driver or barrier can be consumers' perception of the product. To which extent a consumer needs a product or to which extent the product is a better replacement of the old product, has a strong influence in the decision making process. Another issue can be the context in which the customer needs the application. In case of utilitarian needs, which can be defined as usefulness and mobility, customers do not see mobile applications as ubiquitous when other alternatives are available. In other words, when a person is at home and is able to access internet via a PC, he will prefer doing so because using a PC to surf the internet is still easier than using a mobile phone. On the other hand, hedonic values like perceived enjoyment and concentration affect intention to use mobile internet. The more fun and cognitive concentration consumers feel with mobile Internet, the more likely they are to use it. (Yang, Lu, Gupta & Cao, 2012) This indicates that the fun factor indeed has a strong influence on consumers' adoption of mobile phone applications.

Furthermore, in the research of Wang and Li (2012), it is shown that perceived enjoyment has the strongest influence of all attributes on the purchase intention. Perceived enjoyment refers to “the extent to which the activity of using the computer is perceived to be enjoyable in its own right, apart from any performance consequences that may be anticipated” (Nysveen et al. 2005a)

appears to be particularly important as a driver for using experiential services, such as contact and gaming services. The effects of perceived enjoyment on intention to use mobile services are both direct and indirect.

In addition, Verkasalo, Molina-Castillo and Bouwman (2010) state that perceived enjoyment even adds to the hedonic value of advanced mobile services, having a positive effect on the intended adoption.

Besides that the mobile application should have a certain amount of enjoyment and fun in it, consumers have really emphasized the importance of usefulness and ease of use of the product. The research of Nysveen et al. (2005a) shows this importance for both attributes on the attitude towards the products and consequently the intention to use. One conclusion of this research is that certain basic levels of usefulness must be present for all kinds of mobile services no matter what the objective is. One can say that usefulness and easiness of use are somewhat overlapping. The definitions of Davis (1989) show otherwise. He states that perceived ease of use is “the degree to which a person believes that using a particular system would be free of efforts” while the definition of perceived usefulness is: “the degree to which a person believes that using a particular system would enhance his or her performance". In addition to this, the research of Nysveen et al. (2005a) showed the positive influence of perceived ease of use on the intention to use mobile services, especially when it concerned female and/or older users. In their case, it is really important to stress the user friendliness of the product in order to increase the possibility of adopting the product. The research of Herzenstein, Posavac and Brakus (2007) states that consumers’ self-regulation guides their product evaluations. Consumers perception of new products is riskier than their perception of existing products but they also have the potential benefit of addressing unmet needs or satisfying needs better. Whether or not the consumer adopts the product depends on their self-regulation. For example, "promotion-focused consumers may differ from prevention focused consumers in how they weigh the needs satisfied by a new product against its costs since prevention focused consumers try to avoid losses and negative outcomes and may be more sensitive to new product risk" (Herzenstein, Posavac and Brakus, 2007) Therefore, perceived risk can also play an important role. When a person already knows a brand from experience, it is more likely that they will try a new product from the same brand, the perceived risk seems lower.

Next, the perception of risk is important when consumers have to adopt new products in the near future. Specifically, it is shown that consumers who are considering adopting in the near future focus more on uncertainties related to the drawbacks of adoption than those considering adopting in the distant future. (Castaño, Sujan, Kacker, and Sujan, 2008) When consumers are considering adopting new products in the distant future, they tend to focus more on the symbolic uncertainties and performance, which are more related to the benefits. In addition to this, adoption decisions in a near time frame seem to be anxiety driven whereas feelings of optimism tend to dominate under a distant time frame. (Castaño, Sujan, Kacker, and Sujan, 2008)

identification by recognizing identities and corresponding wishes and needs can result in unique customer value. This, in turn, can enhance consumers’ perceived quality of the services received. (Wang and Li, 2012)

According to a research of Prins and Verhoef (2007), both mass advertising and direct marketing communication influences the adoption behaviour of existing consumers. However, mass communication does not affect this behaviour strongly, probably due to the focus on creating awareness and information provision whereas direct marketing is more action-oriented. When it comes to competitive brand advertising, it is proven to lengthen adoption timing of existing consumers. Because a successful introduction is crucial when it comes to the adoption process, this information about the influence of advertising can be very helpful. Mass advertising is very useful in creating awareness about the new product and direct marketing can decrease the individual customer adoption timing.

2.2.2 Social based drivers and barriers

When it comes to behavioural influences, it seems that also social factors must be taken into account. The behaviour of friends and family does influence the process of adoption and diffusion. The research of Taylor, Voelker and Pentina (2011) shows that the number of applications downloaded by friends, decide the likelihood that you download any application. The more an individuals' friends download, the more likely he is to download applications, regardless the type of application his friends have. In other words, social contagion causes the spread of mobile phone applications and therefore can be seen as a behavioural influence in the adoption process. This indicates that an individuals' reference group, actually defines your use of applications.

In line with this finding, Thompson and Sinha (2008) find that higher levels of participation in a brand community lead to both loyalty and oppositional loyalty in adoption behavior. Nowadays, consumers are encouraged to participate in user groups or to registrate on websites of certain brands. Also liking a page on Facebook is part of this phenomenon. By including customers in these little brand specific worlds, the sense of belonging is created and causes an increasing customer loyalty. Higher levels of participation increase the likelihood that a person will adopt a new product from the preferred brand and accelerates the time to adoption. At the same time, participation reduces the likelihood that a person will adopt a product from a competing brand and decelerates the time to adoption. (Thompson and Sinha, 2008)

2.2.3 Technology based drivers and barriers

With such highly technical and innovative products like mobile phone applications, the perceived risk on a technological base by the consumer is very important. Often, personal data or even financial data are being used by an application and therefore a target of criminal activities. Knowing that privacy issues are high priority and personal data are secured, can only reduce the resistance of consumers using the product.

In a conjoint analysis directed at the adoption of mobile games, the conclusion was that mobile systems have to be reliable and data-transmission has to be secure, while the systems have to be easy to navigate and fit into the daily routine of users. (Kleijnen, M., de Ruyter, K., Wetzels, M., 2004) This underlines the conclusions of other research that consumers want to limit their losses in the adoption process and their perception of risk can be decisive.

One other important aspect of today's highly technical and innovative products is design. Product developers are faced with difficult questions, about which they must make decisions, e.g., about: which products to create and how to optimize the user experience. In order to reduce uncertainty, it is important to ensure that the users’ perspectives can have an influence on such decisions. (Pals, Steen, Langley And Kort, 2008) Including user' perspective can only increase the chance on a successful adoption of the product because the end-user knows best how he wants the product to look like.

2.2.4 Price

3 Conceptual Framework

In the literature review, a distinction has been made between different drivers and barriers of mobile phone service or application adoption and five of these are chosen to be included in this research: risk, enjoyment, ease of use, price and personalization. From the literature research, these can be seen as the most important drivers or barriers of adopting mobile phone applications.

These drivers and barriers will be studied again in this research but the main difference is that the market segment, in which the end users are situated, influences the possible relations between the attributes and the adoption of applications. This moderator has been chosen because the market of het Kadaster exists of two groups; consumers and professionals. Professionals like real estate agents and employees of municipalities make extensive use of cadastrial information, mainly in their working environment. Consumers request these types of information when they buy a house, apply for a construction permit but also as a hobby, since Geocaching is becoming more popular.

Figure 1: Conceptual Model

A research of Dodds, Monroe and Grewal (1991) says that price can be both an indicator of the amount of sacrifice needed to purchase a product and an indicator of the level of quality. Higher prices lead to higher perceived quality and consequently to a greater willingness to buy. At the same time, the higher price represents a monetary measure of what must be sacrificed to purchase the good, leading to a reduced willingness to buy. In other words, the price has a negative effect on purchase intention. This is being strengthened by the research of Kalish (1985), who finds that the actual adoption depends on price and the individual evaluation of the product. Furthermore, this research states that the adoption of the product is contingent on the value of the product to the customer being greater as the price of the product.

As Herzenstein et al. (2007) stated before, the adoption of products depends on the needs and goals of consumers and the price and other costs of the product. The perception whether a price is too high or not depends on the consumer personally but also on the needs and goals of this particular person. The value attached to a product can differ among consumers and thereby the effort and money they are willing to spend. Employees of a business firm may be willing to spend more money on an application, provided that it is helpful for their proficiency, in contrast with consumers in the private market who adopt the application for their own benefits. Since this indicates that the effect of price for business is less stronger as the effect for consumers, it can be assumed that price has a stronger influence on adoption by consumers. Together, these studies suggest that price has a strong influence on the adoption of products in general. Thus:

H1a: The price of a mobile phone application has a negative effect on the adoption of the application by consumers and professionals

H1b: The influence of price on the adoption of a mobile phone application will be stronger for consumers compared to professionals

Fan and Poole (2006), the goal of personalization is seen as increasing sales directly and through cross sales besides building brands and increasing customer loyalty. In this case, increasing sales directly can also be seen as adoption rates because a lot of new customers are needed in order to do this.

The same research of Fan and Poole (2006) also states that continuous information about the particular consumer is needed to guarantee success. Information like personal preferences and interests cognitive ability, motives, demographic or psycho-cultural profiles, user behaviors and specific contexts is important. Thereby, Nysveen et al. (2005b) state that mobile services should enable users to express their personal and social identity with the use of on time, up to date and personalized services. It is expected that personal and social identity is more important to users in the private market compared to users in business markets and municipalities because users in private markets are confronted with social pressure (Nysveen et al. 2005a) whereas the working environment is much more relevant in the business market and for the municipalities. This assumption is being strengthened by a research of Chakraborty, Lala and Warren (2003), who researched the effectiveness of different attributes on the adoption of a business to business website. From this research has been shown that attributes like organization, privacy and security are being rated as very important whereas personalization has been rated as one of the least important attributes and business to business consumers seem to concentrate on the utilitarian aspects of services. This contributes to prior findings and therefore it can be expected that personalization has a stronger positive influence on the private market segment. Therefore the following hypotheses are formulated:

H2a: The level of personalization will have a positive effect on the adoption by consumers and professionals

H2b: The influence of personalization on the adoption of the application will be stronger for consumers compared to professionals

Castaño et al. (2008) state that when consumers are close to adopting a new product, uncertainties about money, time and effort are more important in consumers' eventual decision whether to buy the product or not. In some product categories, even the symbolic uncertainties may be present. These also depend on the perception of the consumer. Clearly, consumers have uncertainties when it comes to buying new products. These uncertainties include all the reasons why a product may not perform well. If these uncertainties are strong enough, it may decrease the chance of buying the product and therefore delay the adoption process. Increasing risks cause a decreasing adoption level and diffusion rates. (Rogers, 1995) Furthermore, the perception on the reliability of the service, which also can be seen as a risk factor, significantly influences both the social and business use of mobile ICT (Meso, 2005) Other research states that consumers’ concerns over the security and privacy of their mobile transactions remain high. The emerge of location-based services is causing consumers' sense of privacy and security to decrease. (Mahatanankoon, Wen and Lim, 2005) These privacy issues seem more relevant for consumers instead of professionals because consumers are affected in their own private environment, whereas professionals are only affected in their working environment, which is assumed to be less important.

and promotion to advancement and growth. Since in business markets, the goal usually is to increase growth and profits, it can be assumed that the level of risk will be less important to customers in business markets. Therefore, it is assumed that the strength of the influence of risk, depends on the type of market segment. Thus, the following hypotheses can be proposed:

H3a: The level of risk will have a negative effect on the adoption of the application by consumers and professionals

H3b: The influence of risk on the adoption of the application will be stronger for consumers compared to professionals

A very important attribute of mobile internet services is the extent to which it is enjoyed by the consumer. Enjoyment, which can be seen as hedonic value, affects the intention to use mobile internet according to the research of Yang, Lu, Gupta & Cao (2012) The research has shown that the more fun users find the experience with mobile internet, the more likely they are to use it. In addition, Verkasalo (2010) find that enjoyment is directly related to the intentions to use mobile services, especially when it concerns the use of maps in the service. Furthermore, the research of Wang (2012) shows that enjoyment is the most influential factor that can positively affect the intention of consumers to buy a mobile service. Next to this, it is proven that the enjoyment by consumers, influences the brand loyalty positively. Together, these researches provide a base for the assumption that a higher level of enjoyment will increase the adoption of mobile phone applications by consumers.

In addition, Nysveen et al. (2005b) find that the influence of enjoyment is even stronger when it concerns experiental services such as contact and gaming services. Users in private markets are more likely to use mobile phone applications to experience a certain amount of enjoyment.

Furthermore, the research of Chakraborty, Lala and Warren (2003) shows that the effectiveness of the entertainment attribute on website, is to be rated as the least important by business to business customers. Striking is that entertainment is often rated as the most important attribute in business to consumer websites. This indicates the difference in importance of the enjoyment factor for consumers vs. professionals. In addition, Bruner and Kumar (2005) found that the enjoyment is a more powerful determinant of attitude toward use in the consumer environment instead of usefulness, which has been found by Davis (1992) in the working environment. It is assumed, that users in the business markets have a strict work related purpose of use for mobile phone applications and therefore the level of enjoyment will be of less importance. In line with this, the following hypotheses are presented:

H4a: The level of enjoyment will have a positive effect on the adoption by consumers and professionals

H4b: The influence of enjoyment on the adoption of the application will be stronger for consumers compared to professionals

The research of Nysveen et al. (2005b) states that the intentions to use mobile services is significantly affected by the ease of use of the product. This ease of use even influences the attitude towards using the product, which is indirectly an indicator for adoption. Next to this, Nysveen et al. (2005b) state that the ease of use has a stronger effect on goal-directed services, which have more of a informational character. The definition of ease of use given by Davis (1989) "“the degree to which a person believes that using a particular system would be free of efforts” indicates that the level of easiness of use very much depends on the perception of the consumer.

Recent research has shown that the ease of use plays an important role in accepting technology. (Choi and Totten, 2012) In addition to this, a research by Venkatesh (2000) states that the perception of usefulness of a service or product in a working place differs from the perception in a leisure setting. The perception of ease of use in a working place depends only on the level of usefulness, which is correlated with the ease of use (Nysveen et al. 2005a), whereas the perception in a leisure setting depends on several factors like enjoyment, expressiveness and ease of use. This indicates that professionals attach more value to the easiness of use of a products compared to consumers because consumers will evaluate the product on more factors besides ease of use. From this may be expected that the perception of ease of use is a stronger influence in a working place compared to use in a private environment. Thus:

H5a: The level of ease of use will have a positive effect on the adoption by consumers and proffesionals

4 Research Design

In this section of the thesis, the research design will be discussed. In this research design, the following topics are addressed: the research method, in which the variables and questionnaire are shown, the data collection and the plan of analysis, in which the analysis of the data is discussed. The complete version of the questionnaire can be found in Appendix 1.

4.1 Research method

The aim of this research, as has been stated in the research question, is to find out which attributes contribute to a successful adoption of mobile phone applications by consumers and professionals. Therefore it is important to apply the right research methodology to obtain the right answers. In case of this research there has been chosen for a Conjoint Analysis. This kind of analysis seems to fit the purpose of this research because it is suitable to understand how respondents develop preferences for any type of object, which can be a product, service or idea (Hair, Black, Babin & Anderson, 1992) It is based on the idea that consumers evaluate the value of an object by combining the separate amounts of value provided by each attribute. Since the mobile phone applications exists of multiple attributes, which influence the adoption by consumers and professionals, conjoint analysis can be used to define the importance of the attributes and moreover, to show the difference in importance for the market segments. The objectives of this research also meet the objectives of a conjoint analysis: determining the contributions of predictor variables and their levels in the determination of consumer preferences and establishing a valid model of consumer judgments (Hair et al. 1992)

The main objectives have been discussed in this chapter already but it is also important to define the attributes that are studied in the conjoint analysis. As has been discussed in the literature review and has been shown in the conceptual model, a few key attributes have been identified as expected to have major influence on the adoption of mobile phone applications: price, risk, enjoyment, ease of use and personalization. Of course it is not possible to include all attributes that may have an influence, therefore the most important attributes have been chosen.

The first step in designing the conjoint analysis is choosing the methodology, which very much depends on the number of attributes that will be used. The Choice Based Conjoint analysis uses the unique form in which profiles are presented in sets of two or three profiles at a time, the respondent needs to choose the most preferred one. The advantage of this method is that it can be compared with consumers choosing a product in a shop, where they compare products as well. Models estimated from choice data have an advantage in predicting choice behaviour (Elrod, Louviere and Davey, 1992) In a Choice Based Conjoint analysis, interaction effect can be taken into account. For this particular research, the choice has been made to use Choice Based Conjoint analysis. Due to the difficulty of a task in which respondents have to rank the profiles, it is easier to let respondents choose between two or three profiles. An important factor is that the profiles in this research are very detailed and have to be studied in order to be understood properly, therefore it would be too difficult to rank a large amount of these profiles.

measurement scale and the estimation method will be discussed here in order to complete the research method.

At first, a preference model needs to be defined for the conjoint analysis. By far, the part-worth model is the most commonly used and assumes that each level of the attribute has a unique part-utility associated with it. This model also estimates the most parameters because it permits for the most general functional form. (Green and Srinivasan, 1990) The part-worth model is also being used in this research.

Secondly, the data collection method has to be defined. In this research there has been chosen for the full profile method which means that respondents evaluate profiles, based on all attributes, simultaneously. Green and Srinivasan (1990) mention a few advantages of these methods like: the ability of the full profile method to measure overall preference judgements directly using behaviourally oriented constructs, which can be linked to the context of this research in which the chance on a successful adoption by consumers and professionals is being measured. Furthermore, full profiles seem to be more relevant because they provide a complete representation of the product. When the research concerns a possible introduction of a new product, like the current research, these measures can be useful.

Thirdly, the stimulus set construction needs to be discussed. It is important to know which combination of attribute levels should be present in the profiles and which amount is needed in order to do an appropriate analysis. The defining of profiles must be orthogonal, which means no correlation between attributes, and balanced, so that each level in a factor appears the same number of times. (Hair et al. 1992) In order to meet these criteria, a fractional factorial design is the method used to select the subset of profiles. This method, in contrast to factorial design, only uses a subset of the possible profiles needed to estimate the results based on the assumed composition rule.(Hair et al. 1992) In this way, the number of profiles can be reduced while still an optimal design is maintained.

In order to obtain a balanced set of profiles, the program SSI Web 6.0 of Sawtooth Software has been used. With the use of this program, twelve sets of each 3 profiles have been created. Next to this, two hold out sets are taken into account. Given the number of five attributes with each respectively two and three levels, the software produced twelve sets.

In table 1, the efficiency ratings of this design can be seen. These ratings give an indication how well the model can be estimated. In this case, all scores are above 0.96., which are very good scores. At the same time, these scores indicate that one version of sets is enough to estimate the model properly. Because fourteen (twelve sets plus two holdout sets) sets of profiles is still a considerable large amount for respondents to rate, the efficiency rates of several amounts of tasks are compared, in order to check the possibility of reducing the amount of sets. The number of two holdout sets was being maintained during this process. Striking was that the efficiency rates were slowly increasing when the number of tasks dropped to respectively eleven, ten and nine tasks. The number of nine tasks turned out to be the best possible solution. To be complete, a number of eight classes have been calculated but resulted in a decreasing efficiency. (approximately 0.94).

Efficiency

12 tasks Efficiency 9 tasks

Risk 0.9680 0.9827

Enjoyment 0.9680 0.9477

Ease of Use 0.9654 0.9832

Price 0.9697 0.9824

Personalization 0.9882 0.9812

A none option is excluded from the sets of profiles. Reasoning behind this is the risk that a lot of respondents will choose the none option, without intensively studying the other sets of profiles. As has been mentioned before, the mock ups need to be studied carefully in order to fully understand the purpose. Therefore the none option has been left out of the design. In Table 2, the sets of profiles can be seen. To be able to verify results during the analysis, two holdout sets have been included.

As a fourth step of the overview provided by Green and Srinivasan (1990), the stimulus presentation method will be discussed. In their own research, they stress the importance of pictorial material used as a presentation method, which increases the interest of the respondent, provide an easier and less ambiguous ways of conveying information and this should enhance the realism of representing marketplace conditions. The research of Vriens, Loosschilder, Rosbergen, and Wittink (1998) states that computer-generated photo-realistic pictorial representations are an attractive way to obtain consumer input about product design attributes. Together, these reasons are convincing enough to make use of pictorial materials in this research and therefore there has been chosen to use mock ups. Mock ups can be defined as a scale or full-size model of a design or device, used for teaching, demonstration or design evaluation. In these mock ups, the difference between the attributes will be emphasized. This is operationalized by designing a mock up which is based on the lay out, website, products and services of het Kadaster. Together with these mock ups, a short scenario is shown to respondents in which they need to imagine that they are a customer that is looking for specific information, which is shown in the application. In the application, respondents are able to request for information about purchase prices of a building, plot area and buildings in the neighbourhood. Per attribute, two levels will be used which vary from a low to high range. In each mock up, a certain level of an attribute will be used, every time in a different combination. These attributes and levels are operationalized in the following way: risk is operationalized by showing a short text about the security of the applications, one with low risk and one with high risk. Enjoyment is operationalized by giving the opportunity for playing a game and in case of a low enjoyment level this option is not available. Ease of use is operationalized by creating an application in which it is very easy to get the information, literally with one click. The low level of ease of use shows a more complicated application with more options available. For the attribute of price, a range from € 2,99 to € 9,99 will be used. These prices are in line with the real prices in the Apple Store. A minimum price of € 2,99 has been chosen because het Kadaster does not sell products with a lower price and therefore this price is more realistic. According to Hair et al. (1992), the range of the levels should be set somewhat outside existing values but not to an unbelievable level. Therefore the maximum price of € 9.99 has been chosen. Finally, in case of personalization an application is shown in which the respondent is called by his or her name (Henny de Vries, mentioned in the scenario) and in which the latest search results and favourites are available. With a low level of personalization, these options are not shown. Finally, the measurement scale of the dependent variable has to be defined. Respondents are asked to choose one profile out of a set, this will result in dummy variables which will be analyzed using a Choice Based Conjoint analysis on segment level in Latent Gold Choice 4.5.

4.2 Data collection

which can be divided into consumers and professionals. These participants are approached with the use of a questionnaire. The questionnaire is composed with the help of online research company MWM2, which offers online questionnaire software.

As had been said before, the questionnaire exists of mockups which will be shown to the participant. These mock ups are presented in sets of three at a time. Respondents are asked which profile they would choose, if the given profiles were the only options available.With the mock up comes a short scenario to clarify the purpose of the mock ups and the price of each profile is written underneath the particular mock up. To measure the moderator, participants are asked to which target group they belong; consumers, companies or municipalities. At the end of the questionnaire, participants are asked some demographics like gender, age, education but also some questions about their usage of mobile phone applications and which device they use. These questions have to be answered in categories. The full version of the questionnaire can be seen in Appendix 1.

Set Number Card ID Risk Enjoyment Ease of Use Price Personalization

1 1 High Low Low € 2,99 Low

2 Low High High € 9,99 High

3 Low High Low € 6,49 Low

2 4 Low High Low € 9,99 Low

5 High Low High € 2,99 High

6 High Low High € 6,49 High

3 7 Low Low Low € 2,99 High

8 High High Low € 9,99 High

9 Low Low High € 6,49 Low

4 10 High High High € 2,99 Low

11 Low Low Low € 6,49 High

12 High High High € 9,99 Low

10 holdout 28 Low High Low € 2,99 Low

29 Low High Low € 6,49 Low

30 High Low High € 9,99 High

5 13 High High Low € 6,49 High

14 Low Low High € 9,99 Low

15 Low High High € 2,99 High

6 16 High Low Low € 2,99 Low

17 Low High High € 6,49 High

18 High Low Low € 9,99 Low

7 19 High High High € 6,49 Low

20 Low High Low € 2,99 Low

21 High Low High € 9,99 High

8 22 Low Low Low € 9,99 High

23 High High High € 6,49 Low

24 Low Low Low € 2,99 High

9 25 High Low Low € 6,49 Low

26 High High Low € 9,99 High

27 Low Low High € 2,99 Low

11 holdout 31 High Low High € 6,49 High

32 Low Low High € 9,99 Low

4.3 Plan of Analysis

5 Results

This chapter will discuss the results of the questionnaire. After t

completed filling out the questionnaire. Originally, 1215 respondents filled out the questionnaire but a number of these were deleted from the results due to several reasons:

 About 100 respondents answered "n

mobile phone applications?" Because judging the usability of applications is very important in this research, it was chosen to leave these respondents out of the results.

 A few respondents did not fill out their age

Because there is a possibility that these respondents did not take the questionnaire seriously, they were also deleted from the dataset.

5.1 Descriptives

A detailed analysis of the descriptive statistics of resulted in the following:Respondents

target group (of het Kadaster) they belong: consumers, real estate agents or government. These were chosen because these are the most common client groups of het Kadaster and in relation to this research; real estate agents and government are the professionals

professionals. More than half of the respondents were real estate agents and only 12% works with the government a third of all is consumer. Striking is the large amount of real estate agents that responded to the questionnaire, while the population that was approached by email

over the target groups. This may be an indicator for the interest that real estate agents have for a possible application of het Kadaster. The distribution of men and women is far

is woman. This may be due to the fact that real estate agents is often seen as a mens proficiency. average age of respondents was 46 years old.

respondent is 82 years old.

2% 11%

8%

62% 17%

This chapter will discuss the results of the questionnaire. After two weeks, 1107 respondents fully completed filling out the questionnaire. Originally, 1215 respondents filled out the questionnaire but a number of these were deleted from the results due to several reasons:

bout 100 respondents answered "never" to the question: "How often do you make use of mobile phone applications?" Because judging the usability of applications is very important in this research, it was chosen to leave these respondents out of the results.

A few respondents did not fill out their age or filled in a non-realistic number (0,5,10,444) Because there is a possibility that these respondents did not take the questionnaire seriously, they were also deleted from the dataset.

A detailed analysis of the descriptive statistics of the data has Respondents were asked to which they belong: consumers, real estate agents or government. These were chosen because these are the most common client groups of het Kadaster and in relation to this research; real estate agents and government are the professionals and consumers are non . More than half of the respondents were real % works with the government while a third of all is consumer. Striking is the large amount of real responded to the questionnaire, while the t was approached by email was evenly divided

This may be an indicator for the interest that real estate agents have for a possible application of het Kadaster.

nd women is far from even. From all respondents, 75% is man and only 25% This may be due to the fact that real estate agents is often seen as a mens proficiency.

espondents was 46 years old. The youngest respondent is 19 years old and the oldest

Over 60% of the respondents indicated that their highest education was HBO. This is not very strange because the required education to be a real estate agent is also on HBO level. 0.3 % of the respondents indicated that their highest education level is primary school. Because this is an extremely low percentage and could not be seen in the chart, it has been left out.

59% 12%

Figure 2: Target groups

VMBO/MAVO/LBO MBO

HAVO/VWO HBO University

wo weeks, 1107 respondents fully completed filling out the questionnaire. Originally, 1215 respondents filled out the questionnaire but

uestion: "How often do you make use of mobile phone applications?" Because judging the usability of applications is very important in

realistic number (0,5,10,444) Because there is a possibility that these respondents did not take the questionnaire seriously,

n and only 25% This may be due to the fact that real estate agents is often seen as a mens proficiency. The The youngest respondent is 19 years old and the oldest

Over 60% of the respondents ndicated that their highest This is not very strange because the required education to be a real estate agent is 0.3 % of the respondents indicated that their highest education level is primary school. Because this is an extremely low percentage and could not be seen in the chart, it has been left out.

29%

Consumers Real estate Government

Next to this, respondents were asked about their usage behaviou

These questions were included in the questionnaire because het Kadaster was interested in this information but also to be able to exclude respondents who never use mobile phone applications.

The following question was: "How often do you buy a product or service via a mobile phone application"? The results can be seen in figure 6. Almost 40% of the respondents indicates that they rarely buy a product of service. More striking is the comparison between figure 4 and 5: almost everybody uses mobile phone applications daily, but only a few (2.1%) actually buys a product on a daily basis

Furthermore, respondents were asked where they make the most use of an application; at home, at the office or at another location.

multiple answers. In all target groups, the most popular place was at home. Mo

respondents prefers their own house when making use of apps. The other locations are chosen by approximately 30% of the respondents. Striking is that real estate agents make more use of other locations than the other target groups. Next to t

and Android as a platform for mobile phone applications. Only a few respondents make use of an tablet, which is surprising because tablets are very popular in the markets these days.

72.7% 15.4% 0 200 400 600 800 1000 Daily Weekly

How often do you use mobile phone

applications?

Figure 5: Usage behaviour

Next to this, respondents were asked about their usage behaviour of mobile phone applications. These questions were included in the questionnaire because het Kadaster was interested in this information but also to be able to exclude respondents who never use mobile phone applications.

From the chart in figure 5 be seen that

quarters of the respondents daily uses mobile phone applications and 15 percent uses them weekly. This, again, underlines the popularity of

apps nowadays. The

respondents who indicated that they never make use of apps, are already excluded from this chart.

actually buys a product on a daily basis

ondents were asked where they make the most use of an application; at home, at Within this question, respondents had the possibility to choose In all target groups, the most popular place was at home. More than 80% of respondents prefers their own house when making use of apps. The other locations are chosen by approximately 30% of the respondents. Striking is that real estate agents make more use of other locations than the other target groups. Next to this, the majority of respondents uses iOS (via Iphone) and Android as a platform for mobile phone applications. Only a few respondents make use of an tablet, which is surprising because tablets are very popular in the markets these days.

3.6% 8.3%

Monthly Rarely

How often do you use mobile phone

applications?

Daily Weekly Monthly Rarely

How often do you buy a product or

service via a mobile phone application?

Figure 6: Buying behaviour r of mobile phone applications. These questions were included in the questionnaire because het Kadaster was interested in this information but also to be able to exclude respondents who never use mobile phone applications.

From the chart in figure 5 can be seen that almost three quarters of the respondents daily uses mobile phone applications and 15 percent uses them weekly. This, again, underlines the popularity of

apps nowadays. The

respondents who indicated that they never make use of apps, ready excluded from this

ondents were asked where they make the most use of an application; at home, at Within this question, respondents had the possibility to choose re than 80% of respondents prefers their own house when making use of apps. The other locations are chosen by approximately 30% of the respondents. Striking is that real estate agents make more use of other his, the majority of respondents uses iOS (via Iphone) and Android as a platform for mobile phone applications. Only a few respondents make use of an

22.1%

Never

How often do you buy a product or

service via a mobile phone application?

5.2 Model specification

To determine what the most appropriate specification for the utility function is, it is important to look at the representation of the attributes; a vector or a part worth model. In this case, price has to be determined because price is the only attribute that can be measured on an ordinal scale and therefore can be checked for linearity. The other attributes are nominal, so a distance between the options cannot be determined. The distances between the parameters can be seen in figure 7. From this graph can be concluded that price is part worth. The parameter of € 2.99 is positive and the parameters of the other prices are decreasing, indicating that these prices are less favourable by respondents. From this can be concluded that consumers rate price the cheaper the better, but the difference between €2,99 and €6,49 is larger than the difference between € 6,49 and € 9,99. Therefore cannot be concluded that the price is linear, although the difference is very small.

5.3 Model fit

After discussing the parameters and the hypotheses, the predictive validity of the model can be studied. This indicates how well the model predicts the choices. Therefore, several indicators have been used and these can be seen in table 3.

Baseline (0) Baseline Model R2 _{(0) R}2

0.6666 0.6726 0.4691 0.2964 0.3026

Table 3: Model fit

First, we can see the score of the baseline (0) model, which can be seen as the equal probability model and is a model without any specifications. This model will score a 33.34% ((1 - 0.6666) *100). So without any specifications, the model will predict 33.34% of the choices right. The following baseline model is a model with only an average choice probability. This model will predict 32.74 % ((1-0.6726) *100) of the choices right. The next column represents the score of our model. Our model predicts 53.09% of the choices right. So our model scores better than both baseline models.

The next indicator is the R square, which tells the reduction of prediction errors. In this case, the R square of a baseline model, without any information is 0.2964. The R square of our model is 0.3026.]

5.4 Model estimation general model

In order to be able to give an answer to the hypotheses, a general model needs to be estimated. This model has only one class but will provide answer to hypotheses H1a, H2a, H3a, H4a and H5a. In the following section the model will be interpreted and the hypotheses will either be accepted or rejected. Finally, the fit of the model will be discussed.

A general model has been estimated with Latent Gold Choice 4.5. As attributes have been used: Risk, Enjoyment, Ease of Use, Price and Personalization. Since this model only will have one class, it is not

-1 -0.5 0 0.5 1 1.5 2.99 6.49 9.99 Parameters

needed to include covariates. In a later stadium, it is being studied if the target group indeed has an influence on the preferences for attributes.

First of all, the relative importance of the attributes has been studied. The relative importance indicates which attribute is rated as the most important and which attribute as the least important. It shows the maximum effects for each attribute. In case of relative importance, this adds up to one. Because this model is not divided in segments, these importance numbers apply to the whole population. As can be seen in table 4, the price is by far the most important attribute. The second important attribute has only half the importance of price; Ease of use.

The third important attribute is enjoyment, but the score is rather low. Risk and personalization are the least important attributes. The scores of these attributes are that low, that it might be said that these attributes do not really matter at all.

The next step in the interpretation of the model is studying the parameters. In the questionnaire, the parameters were divided into two levels; high or low. In case of price, there were three levels estimated: 2.99, 6.49 and 9.99. For each of these levels, the model had parameters estimated. These can be seen in table 5.

Table 4: Relative importance of attributes

Relative Importance Risk 0.0185 Enjoyment 0.1430 Ease of Use 0.2592 Price 0.5374 Personalization 0.0420 Parameters Risk High -0.0317 Low 0.0317 Enjoyment Low 0.2450 High -0.2450

Ease of Use Low -0.4439

High 0.4439 Price € 9.99 -0.8506 € 6.49 -0.1400 € 2.99 0.9906 Personalization Low 0.0719 High -0.0719

From these results can be concluded that a high level of risk, has a very small negative influence on respondents. They are more likely to choose a mock up that represents a low level of risk. However, this influence can be seen as very small. Surprisingly, the level of enjoyment also has a negative influence on the preference of respondents. Apparently, respondents do not want any level of joy in an informative mobile application.

In comparison with the level of ease of use, this is the other way around. Respondents really do prefer a high level of ease of use and therefore the effect is also rather strong.

The parameter of price has also been discussed before (figure 6), but also from the parameter estimates in table 4 can be seen that the highest price has a really strong negative effect. The second price of € 6.49 still has a small negative effect and the lowest price of € 2.99 has a really strong positive effect. From this can be concluded that respondents really prefer a lower price.

The last attribute, personalization, again has really small effects on the choice of respondents. From the table can be seen that a low level of personalization has a slight positive effect and the high level has a small negative effect. From this can be concluded that personalization in a mobile phone application is not preferred by respondents.

Overall can be concluded that the price and ease of use have the largest positive influence whereas risk, enjoyment and personalization have a negative influence. In line with the results of the relative importance, again price and ease of use have the most influence because these parameters show the largest effects. In order to find out if these effects really do have a significant effect, the p values and Wald statistics have been studied. In case of the p values, a significance level of < 0.05 has been used. These results can be seen in table 6.

5.4.1 Hypotheses

Now the results have been studied and discussed, the acceptation or rejection of the hypotheses follows.

H1a: The price of a mobile phone application has a negative effect on the adoption of the application by consumers and professionals

This hypothesis can be accepted. The parameters in table 4 and figure 6 showed that the cheaper the price, the more prefered it is by respondents and a high price will have a negative effect on the ratings by respondents. The p value in table 5 shows that the effect is significant and therefore the hypothesis can be accepted.

Wald P value Risk _{4.8805 0.027} Enjoyment _{249.7177 0.000} Ease of Use _{930.4753 0.000} Price _{3762.5546 0.000} Personalization _{25.5695 0.000}

H2a: The extent of personalization will have a positive effect on the adoption by consumers and professionals

This hypothesis has to be rejected. The effect of personalization is, although very small, negative. A high level of personalization is not causing higher ratings by respondents. Furthermore, the effect is significant on a < 0.05 level.

H3a: The amount of risk will have a negative effect on the adoption of the application by consumers and professionals

This hypothesis can be accepted. A high level of risk has a very small negative effect on the preference of mobile phone applications by respondents. A low level of risk has a very small negative effect. From table 5 can be seen that this effect is significant.

H4a: The level of enjoyment will have a positive effect on the adoption by consumers and professionals

This hypothesis should be rejected. The level of enjoyment has a negative effect on the preference of consumers. Apparently, respondents do not want any level of "fun" in an informative mobile phone application. Again, the effect is significant.

H5a: The level of ease of use will have a positive effect on the adoption by consumers and professionals

5.5 Model estimation Latent Class model

The next step is estimating a Latent Class model. Again the dependent variable will be the choice of respondents and this will be defined by the same attributes that have been used before. Next to this, covariates have been included in the model; age, gender, target group and education. It is expected that the target group a respondent is in, will have an influence on their preferences of attributes. This section of the results chapter is aimed on to either accept or reject hypotheses: H1b, H2b, H3b, H4b, H5b. By studying possible differences in classes, an answer should be found.

5.5.1 Defining number of classes

In order to find out which solution of number of classes is the best to use, a solution up to ten classes has been conducted. Five information criteria have been studied and compared with each other to define the number of classes. BIC, AIC, AIC3 and CAIC (based on LL) - these statistics (information criteria) weight fit and parsimony by adjusting the LL to account for the number of parameters in the model. The lower the value, the better the model. Next to this, the AWE is used. This criterion is similar to BIC but takes classification performance into account. The criteria can be found in table 7.

AWE AIC BIC CAIC AIC3

1 class 16074.9261 15984.8132 16014.8697 16020.8697 15990.8132 2 classes 15676.9102 15008.9768 15119.1838 15141.1838 15030.9768 3 classes 15101.2892 14183.4574 14373.8149 14411.8149 14221.4574 4 classes 15020.0932 13773.9270 14044.4351 14098.4351 13827.9270 5 classes 15023.8952 13503.8187 13854.4774 13924.4774 13573.8187 6 classes 15208.5277 13388.9377 13819.7469 13905.7469 13474.9377 7 classes 15527.4347 13306.6821 13817.6418 13919.6418 13408.6821 8 classes 15.796.6484 13306.6821 13827.7079 13988.3247 13354.5976 9 classes 15997.4774 13183.0639 13854.3247 13988.3247 13317.0639 10 classes 16346.8078 13135.2799 13886.6913 14036.6913 13285.2799

Table 7: Information criteria up to a ten class solution