MSc Thesis Marketing Management

MSc Thesis Marketing Management

by: Hylke de Haan – student number 1646206 Title:

‘An empirical exploration of the quality-value-loyalty chain for mobile operating systems: the moderating roles of brand attachment, product involvement and brand usage’

Keywords:

Mobile operating systems E-service quality dimensions Perceived value (hedonic and utilitarian)

2

PREFACE

First of all, I would like to thank my supervisor dr. Janny Hoekstra for her positive and critical input and feedback, what has been crucial for the completion of this project. I really enjoyed working with dr. Hoekstra on this thesis, and her meetings and feedback were very helpful for taking further successful steps, especially when I started working on my first job last June. Furthermore, I would like to thank dr. Jia Liu for her useful feedback and assessment of this study.

3

ABSTRACT

This study examines the quality-value-loyalty chain for mobile operating systems. We research whether two different types of perceived value (utilitarian and hedonic value) can lead to brand loyalty of mobile operating systems. In addition, several e-service quality dimensions of the mobile operating system are included as antecedents to utilitarian and hedonic value. We propose that these relationships are moderated by product involvement, brand attachment and brand usage. The study uses survey data from 153 respondents.

Factor analysis confirms that e-service quality of a mobile operating system consists of two dimensions: user interface and content. These constructs both have a positive significant effect on the perceived utilitarian value; however, only the user interface is positively related to perceived hedonic value. For these relationships, no moderating effects are found for product involvement, brand attachment and brand usage, although these moderators all have a direct effect on hedonic value. Utilitarian and hedonic value both have a positive effect on brand loyalty, thus utilitarian value has a stronger effect. This means that loyalty of customers can be most effectively gained by investing heavily in the functional features of a product.

4

TABLE OF CONTENTS

2.2.1 | Antecedents of brand loyalty ... 8

2.2.2 | Antecedents of utilitarian and hedonic value ... 9

2.2.3 | Moderating effects ... 11

3 | Research Design ... 14

3.1 | Data collection and sample characteristics ... 14

3.2 | Survey development and measurement ... 15

3.3 | Factor analysis ... 17

3.3.1 | E-service quality dimensions ... 18

3.3.2 | Other constructs ... 20

3.4 | Reliability analysis ... 22

3.5 | Assumptions of multivariate analysis ... 23

3.6 | Econometric model ... 26

4 | Results ... 28

4.1 | Regression results for brand loyalty ... 28

4.2 | Regression results for utilitarian value ... 29

4.3 | Regression results for hedonic value ... 31

5 | Discussion, Conclusions and Implications ... 33

6 | Limitations and Recommendations for Future Research ... 37

5

1 | Introduction

Since smartphone technology is becoming more affordable and pervasive (Fogg, 2009), the mobile industry is at a turning point. By 2012, 58% of all Dutch consumers own a smartphone. It is the first year there are more smartphones sold than normal mobile phones (Telecompaper, 2012). In the last quarter of 2012, the two mobile operating systems Google Android and Apple iOS dominated the worldwide smartphone market with a combined market share of 85.3%, with Google Android as market leader with a share lof 66.2%. In comparison, a year before the combined market share of Android and iOS was only 65.4%. Nokia Symbian once again experienced a huge loss of market share of 14.5% within a year, with a remaining share of only 4.2% (Gartner, 2012). However, Nokia teamed up with Microsoft to regain its market position with the Windows 8 operating system, and contracted HTC and Samsung to produce top-end mobile phones for its operating system.

These developments indicate that the mobile operating system is gaining prominence as one of the most important considerations in the consumers’ decision making process for buying a smartphone. Moreover, it signals that consumers, instead of being loyal towards their mobile operator or phone brand, might become more loyal towards the mobile operating system. The question arises how consumers become loyal to a mobile operating system, while most marketing expenditures in the mobile industry are spent on promoting the network operator and phone brands. This development is in contrast to the comparative PC industry, where the focus is more on the operating system and indeed consumers tend to be loyal to a certain operating system.

Many technology companies have struggled with branding, since the speed and brevity of technology product life cycles cause unique branding challenges. Financial success is no longer driven by product innovation alone, or by the latest and greatest product specifications and features. Marketing skills are playing an increasingly important role in the adoption and success of high-tech products (Keller, 2008). One of the challenges in marketing e-services is that they are less tangible than products and more likely to vary in quality. Another reason why branding for high-tech companies is difficult, is because little is known about consumer behavior of smartphone users (Wagner, 2011).

6 A scale is constructed to measure the most important e-service quality dimensions. Next, we examine how they relate to the utilitarian and hedonic components of perceived value, which in turns results in brand loyalty. Furthermore, we investigate whether usage of either the Apple iOS or Google Android mobile operating system is of influence on the relationship between the e-service quality dimensions and perceived value. In addition, product involvement and brand attachment are also enclosed as moderators on this relationship. All constructed scales are measured on a seven-point Likert scale. Data is collected by an online survey among smartphone users in the Netherlands.

Developing insights into the determinants of value perception and brand loyalty for a mobile operating system is important for both marketing researchers and the managers in charge of designing the service. The results can be used by firms on a strategic level by improving and invest more heavily in the most important service attributes in order to increase the perceived value of the product. Investments and improvements in those e-service quality dimensions in order to increase the perceived value could enhance loyalty. On a marketing level, this information can be translated into marketing campaigns, whereas more emphasis and expenditures can be put on the most important e-service quality dimensions in order to gain brand loyalty.

7

2

|

L

R

2.1 | Conceptual model

Figure 1: Conceptual model

8

2.2 | Hypotheses

2.2.1 | Antecedents of brand loyalty

Perceived value

In literature, perceived value is traditionally acknowledged as a function and trade-off between sacrifice (price) and utility. Zeithaml (1988) defines perceived value as ‘the overall assessment of the utility of the utility of a product/service as measured by received benefits deducted by perception of sacrifice’. Moreover, Monroe (1990) explains the concept as ‘representing a trade-off between the quality of the benefits they perceive in the product relative to the sacrifice they perceive by paying the price’.

However, the consumption process for certain products is more than only a search for utilitarian performance; it can also include a search for experience and emotions (Holbrook & Hirschman, 1982). When consumers choose expensive devices, such as mobile phones, hedonic reasons dominate but the behavior is justified in terms of utilitarian value (Heath & Soll, 1996). The view of value as a trade-off between only quality and price is too simplistic (Bolton & Drew, 1991).

Holbrook & Hirschman (1992) argue for an experiental perspective that includes the symbolic, hedonic and esthetic aspects of the consumption process. In addition, Batra & Ahtola (1990) supported the presence of distinct utilitarian and hedonic components, which are referred to as ‘thinking and feeling’ dimensions. A multidimensional conceptualization of perceived value is needed to adequately capture the presence of both cognitive (utilitarian) and affective (hedonic) factors in the nature of value, as consumption experiences usually involve more than one type of value simultaneously (Holbrook, 1994; Sweeney & Soutar, 2001).

9 worldwide market share of more than 66%. In this case, consumers have a clear choice in operating systems for a phone, but prefer Android most of the times in their decision making process for a smartphone and thus tend to be loyal towards the operating system. Since consumer perceived value is associated with brand loyalty as a post consumption outcome (Parasuraman & Grewal, 2000), we hypothesize that both components of perceived value have a positive effect on brand loyalty:

H1: Perceived utilitarian value is positively related to brand loyalty

H2: Perceived hedonic value is positively related to brand loyalty

2.2.2 | Antecedents of utilitarian and hedonic value

Augmented level of e-service quality

In the literature on services, two approaches to study e-services can be distinguished: one that proposes the use of existing service quality theory as a basis for further empirical research (Parasuraman & Grewal, 2000; Zeithaml et al., 2000), and a more empirical approach, which has generated new categories for e-services (Dabholkar, 1996).

This study uses the latter approach to determine the key e-service quality dimensions of a mobile operating system. Grönroos et al. (2000) propose that for e-services the traditional service concept, consisting of the core service, facilitating and supporting services, needs to be extended with an additional augmented level with elements of customer participation and communication. Van Riel et al. (2001) further state that the core, supporting and complementary services can be interpreted as ‘what’ customers receive, whereas the user interface describe ‘how’ the service is delivered. In this study e-service quality is measured on the augmented level. This does not mean that the basic, more generic level is not important, but consumers simply expect a high general level of quality associated with this factor (such as reliability, support and security) for all mobile operating systems. In other words, the e-service quality attributes on this level serve as prerequisites. Hence, the generic e-service quality level can be labeled as the minimal requirements for a mobile operating system and differentiation occurs at the augmented level.

Augmented e-service quality dimensions: User interface and Content

10 space where interaction occurs between humans and machines, and through which the customer accesses the services (Van Riel et al., 2001).

In addition to Grönroos et al. (2010), we extend the augmented level with a second component: content. Content, e.g. (website) information, is widely acknowledged as an important determinant of e-service quality (Abels et al., 1999; Santos, 2003; Fassnacht & Kroese, 2006; Sohn & Tadisina, 2008). Content can be described as all software, such as (web) services and applications, installed on or available for the mobile operating system. Availability and functionality of relevant applications are of strategic importance for the success of a mobile operating system, along with a well-structured application store.

The user interface can be interpreted as the facilitator by which the content is accessed. The user interface is a fixed foundation, which is mainly developed by the internal organization (Apple or Google). Content is more flexible in terms of offering, choice and change. It is a result of combined action of the internal party and thousands of external developers for the available software. These developers are very different in size and in what type of applications they develop.

Content in a broader context also includes the pre-installed software. To demonstrate its importance, with the introduction of the new operating system iOS6 by Apple, the traditional navigation software of competitor Google Maps was replaced by inferior self-created Apple navigation software full of bugs. The introduction has led to massive critics and even apologies by the company itself. Consequently, this probably resulted in a lower perception of the overall utilitarian value of the new mobile operating system.

H3: E-service quality of user interface is positively related to perceived utilitarian value

H4: E-service quality of content is positively related to perceived utilitarian value

The way an operating system is designed or the ease of how it is used can also make it positively fun to use. Dabholkar (1996) found a strong positive effect of enjoyment of using self-service technology on perceived overall self-service quality. Enjoyment captures the hedonic dimension of consumption and it is proposed that the quality of the user interface has a positive influence on the perceived hedonic value.

H5: E-service quality of user interface is positively related to perceived hedonic value

11 2.2.3 | Moderating effects

Product involvement

According to Malär et al. (2011), product involvement is defined as the personal relevance of the product, which is determined by the extent to which the product is interesting and important to the consumer. Park and Young (1986) make a distinction between cognitive involvement and affective involvement. The first one is derived from utilitarian motives, the second from emotional motives. When personally relevant knowledge is activated in memory, a motivational state is created that energizes or drives consumers’ cognitive behavior (attention, comprehension and information search) or affective responses (emotions). Therefore, on the one hand product involvement can have influence on the relationship between e-service quality dimensions and utilitarian value by the motivation of cognitive behavior.

H7: The relationship between the e-service quality of user interface and the perceived

utilitarian value is positively moderated by product involvement

H8: The relationship between the e-service quality of content and the perceived utilitarian

value is positively moderated by product involvement

On the other hand, the activation of a motivational state that drives the affective responses can lead to influence of product involvement on the relationship between the e-service quality dimensions and hedonic value. The stronger the nature of these affective responses, the more influence it probably will have on this relationship.

H13: The relationship between e-service quality of user interface and perceived hedonic value

is positively moderated by product involvement

H14: The relationship between e-service quality of the content and perceived hedonic value is

positively moderated by product involvement

Brand attachment

12 Brand attachment is more than emotions; it is reflected by mental representations (rich cognitive schemata) that include brand–self cognitions, thoughts, and autobiographical brand memories (Berman & Sperling, 1994; Mikulincer & Shaver, 2007) that measures of emotions may not capture. Hence, we hypothesize that the cognitive aspect of brand attachment can have a positive effect on the relationship between the e-service quality dimensions and utilitarian value:

H9: The relationship between e-service quality of user interface and perceived utilitarian

value is positively moderated by brand attachment

H10: The relationship between e-service quality of content and perceived utilitarian value is

positively moderated by brand attachment

Though cognitive in its representation, this brand–self linkage is inherently emotional (Mikulincer & Shaver, 2007; Thomson et al., 2005), involving myriad and potentially complex feelings about the brand. This leads to the proposition that the emotional aspect of brand attachment can have a positive moderating effect on the relationship between the e-service quality dimensions and hedonic value:

H15: The relationship between e-service quality of user interface and perceived hedonic value

is positively moderated by brand attachment

H16: The relationship between e-service quality of content and perceived hedonic value is

positively moderated by brand attachment

Brand usage

13 H11: Brand usage has a significant positive moderating effect on the relationship between

e-service quality of user interface and perceived utilitarian value

H12: Brand usage has a significant positive moderating effect on the relationship between

e-service quality of content and perceived utilitarian value

H17: Brand usage has a significant negative moderating effect on the relationship between

e-service quality of user interface and perceived hedonic value

H18: Brand usage has a significant negative moderating effect on the relationship between

14

3

|

R

D

3.1 | Data collection and sample characteristics

The target group consists of smartphone users in the Dutch mobile phone market. The respondents are required to either use the Apple iOS or the Google Android mobile operating system. Potential respondents were approached by e-mail or Facebook to participate in an online survey with a link to the website ThesisTools.com. Respondents were gathered with an incentive to participate, namely to win a prize (a tablet).

A total of 153 respondents participated in this survey. After selection, 38 respondents were denied for analysis and a total dataset of 115 respondents remained. In all 38 cases, respondents were excluded due to data cleaning on missing values, more specific because of abandoning the survey before completing.

At first sight no outliers are identified, mainly due to the pre-established range of answers respondents could give. The only open question about age does not result in outliers. Furthermore, two respondents failed to fill in the demographic question about gender, though their gender could be ascertained based on their email address.

The descriptive statistics are displayed in table 1. The sample is approximately equally distributed along brand usage (41.7 vs. 58.3%). Most participants are male (73.9%) and the average age of the respondent is 25.19 years (age ranging from 19-51 years). Given the overall high education of respondents, in combination with an overall low income and age distribution, it is assumed that a fair share of the sample consists of students.

Table 1: Descriptive Statistics (N =115)

Variable Count Percentage Brand usage # % Apple iOS 48 41.7 Google Android 67 58.3 Gender # % Male 85 73.9 Female 30 26.1

Age (in years) # %

15

3.2 | Survey development and measurement

Due to the lack of research on mobile operating systems, we develop a new measurement scale to measure the two augmented e-service quality dimensions of a mobile operating system, user interface and content, based upon existing e-service literature where possible.

The user interface scale compromises four separate constructs: customization, look and feel, ease of use and structure and layout. The scale is partly based on Santos (2003), who identifies ease of use, appearance, structure & layout and content as the most important features of the augmented e-service quality dimension. Given the importance of content to the success of a mobile operating system, content is considered as a separate e-service quality dimension, and the other constructs constitute the user interface. The scale is completed with the inclusion of customization.

Site aesthetics (Zeithaml et al., 2000), website design (Dabholkar, 1996; Szymanski & Hise, 2000), appearance (Cox & Dale, 2001; Santos, 2003) all refer to the aesthetic or sensory e-service quality aspect of the operating system. This is labeled as the ‘look and feel’ dimension and is defined as the extent into which the operating system has an attractive visual design, an user-friendly handling and a proper use of color, graphics, images and animations. Meuter et al. (2000) propose that ease of use is one of the most important features of the user interface and affects service quality perception. Ease of use can be interpreted as the extent into which the required actions are natural, logical, inevitable and easy to understand.

Structure and lay-out refers to the extent into which the operating system is logically organized and has a clear structure (Santos, 2003). Customization (or: personalization) is considered by Zeithaml et al. (2000) as an important e-service quality dimension. In this study, customization is defined as the extent into which the menus and design of the operating system can easily be tailored to consumer preferences.

16 For the constructs perceived value, brand loyalty, brand attachment and product involvement existing scales are adopted.

For perceived value, the 19-item PERVAL scale by Sweeney & Soutar (2001) is utilized. Sweeney and Soutar (2001) distinguish between utilitarian and hedonic components of perceived value. They propose that perceived value is a construct of on the one hand price-worthiness and

functional value (utilitarian motive), and on the other hand emotional and social value (hedonic

motive). Functional value is defined as the utility derived from the perceived quality and expected performance of the product, and five items from this scale are used to measure the utilitarian value construct. The price-worthiness scale is not included since no direct costs for consumers can be attributed to both mobile operating systems, in contrast with the comparative personal computer industry. Apple iOS is exclusively part of the Apple iPhone, and Google Android is available as open source software. Price is rather complex as a trade-off between phone brand, network operator and in lesser extent the mobile operating system. Emotional value is defined as the utility derived from the feelings or affective states that a product generates, and social value can be explained as the utility derived from the product’s ability to enhance social self-concept. Three items from both scales are selected to constitute a six item scale to measure the hedonic value construct.

T

ABLE

2

:

M

EASUREMENT OF

C

ONSTRUCTS

In general, I have a strong interest in smartphones My smartphone is very important to me

I get bored when other people talk about smartphones My smartphone means a lot to me

Brand loyalty BL1 BL2 BL3 BL4 BL5

I say positive things about [Brand X] to other people I recommend [Brand X] to someone who seeks my advice I encourage friends and relatives to do business with [Brand X] I consider [Brand X] as my first choice

I want to do more business with [Brand X] in the next few years

Utilitarian value UV1 UV2 UV3 UV4 UV5

[Brand X] has consistent quality [Brand X] is well made

[Brand X] has an acceptable standard of quality [Brand X] has poor workmanship

[Brand X] would perform consistently

Brand attachment

BA1 BA2 BA3 BA4

To what extent is [Brand X] part of you and who you are?

To what extent do you feel that you are personally connected to [Brand X]? To what extent do you feel that you are emotionally bonded to [Brand X]? To what extent does [Brand X] say something to other people about who you are?

Hedonic value EV1 EV2 EV3 SV1 SV2 SV3

Usage of the operating system makes me feel good Usage of the operating system gives me joy

Usage of the operating system makes me feel relaxed I consider my operating system as a status symbol I feel proud of being the user of this operating system Usage of this mobile operating system makes a good impression on other people

E-service quality of user interface

UI1CUS UI2CUS UI3LF UI4LF UI5LF UI6EoU UI7EoU UI8EoU UI9SL UI10SL

The design can easily be adopted to my personal preferences The menus can easily be tailored to my personal preferences My operating system has an attractive visual design

My operating system has an user-friendly handling

My operating system has a proper use of color, graphics, images and animations

My mobile operating system requires a minimal number of actions to get where I want

The required actions to perform a task are naturally and logical It is easy to understand which actions need to be taken to perform a task

The content on the operating system is logically organized My operating system has a clear structure

E-service quality of content

C1SI C2SI C3SI C4AC C5AC C6AC C7AS C8AS C9AS C10AS

My operating system gives me the ability to access and download relevant content and applications

The content on my operating system is well integrated with other (web) services I often use

All relevant media formats are supported

I am satisfied about the amount of services and applications on my operating system

The content on my operating system is relevant to me

I am satisfied with the content and applications on my operating system I am satisfied with the amount of applications available for my operating system

I am satisfied about the price of paid applications I am satisfied with the availability of free applications

17

3.3 | Factor analysis

The survey consists of 44 items from which the constructs are abstracted. Factor analysis is used to identify a smaller set of salient variables from a larger set for use in subsequent multivariate analysis. Two separate factor analyses are conducted. First, all 20 items representing the e-service quality attributes are proceeded to test whether the proposed two e-service quality dimensions are indeed the underlying dimensions of the mobile operating system. Next, the remaining 24 items are subject of analysis to support the existence of the separate constructs in literature.

Factor analysis attempts to identify underlying variables, or factors, that explain the pattern of correlations within a set of observed variables. The factor analysis process is divided in two steps: first, the items are tested on factorability, and second, the factors are extracted.

To identify underlying factors, several criteria for factorability were used (Hair et al., 2006; Malhotra, 2007):

• Correlation between the items (≥ .3, examine ≥ .8); • Bartlett’s test of sphericity;

• Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy (≥.6); • Diagonals of the anti-image matrix; (≥ .5);

• Communalities (≥ .4);

• Primary factor loadings ( ≥ .55);

• Cross-loadings: (primary-secondary discrepancy ≥ .3).

After checking for factorability of the data and deletion of undefined items, the accurate number of factors can be determined. For this end, a three step process is used (Field, 2000; Rietveld & Van Hout, 1993):

1. Retain only those factors with an eigenvalue larger than 1 (Guttman-Kaiser rule); 2. Keep the factors which, in total, account for minimal 60 percent of the variance; 3. Make a scree-plot and keep all factors before the breaking point or elbow.

18 3.3.1 | E-service quality dimensions

The check for factorability on the service quality items resulted in the elimination of 7 of the 20 items. First, the results in the correlation matrix indicate that one variable correlated possibly too highly with two other variables (> .862). A face validity check concluded that two items basically measured the same construct in an identical way. Therefore, one item was deleted. No variables had too low correlations (< .3). Furthermore, the Bartlett’s test ( .000) secured that the variables are related in the population. The high KMO score of .895 indicates good appropriateness for factor analysis (> .6). The diagonals of the anti-image correlation matrix are all > .8. However, three items were deleted due to large loadings on multiple factors. Also another variable had a too low primary loading of .405. After deletion of those 4 items, another 2 items proved to have too high cross-loadings, and consequently are removed from the data set. After deletion of items, a total of 13 items

remained. The remaining items had primary loadings ranging between .598 - .815, minimal communality of .429, and are preserved for the extraction of factors (table 3).

Extraction of factors

According to the Kaiser’s criterion, a total of three factors can be retained (table 4). However, a variance of 1.065 is still low and only slightly more than a single variable forming a factor, since each variable has a variance of 1.0 due to standardization. The minimal percentage of variance approach suggests a minimum of two factors (62.9%) to reach a satisfactory level of variance. In addition, the scree plot has a distinct break between the steep slope of factors, with large eigenvalues and a gradual trailing off associated with the rest of the factors (figure 2). The scree plot strongly indicates a total of only 2 factors, since there is a clear point identifiable where the scree begins. Experimental evidence indicates that the point at which the scree begins denotes the true number of factors. Furthermore, the rotated component plot is analyzed in order to verify the right amount of

Table 4: Eigenvalues and variances

Factors Eigenvalue % of variance Cumulative %

1 6.704 51.568 51.568

2 1.472 11.323 62.891

3 1.065 8.192 71.083

4 .723 5.559 76.642

Table 3: Rotated factor matrix

Factors

Items User interface Content Communality

20 3.3.2 | Other constructs

For all the non e-service quality scales, factor analysis is used to check if the constructs from literature are derived from and supported by the data. For this set of items, the same process for analysis is used as for the e-service quality items.

The correlation matrix assigned five sets of variables with a suspected value of > .8 and needed to be examined for removal from factor analysis. Still all sets of variables scored < .9, and after a face validity check it was concluded that each variable has a different composition and a unique contribution. Again, a good KMO score ( .858) indicated appropriateness for factorability. The null hypothesis that the correlation matrix is an identity matrix was rejected by the Bartlett’s test of sphericity ( .000). The diagonals of the anti-image correlation are > .728. The remaining items had primary loadings ranging between .581 - .870 and had a minimal communality of .516 (table 7). Consequently, all items met the criteria for factorability and none of the 24 items were removed for extraction of factors.

The eigenvalue represents the total variance explained by each factor and only factors with an eigenvalue greater than 1.0 are retained. As can be seen in the table 5, a maximum of six factors can be extracted. At least three factors should be extracted to meet the requisite of minimal 60 percent of total variance. More ideally, six factors account for 78.2% of total variance. No conclusion can be drawn

from the shape of the scree plot. In contrast to the former scree plot, no clear distinction can be made at which point the scree begins. It is rather a more gradual line. Also no conclusion can be drawn from the rotated component plot. Interpretation proved to be too difficult with 24 items resulting in six factors in one plot.

The results of the factor analysis indicate that the data successfully can be reduced to six underlying factors. A total of five factors were expected from literature: brand loyalty, brand

attachment, product involvement, utilitarian value and hedonic value. However, the factor

analysis identified the division of the hedonic value construct into two separate constructs:

emotional and social value. This is supported by the theory by Sweeney & Soutar (2001) and

Punnyimoorthy et al. (2007). The means of both constructs are also significantly different (EV: 5.02; SV: 3.01, p > 0.05). Nevertheless, one scale for hedonic value is included in the conceptual model.

Table 5: Eigenvalues and variances

Factors Eigenvalue % of variance Cumulative %

21 Arnold & Reynolds (2003) identify social motivation as one of six types of hedonic motivations, along with adventure, gratification, role, value and idea motivations. In this study, emotional value is measured on a more generic hedonic level, whereas social value can be attributed as measurement of one of six more sub-levels of hedonic value. Consequently, since generic levels of utilitarian and hedonic value are desired in this study, it is justified to delete the three social value items from the hedonic value scale. Hence, the hedonic value scale will consist of the three emotional value items.

Table 6: Rotated factor matrix using VARIMAX procedure and communalities

22

3.4 | Reliability analysis

Internal consistency reliability is used to assess the reliability of a summated scale where several items are summed to form a total score. A coefficient alpha of .60 or less generally indicates unsatisfactory internal consistency (Malhotra, 2007). As can be seen in table 7, the values of the scales range from .80 to .93, indicating construction of good, reliable scales. Table 7 also displays the means, standard deviations and correlation coefficients for all constructs. The correlation coefficients confirm the existence of significant positive correlations between each pair of variables that are supposed to have a positive direct relationship in the conceptual model (p < 0.05).

Table 7: Cronbach’s Alpha, Means, Standard Deviations and Correlation Coefficients (N = 115)

Variable Items

α

1. User interface ₇ .90 _{5.39 1.10} 1.00 2. Content ₆ .86 _{5.63 1.07} .551a 1.00 3. Utilitarian value ₅ .92 _{5.56 1.07} .555a .567a 1.00 4. Hedonic value ₃ .83 _{5.03 1.37} .465a .306a .468a 1.00 5. Brand loyalty ₅ .91 _{4.93 1.42} .502a .542a .492a .612a 1.00 6. Product involvement 4 .93 4.65 1.30 .329a .305a .261a .365a .348a 1.00 7. Brand attachment 4 .80 3.05 1.61 .125 .089 .147 .383a .374a .301a 1.00

23

3.5 | Assumptions of multivariate analysis

Before multivariate data analysis could be performed, several assumptions of multivariate analysis should be tested (Hair et al., 2006; Aiken & West, 1991).

1. Acceptable degree of multicollinearity

Less than complete multicollinearity (e.g. less than perfect correlation between the predictors) is desirable. The Variance Inflation Factor (VIF) is a widely used measure of the degree of multicollinearity between an independent variable

with the other variables in a regression model. By measuring the VIF it is assured that no variable in the model is measuring the same relationship as is measured by another variable. A rule of thumb is that if the VIF exceeds a value of ten, multicollinearity is regarded too high (Hair et al., 2006). Since neither of the independent variables has a variance inflation factor greater than ten (all VIFs are below a value of five, table 8), there are no apparent multicollinearity problems.

2. Multivariate normal distribution

Regression assumes that the error terms of the variables have normal distributions. Non-normally distributed variables can distort relationships and significance tests. In order to test whether the two error terms of the samples are normally distributed, first two nonparametric tests are executed: the Mann-Whitney U (MWU) test

and the Kolmogorov-Smirnov (KS) test. The MWU test compares the difference in location of two populations, and the KS test assures the similarity of both distributions, taken into account the differences in median, dispersion and skewness. Both tests indicate that the variables ‘user interface’, ‘content’ and ‘product involvement’ are not normally distributed (p > 0.05). Moreover, the KS test also denotes ‘brand loyalty’ as not being normally distributed (p = 0.158).

Table 8: Variance inflation factor

Variable VIF (N=115) VIF (N=107) User interface _2.073 1.792 Content _2.324 1.610 Utilitarian value _2.713 1.761 Hedonic value _1.735 1.708 Product involvement 1.320 1.228 Brand attachment 1.272 1.427

Dependent: brand loyalty

Table 9: MWU and KS test

24

3. Deletion of outliers

More specific information about the distribution of the samples can be withdrawn from the skewness and kurtosis values. Skewness assesses its symmetry about the mean of a distribution, whereas kurtosis refers to the relative peakedness or flatness of the curve defined by the frequency distribution (Malhotra, 2007). Interpretation of the Z-scores of the skewness and kurtosis values lead to better understanding of distribution of the sample. Analysis of the Z-scores in table 10 indicate non-normal distribution, since several variables drop outside the acceptance range of -1.96 to 1.96 (normal distribution equals 0). By identifying outliers, deleting them and test the effects of deletion on the skewness and kurtosis values of the total dataset (trial-and-error), it is attempted to improve the Z-scores of the sample set. Hence, it must be assured that all variables are measured reliably, without error or outliers.

Kurtosis. Two variables which failed to meet both nonparametric tests, ‘user interface’ and

‘content’, have high positive Z-scores of between 6.60 and 9.27 (table 10), meaning a higher peakedness than normal distribution. After identification and the removal of eight outliers, Z-scores are well within acceptable limits (table 11). The high negative Z-Z-scores of ‘brand attachment’ can be explained by the fact that this variable has a lot of minimal scores on the 7 points scale, meaning many respondents are not attached to a mobile operating system or derive social value from.

Table 10: Initial skewness and kurtosis values (N = 115)

Variable Mean SD Skewness

25

Skewness. Initially, the Zscores on skewness farly exceeded limits by negative scores as low as

-7.39. Negative skew values indicate that the left tail of the distribution is heavier than the right one. The removal of items also dramatically improved Z-scores of skew values (table 11). For example, the highest Z-score on ‘user interface’ (-7.39) transformed to a more acceptable level of -2.29, although several items still exceed the requirement by a minimum. In addition, deletion of more items did not have a significant positive effect on skew values, and histograms indicate a reasonable normal distribution. In conclusion, the deletion of eight outliers improved Z-scores dramatically and reduced the standard deviation, and a reasonable normal distribution can be assumed.

Table 11: Revised skewness and kurtosis values (N = 107)

Variable Mean SD Skewness

26

3.6 | Econometric model

In order to measure the relationships in the conceptual model, multiple regression analysis is used for all equations. Multiple regression analysis is a statistical technique to determine the mathematical influence of two predictor variables on the criterion variables. The additional moderated multiple regression model is formed by including a moderator Z and its products with the two predictors X1 and X2, the so-called interaction terms. The regression analysis is

split into three models in accordance with the 18 hypotheses, which can be classified by three different criterion variables: utilitarian value, hedonic value and brand loyalty. All variables are appropriate for regression analysis. The nominal variable brand usage is dummy coded with Apple being controlled for in order to be convenient for regression. All other variables did not need further preparation for regression.

In the first regression model with brand loyalty as criterion (Y) is defined by the following equation, whereas the predictors β 1 and β 2 are respectively utilitarian and hedonic value and

ɛ represents an error term.

(1) BL = α + β 1 UV + β 2 HV + ɛ

Whereas BL = brand loyalty, UV = utilitarian value and HV = hedonic value.

The second regression model with utilitarian value as criterion is defined by the following four equations, whereas the β 1 (user interface)and β 2 (content) are the predictors X1 and X2, β 3

(product involvement, brand attachment or brand usage) equals the moderator Z, β 4 and β 5

are the interaction terms, and ɛ represents an error term. (2) UV = α + β 1 UI + β 2 C + ɛ

(3) UV = α + β 1 UI + β 2 C + β 3 PI + β 4 PI*UI + β 5 PI*C + ɛ

(4) UV = α + β 1 UI + β 2 C + β 3 BA + β 4 BA*UI + β5 BA*C + ɛ

(5) UV = α + β 1 UI + β 2 C + β 3 BU + β 4 BU*UI + β 5 BU*C + ɛ

(6) UV = α + β 1 UI + β 2 C + β 3 PI + β 4 PI*UI + β 5 PI*C + β 6 BA + β7 BA*UI +

β8 BA*C + β 9 BU + β 10 BU*UI + β11 BU*C + ɛ

27 First, regression is conducted with only the two predictor variables in order to measure the direct effects on the criterion variable without distortion due to presence of other variables in the model. In the next three equations, the moderators are included one by one to estimate the separate effects of the moderators on the relationships between the predictor and criterion variables. At last, all moderators are included to determine the influence of the predictors relatively to each other.

The equations in the third regression model are similar to those in the second regression model. Except in this model the criterion variable Y is hedonic value.

(7) HV = α + β 1 UI + β 2 C + ɛ

(8) HV = α + β 1 UI + β 2 C + β 3 PI + β 4 PI*UI + β5 PI*C + ɛ

(9) HV = α + β 1 UI + β 2 C + β 3 BA + β 4 BA*UI + β 5 BA*C + ɛ

(10) HV = α + β 1 UI + β 2 C + β 3 BU + β 4 BU*UI + β5 BU*C + ɛ

(11) HV = α + β 1 UI + β 2 C + β 3 PI + β 4 PI*UI + β5 PI*C + β 6 BA + β 7 BA*UI +

β8 BA*C + β 9 BU + β 10 BU*UI + β11 BU*C + ɛ

28

4

|

R

4.1 | Regression results for brand loyalty

In the first model we hypothesize that perceived utilitarian value and hedonic value are antecedents of brand loyalty. Table 12 displays the regression results of the first two hypotheses regarding direct positive effects of utilitarian and hedonic value on brand loyalty. Addition of the antecedents lead to a good fit of the model with F (2) = 26.11, p < 0.01. The variance in brand loyalty can be explained for 32.1 percent by the antecedents. H1 proposes a direct positive effect of utilitarian value on brand loyalty. This hypothesis is supported (p < 0.01), implying that utilitarian value increases the brand loyalty. Moreover, hypothesis H2 asserting a direct positive effect of hedonic value on brand loyalty is also accepted (p < 0.01). In this respect, utilitarian value ( β = .336) has a stronger effect on brand loyalty than hedonic value ( β = .289).

Table 12: Regression results for brand loyalty

Hypothesis Model 1

Variables Beta

Utilitarian value (UV) 1 .388a

Hedonic value (HV) 2 .271a

R2 .334

Adjusted R2 .321

R2 change

F-value 26.11a

29

4.2 | Regression results for utilitarian value

Main results

Table 13 displays the regression results for hypotheses 3 and 4 regarding the direct relationships between dependent variable utilitarian value and its antecedents, user interface and content. The inclusion of both independent variables notions a good fit of the model with F (2) = 32.47, p < 0.01. Of the total variance of utilitarian value, 37.3 percent can be explained by the independent variables (adjusted R2 = .373), which indicates a moderate strength of association between X1,2 and Y.

H3 suggests a positive significant relationship between user interface and utilitarian value. This hypothesis is accepted (p < .01). There is also support for H4 (p < .01), affirming that content increases the perceived utilitarian value. The effect of user interface on utilitarian value ( β = .355) is about equal in strength compared to the effect of content on utilitarian value ( β = .354).

Moderating results

Table 13 also reproduces three subsequent moderating models to test the moderating effects of the moderators product involvement, brand attachment and brand usage. In each model, the moderator and two interaction terms of the moderator and predictors are included in the first regression model. Model 3 encloses the moderator product involvement and two interaction effects between product involvement and the two predictors. The increase of R2 _in this model is not significant (R2 _{change of 0.12 signified p > 0.10). Furthermore, the interaction} terms PI*UI and PI*C were not significant. In conclusion, H7 and H8 are not supported, in which we hypothesize that the effect of user interface and content on utilitarian value is stronger when consumers are more involved with the product class.

In model 4 we hypothesize that when consumers are more attached to a brand, this will result in a stronger effect of user interface and content on utilitarian value. However, adding of the moderator and its interaction terms did not result in a significant change in R2 (0.21, p > 0.10). The beta coefficients BA*UI ( β = .077) and BA*U ( β = -.019) were also not significant. Thereby, the hypotheses H9 and H10 are rejected.

30 supported.

In model 6 the total model is estimated. The predictor variables are again significant, and noteworthy the interaction BU*UI is also significant ( β = .254). Furthermore, the moderator brand usage has a strong negative direct effect on utilitarian value ( β = -.360).

Table 13: Regression results for utilitarian value

Hypothesis Model 2 Model 3 Model 4 Model 5 Model 6

Variables Beta Beta Beta Beta Beta

User interface (UI) 3 .355a .361a .342 a .189c .205b Content (C) 4 .354a .354a .337a .421a .406a Moderators Product involvement (PI) .046 -.022 Brand attachment (BA) .127 .064

Brand usage (BU) -.346a -.360a

Interaction effects PI * UI 7 -.032 -.021 PI * C 8 .113 .08 BA * UI 9 .077 .101 BA * C 10 -.019 -.088 BU * UI 11 .157 .254b BU * C 12 -.101 -.128 R2 .384 .396 .406 .492 .524 Adjusted R2 .373 .366 .376 .467 .469 R2 change _.012 .021 .107a .140 a F-value 32.47a 13.25a 13.78a 19.53a 9.52a

31

4.3 | Regression results for hedonic value

Main results

Table 14 shows the results of the regression between hedonic value and its antecedents. Hypotheses H5 and H6 hypothesize a positive effect of on the one hand user interface on hedonic value, and on the other hand content on hedonic value. Model 7, with accompanying variables user interface and content, produces a good fit (F (2) 27.03 = p < 0.01). The both variables account for 23.1 percent of the variance in hedonic value (adjusted R2 = .231). H5 states that user interface has a positive effect on hedonic value and this hypothesis is supported (p < 0.01). Thus, most of the variance in hedonic value is accounted for by user interface ( β = .477). On the opposite, the regression results indicate that content has small till no influence on hedonic value ( β = .033, p > 0.10). As a result, H6 postulating content has a significant effect on hedonic value must be rejected. This fact also indicates that the moderators hardly can have any influence on this relationship, since it is literally almost nonexistent.

Moderating effects

In the moderated regression model for hedonic value, the same procedure for analysis is used as in the utilitarian value model. Moreover, the moderators are the same as in the former model. Model 8 includes the moderator product involvement and two interaction effects between product involvement and the two predictors. The change in R2 _{of .063 is significant (p} < .05). However, the effects of the interaction terms PI*UI ( β = .022) and PI*C ( β = .119) are quite small and not significant (p > .10). Thereby, the hypotheses H13 and H14 implying a moderating effect of product involvement are rejected. Moreover, the increase in R2 can be explained by a side effect, namely the direct positive effect of the moderator on hedonic value (

β = .218, p < .05).

The regression model with brand attachment as a moderator resulted in similar findings as in model 8. Again, the inclusion of the moderator and its interaction terms led to a significant large increase in R2 of .131 (p < 0.01). This is mainly due to the large direct effect of the moderator on hedonic value ( β = .383, p < .01). Hypotheses H15 and H16, suggesting a moderating effect of brand attachment on the relationships between user interface/content on hedonic value, are rejected (p > 0.10).

32 In model 11 the total model is estimated. This did not result in notable differences in outcome in comparison to the measured effects in the previous models. No moderating effects are found in this model.

Table 14: Regression results for hedonic value

Hypothesis Model 7 Model 8 Model 9 Model 10 Model 11

Variables Beta Beta Beta Beta Beta

User interface (UI) 5 .477a .458a .386a .363a .328a

Content (C) 6 .033 .010 .001 .071 .025

Moderators

Product involvement (PI) .218b .075

Brand attachment (BA) .383a .306a

Brand usage (BU) -.305a -.248a

Interaction effects PI * UI 13 .022 .038 PI * C 14 .119 .131 BA * UI 15 .045 .056 BA * C 16 -.072 -.101 BU * UI 17 .060 .057 BU * C 18 .063 .075 R2 .245 .308 .376 .335 .462 Adjusted R2 .231 .274 .345 .302 .400 R2 change .063b .131a .090a .217 F-value 16.89a 9.02a 12.18a 10.19a 7.14a

33

5

|

D

,

C

I

This study conceptualizes the quality-value-loyalty chain for mobile operating systems. In accordance with the quality-value-loyalty chain concept (Reichheld, 1996; Parasuraman & Grewal, 2000), it was expected that e-service quality enhances perceived value, which in turn contributes to brand loyalty. In addition, the quality-value-loyalty chain was elaborated into a more complex model. The perceived value construct was split into a utilitarian and hedonic value component, and it was proposed that the e-service quality of a mobile operating system is determined by multiple dimensions: user interface and content.

The first contribution of this study is the confirmation of the proposed two e-service quality dimensions of a mobile operating system (user interface and content) as antecedents of perceived utilitarian and hedonic value. Whereas for the quality of e-services various different constructs are mentioned in literature, no previous study succeeded in dividing the several e-service quality variables into two separate dimensions.

Although the proposed user interface construct was derived from factor analysis, not all sub-constructs did pass the check for factorability. The structure and lay-out variables were rejected, and user interface remained as a construct consisting of customization, ease of use and look and feel. An explanation for this can be that customers demand a logically organized mobile operating system with a clear structure, in the same way that customers demand security and reliability. Hence, structure and lay-out will serve as a prerequisite and does not operate on the augmented level. Perhaps this aspect is more adequately covered on the augmented level by the ease of use concept, in terms of logicality of required handlings and minimal number of handlings necessary. The e-service quality of content was affirmed as a composition of service integration, amount of content and the application store.

34

Antecedents of perceived utilitarian and hedonic value

In accordance with previous studies (e.g. Bauer et al., 2006), direct relationships were expected between the e-service quality dimensions and the perceived value constructs. The e-service quality of the user interface had a direct significant positive effect of similar strength on both perceived utilitarian and hedonic value. Hence, the user interface consists of both utilitarian and hedonic elements. The aesthetic and sensory elements of the look and feel component directly appeal to the hedonic value of the mobile operating system; the ease of use component in terms of the required actions relate more directly to the perceived utilitarian value.

Contrary to the user interface dimension, e-service quality of content had only a significant positive influence on the perceived utilitarian value; content had almost no effect on hedonic value. The positive effect of content on utilitarian value is more obvious because of its functional elements, but the data did not support the existence of a leverage of emotions on the perceived hedonic value. It was assumed that the incorrect functioning of software would create negative emotions, which would result in lower perceived hedonic value. Notably, this is not supported by data. One could argue that content in terms of applications with entertainment as primary objective would definitely enhance the perceived hedonic value. Games for smartphones are widely available and very popular among users; for example, Angry Birds has been downloaded more than 100 million times for both Apple iOS and Google Android. However, in the top 20 list of most installed applications (100 million or more) on Google Android only two applications have entertainment as its main purpose (i.e. Angry Birds and Fruit Ninja: https://play.google.com/store/apps). This indicates that the majority of users still mainly use their smartphone for utilitarian reasons instead of for hedonic fulfillment. Moreover, this provides another motive for the existence of a strong significant positive relationship between content and perceived utilitarian value.

Moderating effects

35 Furthermore, no moderating effects were measured for brand usage. One exception is brand usage as moderator in the total model for utilitarian value (model 6), in which the interaction with user interface is significant. The difference in effect of user interface contrary to models 2, 3 and 4 must be noted. The influence of user interface in those models is much stronger. This points at strong influence of brand usage in the total model, and its presence tends to decrease the influence of user interface and somehow the two variables interact with each other. On its turn, this can explain the significance of the interaction BU*UI.

However, brand usage had a significant direct negative effect on both utilitarian and hedonic value. Since the variables are dummy coded with Apple being controlled for, this means that usage of the Apple iOS will lead to a stronger positive effect compared to usage of Google Android on both utilitarian as well as on hedonic value. The stronger effect on hedonic value was expected because Apple iOS was associated with hedonic terms. However, the stronger positive influence of Apple iOS on utilitarian value is remarkable considering Google Android was characterized with utilitarian associations by the interviewees. An explanation for this can be that Google Android is also widely available for budget phones of inferior quality, which can affect the overall perception of the utilitarian value of the mobile operating system. Hence, Apple iOS is only available for the Apple iPhone, which is positioned and perceived as an exclusive premium product.

Antecedents of brand loyalty

36 On a strategic level, the results can be used by managers in charge of designing the service for improving and invest more heavily in the most important service attributes in order to increase the perceived value of the product. On a marketing level, this information can be translated into actionable marketing campaigns, whereas more emphasis and expenditures can be put on the most important e-service quality dimensions in order to gain brand loyalty.

37

6

|

L

R

F

R

This study has several limitations, which are mostly related to the sample. The main objective of a sample should be its capacity to reflect the entire target population. Considering the overall high education of respondents, in combination with an overall low income, it is assumed that a fair share of the sample consists of students. Therefore, the conclusions drawn from this sample might not be reflective for the entire target population and might have created a sample bias. Hence, it is disputable that a lower education or higher age will result in a significant change in outcome.

The sample size of 153 respondents can also be evaluated as not an adequate sample size with regard to the entire target population of in total approximately ten million people (smartphone users in the Netherlands). Additional research with a greater sample is needed to support the conclusions in this study. Furthermore, the data is limited to one country. Future research in an international context is needed in order to be able to draw conclusions on an international level.

Another limitation is the scope of the study. In this study, the focus is only on the mobile versions of the Google Android and Apple iOS operating system. Therefore, the results are only applicable to the mobile phone industry; further research on the quality-value-loyalty chain of the tablet industry is needed. The study can also be conducted in a broader context with the inclusion of the phone brand and the network operator. In that case, it would be interesting to know the role of price as a moderator between e-service quality and perceived value.

38

7

|

R

Literature:

Abels, E.G., White, M.D. and Hahn, K. 1999. ‘A user-based design process for Web sites’, OCLC

Systems and Services, Vol. 15 No. 1, p. 35-44.

Aiken, L.S. & West, S.G. 1991. Multiple Regression: Testing and Interpreting Interactions. Newbury Park: Sage.

Arnold, M.J. & Reynolds, K.E. 2003. ‘Hedonic shopping motivations’, Journal of Retailing, Vol. 79 No. 2, p. 77-95.

Bauer, H., Falk, T. & Hammerschmidt, M. 2006. ‘e-TransQual: A transaction process-based approach for capturing service quality in online shopping’, Journal of Business Research, Vol. 59, p. 866-876.

Batra, R. & Ahtola, O. 1990. ‘Measuring the hedonic and utilitarian sources of consumer attitude’, Marketing Letters, Vol. 2, p. 159-170.

Beatty, S.E. & Talpade, S. 1994. ‘Adolescent influence in family decision making: a replication with extension’, Journal of Consumer Research, Vol. 21, p. 332-341.

Berman, W.H. & Sperling, M.B. 1994. ‘The Structure and Function of Adult Attachment’ in

Attachment in Adults: Clinical and Developmental Perspectives, Sperling, M.B. & Berman, W.H.,

eds. New York: Guilford Press, p. 3–28.

Bolton, R.N. & Drew, J.H. 1991. ‘A multistage model of customers’ assessments of service quality and value’, Journal of Consumer Research, Vol. 17 No. 4, p. 375–384. Chaplin, N. & John, D.R. 2005. ‘The Development of Self-Brand Connections in Children and Adolescents’, Journal of Consumer Research, Vol. 32 No.1, p. 119–129. Cox, J. and Dale, B.G. 2001. ‘Service quality and e-commerce: An exploratory analysis’,

Managing Service Quality, Vol. 11 No. 2, p. 121-131.

Dabholkar, P.A. 1996. ‘Consumer evaluations of new technology-based self-service operations: an investigation of alternative models’, International Journal of Research in Marketing, Vol. 13 No. 1, p. 29-51.

Escalas, J. E. 2004. ‘Narrative Processing: Building Consumer Connections to Brands’, Journal of

39 Escalas, J.E. & Bettman, J.R. 2003. ‘You Are What They Eat: The Influence of Reference Groups on Consumers’Connections to Brands’, Journal of Consumer Psychology, Vol. 13 No. 3, p. 339– 348.

Fassnacht, M. and Koese, I. 2006. ‘Quality of electronic services: Conceptualizing and testing a hierarchical model’, Journal of Service Research, Vol. 9 No. 1, pp.19-31. Fogg, B.J. 2009. ‘A behavior model for persuasive design’, Persuasive, p. 40.

Field, A. 2000. ‘Discovering Statistics using SPSS for Windows’, London – Thousand Oaks – New Delhi: Sage publications.

Grönroos, C., Heinonen, F., Isoniemi, K. and Lindholm, M. 2000. ‘The Netoffer model: a case example from the virtual marketspace’, Management Decision, Vol. 38 No. 4, p.243-252.

Hair, J. F., B. Black, B. Babin, R. E. Anderson, and R. L. Tatham. 2006. Multivariate Data Analysis, 6th ed., Upper Saddle River, New Jersey: Pearson Education, Inc.

Heath, C. & Soll, J.B. 1996. ‘Mental budgeting and consumer decisions’, Journal of Consumer

Research, Vol. 23, p.40- 52.

Holbrook, M.B. & Hirschman, E.C. 1982. The experiental aspects of consumption: consumption fantasies, feelings and fun’, Journal of Consumer Research, Vol. 9 No. 2, p. 132-140.

Holbrook, M. B. 1994. ‘The nature of customer value: An axilogy of services in the consumption

experience’. Thousand Oaks, CA: Sage.

Kapferer, J.N. 2005. ‘The post-global brand’, Journal of Brand Management, Vol. 12 No. 5, p. 319-324.

Keller, K.L. (2008). Strategic brand management. Building, measuring, and managing brand

equity, 3rd edition. Upper Saddle River, NJ: Prentice Hall.

Malär, L., Krohmer, H., Hoyer, W.D. and Nyffenegger, B. 2011. ‘Emotional Brand Attachment and Brand Personality: The Relative Importance of the Actual and the Ideal Self’, Journal of

Marketing, Vol. 75, No. 4, p. 35-52

Malhotra, N.K. 2007. Marketing Research: An applied orientation, 5th edition, New York: Prentice Hall.

Meuter, M.L., Ostrom, A.L., Roundtree, R.I. and Bitner, M.J. 2000. ‘Self-service technologies: understanding customer satisfaction with technology-based service encounters’, Journal of

40 Mikulincer, M. & Shaver, R. 2007. Attachment in Adulthood: Structure, Dynamics and Change. New York: The Guilford Press.

Monroe, K.B. 1990. Pricing: Making Profitable Decisions, 5th edition, New York: McGraw-Hill. Parasuraman, A. & Grewal, D. 2000. ‘The Impact of Technology on the Quality-Value-Loyalty Chain: a research agenda’, Journal of the Academy of Marketing Science, Vol. 28 No. 1, p.168-174.

Park, C.W. & Young, S.M. 1986, ‘Consumer response to television commercials: the impact of involvement and background music on brand attitude formation’, Journal of Marketing

Research, Vol. 23, p. 11-24.

Punniyamoorthy, M. & Prasanna Mohan Raj,. M. 2007. ‘An empirical model for brand loyalty Measurement’, Journal of Targeting, Measurement and Analysis for Marketing, Vol. 15, p. 222– 233

Reichheld, F.F. 1996. ‘Learning from customer defections’. Harvard Business Review, Vol. 74, p. 56–69.

Van Riel, A.C.R., Liljander, V. and Jurriëns, P. 2001. ‘Exploring consumer evaluations of e-services: a portal site’, International Journal of Service Industry Management, Vol. 12 No. 4, p. 359-377.

Rietveld, T. & Van Hout, R. 1993. Statistical Techniques for the Study of Language and

Language Behaviour, Berlin – New York: Mouton de Gruyter.

Santos, J. 2003. ‘E-service quality: a model of virtual service quality dimensions’, Managing

Service Quality, Vol. 13 No. 3, pp. 233-46.

Sheth, J.N., Newman, B. and Gross, B. 1991. Consumption Values and Market Choices: Theory

and Applications, Cincinnati: South-Western Publishing Co.

Sohn, C. and Tadisina, S.K. 2008. ‘Development of e-service quality measure for the internet-based financial institutions’, Total Quality Management & Business Excellence, Vol.19 No.9, p.903-918.

Sweeney, J.C., and Soutar, G.N. 2001. ‘Consumer perceived value: The development of a multiple item scale’, Journal of Retailing. Vol. 77 No. 2, p. 203-220.

41 Thomson, M., MacInnis, D. J., & Park, C.W. 2005. ‘The ties that bind: Measuring the strength of consumers’ emotional attachments to brands’, Journal of Consumer Psychology, Vol. 15, p. 77– 91.

Wagner, J. 2011. ‘Anytime/anywhere: playing catch up with the minds of smartphone consumers’, International Journal of Mobile Marketing, Vol. 6 No. 1, p. 28-53.

Whan Park, C., MacInnis, D.J., Priester, J., Eisingerich, A.B. and Iacobucci, D. 2010. ‘Brand attachment and brand attitude strength: conceptual and empirical differentiation of two critical brand equity drivers’, Journal of Marketing, Vol. 74, p. 1-17.

Zeithaml, V.A. 1988. ‘Consumer Perceptions of Price, Quality and Value: A Means-End model and Synthesis of Evidence’, Journal of Marketing, Vol. 52, p. 2-22.

Zeithaml, V.A., Parasuraman, A. and Malhotra, A. 2000. ‘A Conceptual Framework for Understanding e-Service Quality: Implications for Future Research and Managerial Practice’, working report No. 00-115, Marketing Science Institute, Cambridge, MA.

Zeithaml, V.A., Berry, L.L. & Parasuraman, A. 1996. ‘The behavioral consequences of service quality’, Journal of Marketing, Vol. 60 No. 2, p. 31–46.

Websites:

Gartner, 2012 on worldwide market share analysis of mobile phones and operating systems:

http://www.gartner.com/resId=2334916

Telecompaper, 2012 on smartphone usage of Dutch consumers: