Bridging the sector gap in E-satisfaction with the development of a double-edged sword

(1)

Master Thesis

Florine Bottema - 10555838

MSc. in Business Administration – Digital Business University of Amsterdam – Faculty of Economics and Business

Supervisor: dhr.dr. J.Y. (Jonne) Guyt Date of submission: 22-06-2018

(2)

Statement of Originality

This document is written by student Florine Bottema, who declares to take full

responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is

(3)

Acknowledgement

First, I would like to thank my thesis supervisor J. Guyt, for his time, patience and

guidance throughout the process. He allowed this thesis to be my own work, and provided me with direction adjustments and constructive feedback whenever needed. This guidance was of critical value during the moments I did not know what direction to go. Furthermore I would like to P. Meinster and G. Sigmond for the time and support they have provided me

throughout my time as an thesis intern at KPMG at the Financial Business Services

department. At last, I would like to express my gratitude towards my friends and family for the love, care and support.

(4)

Abstract

While private service providers are trying to keep their heads up in today’s challenging e-service landscape, federal institutions are struggling to meet society’s divergent demand. Simultaneously, the distinction between the public and private domain is becoming increasingly blurred, as their disciplines are subsequently becoming more intertwined. Triggered by these environmental pressures, and guided by the inconclusive state of the extant literature on e-satisfaction in the public and private service domain, this study aspires to advance the understanding by capturing the complex concept of customer satisfaction in a cross-sector context. By adopting a user-centric approach, the e-services attributes of trust, reliability, usability and availability have been identified and integrated into a sector

universal measurement instrument. Confirmatory factor analysis (CFA), structural equation modeling (SEM) and a multiple group analysis are applied to test the hypothetical model across both sectors. The presumed relationships between trust, usability and availability revealed a positive direct significant relationship with customer satisfaction. Despite the nonsignificant moderating effect of sector, the analysis reveals trust and usability to be significantly different on sector level. 70% of the variance in customer satisfaction can be explained by the validated structural model of this research. In the journey towards the assessment customer e-satisfaction, extension of the proposed model should eventually lead to not only measuring latent needs with existing services but also identifying what it would take for institutions to fulfil these presumed needs should they not be met. Informed by the conclusive highlights, the demonstrated variable variance on the sector level forms fertile ground for future studies.

Keywords: Customer e-satisfaction, public sector, private sector, financial e-services,

(5)

List of Figures

Figure 1. All Hypotheses Represented in the Conceptual Model ... 22

Figure 2. Respondent Occupation ... 26

Figure 3. Respondent Web Access ... 26

Figure 4. Confirmatory Factor Analysis ... 30

Figure 5. Structural Equation Model – Final Conceptual Model ... 33

Figure 6. Distribution of Public Sector Organisations ... 65

Figure 7. Distribution of Private Sector Organisations ... 65

Figure 8. Purpose use of Insurance e-service provider... 66

Figure 9. Purpose use of Banks ... 66

Figure 10. Final Structural Equation Model (Including Covariance) ... 70

Figure 11. Unconstrained model: Public sector ... 71

Figure 12. Unconstrained model: Private Sector ... 72

Figure 13. Structural Weights model: Public sector ... 73

List of Tables

Table 1. Sample Characteristics ... Error! Bookmark not defined. Table 2. Construct Reliability ... 29

Table 3. Construct Validity ... 31

(8)

Table 5. Hypothesis Testing (Part A) ... 34

Table 6. Nested Model Comparisons ... 36

Table 7. χ2-Difference Test ... 37

(9)

1. Introduction

Over the last decade, we have experienced an incremental growth in the presence of Internet-based services in our daily lives. These digital technologies have not only

revolutionised the way humans interact but have also become a core aspect of the business environment of today’s society.

Until a few years ago, offering governmental services through digital channels was a relatively new idea, and hence a new practice in the public sector (Potnis, 2012). However, driven by the increasing pressure of this new information society, today’s public institutions are making progress on delivering readily accessible digital services with user-friendly interfaces (Accenture, 2017).

While implementing innovations, governments attempt to strike a balance between the divergent demands of fulfilling their citizens’ needs and meeting the previously set budgets and restrictions (Verdegem & Verleye). On top of that, organisations in this domain are also exposed to a high level of public criticism, as innovations in this discipline are likely to affect not only society as a whole, but also citizens individually (Donahue, 2005). In the meantime, private sector innovations continue to set even higher expectations (Accenture, 2012).

The private services industry has encountered the disruptive power of digital technology as well. In order for businesses to survive in this new service landscape, they must change with the times by redefining their core strengths and remaining focussed on their target group, not their internal organisation (D’Emilio, Dorton & Duncan, 2015).

Apart from the new opportunities, risks, responsibilities and restrictions the digital information technologies have brought, it has also changed the customers’ needs, as they are constantly in contact with technological interfaces. Due to the endless possibilities, the digital

(10)

era offers consumers in terms of information, alternatives and comparatives, maintaining customer loyalty is more difficult than ever.

Along with private sector actors’ growing involvement in the provision of

governmental services, it is increasingly difficult to determine the prefix of either public or private (Lindgren & Jansson, 2013). This can raise problems in assessing and identifying a particular service’s target group. Therefore, in order for today’s digital service providers to be able to identify and satisfy the needs of their customers, insight on the development of this satisfaction is required.

Although this might seem evident, past research has indicated that the current

literature lacks on empirical data with regard to the demand side of governmental e-services, and that the design and development of these services are too often guided by the supply side (Bertot & Jaeger, 2006; van Dijk et al., 2008). Moreover, there is an ongoing discussion concerning the relation with the construct of customer satisfaction and its contributing dimensions (Zavareh et al., 2012).

It follows from this that, in order to survive, it is of crucial importance for digital service providers to know their customers. By combining the knowledge of customers’ needs and organisational possibilities with regard to resources, organisations are able to smoothly identify the pain points with the best potential output. Taken the increasing

interconnectedness of both sectors in terms of e-services into consideration, the question raised concerns what customers really want to see in today’s digital services, and whether these preferences are universal. Further questions concern whether these dimensions are sector dependent and if measurement requires distinct exclusiveness. Following thereon, this study will be guided by the following research question:

How do the key determinants of customer satisfaction differ in the public and private sector for delivering financial e-services?

(11)

This study aims to contribute to the current literature by capturing the complex concept of customer satisfaction in the light of digital financial services. By establishing a more thorough understanding, the current state of the literature is examined, taking the moderating factor of sector into account.

By adopting a user-centric approach, four main e-service dimensions are identified and tested. Quantitative data is collected by the design of a survey, resulting in a sample of 569 cases. According to the e-service dimensions identified, a hypothetical model is developed and the model fit is validated with relation to the retrieved data. Eventually, two structural models are designed to test invariance across both sectors.

The remainder of this thesis is structured as follows. Section 2 provides an extensive review of the relevant literature, together with the proposed hypotheses. Subsequently, section 3 will combine the proposed hypotheses into a graphical conceptual model. Section 4 will elaborate on the research design and the methodological approach taken on in this research. The statistical analyses conducted will be presented in section 5, followed by section 6, where the results, implications and limitations of the study are discussed. Section 7 will conclude with a comprehensive synopsis of this thesis.

2. Literature Review

In this chapter, the state of current literature concerning customer (e-)satisfaction is assessed in order to develop a solid understanding of the central concept in this research within the digital context. Based on this, the section concludes with the main research question that guides this study. Second, the primary customer e-satisfaction dimensions and established measurement theories are discussed, along with their relevance to both public and private sector e-services. Based on the four main dimensions selected for this study, several

(12)

hypotheses are developed, which form the foundation for the conceptual representation of this study, as discussed in the next chapter.

2.1 Introductory Background

For quite some time, the concept of customer satisfaction has been a topic of great interest for both researchers and organisations. As the mean objective of each commercial organisation is to generate and maximise profits, customer satisfaction plays an important role. It is widely acknowledged that this construct significantly influences future purchase behaviour and could lead to positive word-of-mouth, and eventually, customer loyalty (Suh & Yi, 2006). Therefore, customers comprise a crucial asset for organisations, as they can

generate long-term business benefits (Zhang, Zhang & Law, 2014).

2.2 Customer Satisfaction

In 1983, the importance of this construct was acknowledged by La Barbera and Mazurksky, who identified customer satisfaction as one of the main drivers of repurchase intentions, with dissatisfaction as one of the main drivers for the discontinuation of future purchases. Customer satisfaction is now considered the baseline indicator of business performance, as it provides organisations with an effective and meaningful explanation of their customers’ preferences and expectations towards their product or service (Gerson, 1993).

According to Kotler (2000), the concept of customer satisfaction entails an individual’s feelings of either pleasure or disappointment resulting from the comparison between the set expectations and the outcome of a product or service. Hansemark and Albinsson (2004) defined this concept as a customer’s overall attitude towards a service provider, or more specifically, as an emotional reaction to the difference in what they anticipated and what they eventually receive.

(13)

Together with the rise of the Internet, the digital conceptualisation of customer

satisfaction has developed over time. Furthermore, since it is harder to keep online customers loyal, not surprisingly, the importance and relevance of this concept has increased as well (Zavareh, et al., 2012). Researchers Szymanski and Hise (2000) refer to this so-called ‘e-satisfaction’ as the total effect of all the experiences with a certain digital service provider over time (Anderson & Srinivason, 2003).

2.3 Customer E-Satisfaction

A customer’s state of satisfaction or dissatisfaction plays a crucial role in the resulting decision of whether to start, continue or stop using a specific service. However, this is not the only consequence. The access to the Internet has provided customers and users with more than just the opportunity to look for alternatives; it has also offered (dis)satisfied customers a stage on which to express themselves and give voice to their experiences. Customers can choose to complain to a third party, or even worse, engage in negative online word-of-mouth. They can share their opinion across a worldwide network with just one click, so one can imagine the impact and consequences such messages hold for a business and its target audience (DeMers, 2015).

Especially in today’s world, with digital technology’s increasing influence on the customer’s daily life and decision-making process, the online expressions of (dis)satisfied customers are receiving an increasing amount of attention. Thus, much research has been performed on the influences and drivers behind the concept and development of this e-satisfaction. However, the meaning of this multidimensional concept greatly depends on the type of service, the channel through which it is delivered and in what sector the service is provided.

(14)

2.4 Customer E-Satisfaction in the public and private domain

The concept of customer satisfaction has received a great deal of attention in both the public and private sector, resulting in a wide variety of concept definitions, hypothetical frameworks and measurement tools. For the concept in relation to the public sector, some researchers speak of ‘conceptual vagueness’ or even refer to the ‘ghetto-isation’ of the construct(Hood & Margetts, 2010; Yildiz, 2007).

Additionally, with the growing involvement of private sector actors in the provision of public services, it can sometimes be difficult to determine the prefix of either public or private (Lindgren & Jansson, 2013). For example, nowadays, an organisation can be public in terms of ownership and private with regard to its financing or delivery. Due to the possibility of combining types of ownership, the boundaries between the public and private sector are becoming more blurred and intertwined (Rainey & Bozeman, 2000). However, despite the similarities between the sectors and their growing interconnectedness, the fundamental assumption of the two sectors being different in their service delivery should not be denied (Allison, 2004). In the conceptual paper of Lindgren and Jansson (2013), three main differences were outlined.

First, the target group for the public sector is, in principle, the whole society—all citizens. Public institutions should serve each individual of the society, driven by public interest, and they should carry out the responsibility for serving public justice and the common good. Second, public institutions can sometimes be perceived as monopolies that possess an asymmetrical relationship with their users (Rothstein, 2010). This asymmetrical relationship arises because public institutions have certain responsibilities and obligations, whereas citizens do not have the choice to file their requests at another organisation. Accordingly, due to a lack of competition, these services do not take place in a free market environment, which can also affect the price (Le Grand & Barlett, 1993). Finally, the role of

(15)

the user is different as well. Citizens are more than just users of public services; they have certain rights with regard to these services, such as services relating to social security benefits.

This difference is not only visible at the direct service level, it is also significant in the area of technology innovations implementation. Public institutions are not only exposed to different kinds of pressure, such as interests, restrictions and demands, but motivating factors are also lower compared to the private sector. (Albury, 2005). On top of that, caution and uncertainty are enforced by the high level of public criticism, as public sector innovations are not only likely to affect the society as a whole, but also citizens individually (Donahue, 2005). These factors, in combination with the rules and restrictions to which public sector institutions are liable, often cause digital innovations to become quickly commonplace by the time they are implemented. This pattern continuously repeats itself, and meanwhile, private organisations continue to raise expectations even higher.

Despite some gains, many critics have claimed that the design and development of public services have primarily been guided by the supply-side possibilities rather than user needs (Bertot & Jaeger, 2006). Furthermore, according to van Dijk et al. (2008), the current state of literature lacks empirical data with regards to the demand side of governmental e-services. In addition, the existing literature is also witnessing an ongoing discussion

concerning the relation and underlying differences in customer satisfaction in digital services (Zavareh et al., 2012). Hence, an instrument for assessing customer satisfaction and its attributes concerning e-services across sectors has not yet been proposed. Therefore, the following research question has been developed to guide this study: ‘How do the key determinants of customer satisfaction differ in the public and private sector for delivering financial e-services?’

(16)

This assessment is of great value, as on the one hand, the public and private sector are becoming increasingly interconnected, while on the other hand, the gap in the current

literature concerning incremental technological developments, together with the increasing competition in the field of digitally provided services, demands a user-centered view

assessment of user and customer needs and preferences. However, the question remains how this satisfaction can best be measured and what dimensions are the main dimensions

identified in the literature.

2.5 Dimensions of Customer E-Satisfaction

Despite the large amount of research and numerous identified dimensions and models tested in the area of customer satisfaction, the literature on the digital version of this concept remains relatively sparse. Furthermore, the identified dimensions are not universal, and a large number of concepts are being used to describe the same construct of e-service.

Especially across sectors, it is difficult to determine whether some concepts relate to the same phenomenon since some scholars partly neglect definitions of the concepts they use

(Lindgren & Jansson, 2013). Furthermore, the elements contributing to satisfaction can vary considerably, depending on the specific industry to which it relates. For instance, food, physical environment and employee service comprise the fundamental attributes of customer satisfaction with restaurant service, while voice call quality and Internet service form

essential elements, relevant to a customer’s satisfaction with a telecom company (Khan, Akram, Shah & Khan, 2017). A clear overview of all these dimensions is provided in Appendix A.

Several measurement models, such as the well-known ‘SERVQUAL’ method developed by Parasuraman et al. in 1985, are analysed. However, the four main attributes to be measured (reliability, assurance, empathy and responsiveness) would not be appropriate in

(17)

the digital context. Additionally, the digital version of this method (e-SERVQUAL) does not offer a suitable alternative, as this method is mainly developed for e-services in the retail context (Ariff, Yun, Zakuan & Josuh, 2012).

The measurement model developed by Verdegem and Verleye (2009) is considered more suitable and provides a good foundation for this study, as the researchers adopted measurements that are applicable to both the private and public sectors. The attributes used by this research are: ‘infrastructure, availability, awareness, cost, technical aspects, customer friendliness, security/privacy, content and usability’.

After an extensive analysis of the existing literature, four frequently recurring attributes have been identified in both sectors, namely, trust/security/privacy, usability, reliability/access and awareness. The four selected themes offer flexibility in the sense that these can be applied to services in both sectors, which is crucial in this comparison-based study. Attributes such as customer friendliness, content, infrastructure, cost and technical aspects are very sensitive to channel and sector, and thus are not adopted as main focus attributes in this research.

2.5.1 Trust

Information security is becoming an increasingly salient topic nowadays, especially in the context of financial e-services. As organisations and users are constantly exchanging confidential information over the Internet, the risk of security attacks grows (Alawneh, Al-Refai & Bathia, 2013). A recent report by BNP Paribas and the Boston Consulting Group revealed that trust should be one of the main focusses for financial institutions in today’s digital environment (BCG & BNP Paribas, 2018).

Congruent to the literature, trust is defined as a combination between privacy and security, and as the users’ confidence towards an electronic service concerning risks or

(18)

doubts during the process (Papadomichelaki & Mentzas, 2009). Similar to these researchers, the concepts of privacy and trust are often combined in order to assess the concept of trust due to the direct relationship these two concepts hold. For instance, McLeo and Pippin (2009) incorporated these concepts by interpreting trust as an individual’s belief that their personal information remains strictly confidential and shall not be used for other purposes than entered into the system.

Building further on this, as trust and privacy might not always possess the same meaning, time has revealed that in many cases, these concepts seem to represent a similar idea, even in different fields or contexts (Ting et al., 2016). These same researchers have highlighted the positive relationship trust holds with customer e-satisfaction.

For the purpose of this study, the concepts of privacy and security are also

incorporated in the measurement of trust. Therefore, the following hypothesis is proposed: H1a = Trust positively affects customer satisfaction with digital financial services

Trust is more than just a pivotal asset in building long-term relationships between users and organisations and encouraging future purchase behaviour; researchers also stress the importance of this key dimension with regard to public services (Kim, Donald & Raghav, 2009). This concept not only entails the trust concerning personal information security; in a public service context, it also relates to the trust in the entire governmental system (Alawneh et al., 2013). As such, trust appears to be an incremental e-service element in the creation of customer satisfaction (Janda, Trocchia & Gwinner, 2002). From this, the following

hypothesis is developed:

H1b = Trust has a stronger positive effect on customer satisfaction in public sector e-services than on the satisfaction with private sector e-e-services.

(19)

2.5.2 Reliability

The second main dimension of customer satisfaction adopted in this study is

reliability. This is considered to be an essential element, as it refers to the performance and accuracy of the online service (Janda et al., 2002). Aspects such as information clarity, easy-to-follow procedures and progress status of requests relate to this construct (Barrera, Garcia and Moreno, 2014). This topic has been studied intensively in the e-service industry, and it has been empirically demonstrated that the element of reliability possesses a strong, direct, positive relationship with customer satisfaction in the field of online services (Barat & Spillan, 2014; Wolfinbarger & Gilly, 2003). With regard to the preceding literature, the following hypothesis is proposed;

H2a = Reliability positively affects customer satisfaction with digital financial services.

Accordingly, reliability has been found to be the one of the strongest predictors of loyalty intentions and customer satisfaction towards a digital service (Omar, Saadan & Seman, 2015). However, the strength of this relationship is primarily emphasised in research related to the public sector. Ibrahim et al. (2006) found that, especially with regard to

electronic banking services, customers place moderate importance on the reliability of the e-services they are offered. Following thereon, the following hypothesis is developed:

H2b = Reliability has a stronger positive effect on customer satisfaction in private sector e-services than on satisfaction with public sector e-services.

2.5.3 Usability

The third main attribute this study adopts as a measure is usability. This measurement stresses the importance of whether a service is easy to comprehend and utilises reliable and up-to-date information.

(20)

It follows from this that the concept of usability is an essential element in measuring the state of satisfaction. It has often been stated that usability goes hand in hand with

efficiency and accessibility (Verdegem & Verleye, 2009). Correspondingly, the International Standards Organisation (ISO) describes accessibility as the usability of a service, facilitated for people within the broadest range of capabilities. This implies that a service can be used by any type of user (Jaeger & Bowman, 2005). Based on previous studies, it can be presumed that the determinant of usability possesses a positive, direct relationship with the level of customer satisfaction (Casaló, Flavián & Guinalíu, 2008). Accordingly, a third hypothesis is generated:

H3a = Usability has a positive direct effect on customer satisfaction with digital financial services.

The usability dimension is particularly relevant in services provided in the public domain, as all the members of society ought to have equal access to and be able to use, perceive and understand these services, no matter their skills, knowledge or capabilities (Bertot & Jaeger, 2006). From this, the following hypothesis is proposed:

H3b = Usability has a stronger positive effect on customer satisfaction in public sector e-services than in private sector e-services.

2.5.4 Availability

The same line of reasoning is behind the dimension of availability. In private sector e-services, availability is an important aspect, as customers expect the service they pay for to be available anytime and anywhere they want. Thus, in banking services, availability is

considered one of the main determinants of (dis)satisfaction. Furthermore, system availability is also considered an important driver of customer satisfaction in the governmental context (Kumbar, 2012). Subsequently, a fourth hypothesis is generated:

(21)

H4a = Availability has a positive direct effect on customer satisfaction with digital financial services.

Similar to the dimension of usability, availability is particularly important in public e-services. For example, when one is dissatisfied with the availability of a private sector service, the user has the option to change providers in order to overcome this. However, the possibilities to do such a thing in the public sector are much more limited. On top of that, governmental institutions carry a responsibility to offer their e-services to all groups in the society, as it is their obligation to serve each citizen equally (Lindgren & Jansson, 2013). In congruence with the aforementioned, the last hypothesis is proposed:

H4b = Availability has a stronger positive effect on customer satisfaction in public sector e-services than in private sector e-services.

2.6 Conceptual Framework and hypotheses

In order to expand the literature regarding the dimensions of e-satisfaction in the public and private sector, four main dimensions have been identified as independent variables.

The first hypotheses (A) relate to the direct relationships between the independent variables, and customer satisfaction are assessed. Subsequently, the moderating effect of sector is included in the remaining hypotheses (B).

Due to the channel- and sector sensitiveness of the attributes of customer friendliness, content, infrastructure, cost and technical aspects, these are not adopted as main focus

attributes. However, these will be later on adopted as single-item measurement scales, and thus adopted as observed variables.

A summary of the proposed hypotheses can be found below:

(22)

H1b = Trust has a stronger positive effect on customer satisfaction in public sector e-services than on the satisfaction with private sector e-services.

H2a = Reliability positively affects customer satisfaction with digital financial e-services. H2b = Reliability has a stronger positive effect on customer satisfaction in private sector e-services than on satisfaction with public sector e-e-services.

H3a = Usability has a positive direct effect on customer satisfaction with digital financial e-services.

H3b = Usability has a stronger positive effect on customer satisfaction in public sector e-services than in private sector e-e-services.

H4a = Availability has a positive direct effect on customer satisfaction with digital financial e-services.

H4b = Availability has a stronger positive effect on customer satisfaction in public sector e-services than in private sector e-e-services.

The conceptualisation, as depicted in Figure 1 is based upon the hypotheses developed in the previous chapter.

(23)

3. Method

This section elaborates on the methodology taken on in this study, including the research approach, design, data sources, research strategy, data collection method and types of analyses conducted.

3.1 Research Strategy

The research strategy adopted in this study is the usage of an online survey. The sampling method used during the distribution of this survey is snowball sampling, also known as chain-referral sampling. This entails the acquisition of respondents through

referrals made by other people who share and possess certain characteristics, or belong to the same interest group, via the researcher (Biernacki & Waldorf, 1981).

3.2 Data Collection Method

In order to test the moderating effect of sector on the relationship between the attributes of customer satisfaction and its overall concept, the online survey has been designed with the ‘randomiser’ tool in Qualtrics. This tool enables to (equally) assign

respondents to one of the developed sets of questions. In this specific situation, two identical sets of questions have been created: one for the public sector and one for the private sector. The only difference lies in the first question, where the respondent is asked to select the two organisation(s) that he/she has been in contact with, over a period of the last 12 months (please refer to Appendix B).

This feature avoids respondents in making their own comparisons across-sector, instead of subjectively assessing the services of the one sector presented. Avoiding these inside comparisons has circumvented complications arising from comparisons made within individuals, such as the requirement of higher reliability (McHorney, Ware, Lu, &

(24)

questions concerning the control variables (age, gender, education level, occupation and web access).

The questions in the survey were based on e-services provided in the Dutch financial services industry. This specific industry was selected due to its great variety and a large number of service providers in both the public and the private sector. The latter is of critical value, as the number of governmental institutions, active in a similar industry, compared to the private sector, is relatively sparse. The financial e-services industry allowed inclusion of a number of similar governmental institutions such as the Dienst Uitvoerend Onderwijs (DUO), Mijnoverheid and Uitvoeringsinstituut Werknemersverzekeringen (UWV). With regard to the private institutions, organisations such as banks (ING, Rabobank, Knab bank, Van Lanschot) and insurance providers (Ditzo, Nationale Nederlanden) were included.

3.3 Restructuring the retrieved data

As stated above, the survey’s first question relates to the two organisations the respondent has been in contact with over the last year. Albeit, the respondents are not required to select two options; they could also choose to select either one or more

organisations. The multi-response option, applied in the first question, serves as a foundation for the rest of the questions, as the participant’s selection appeared below every statement.

On the one hand, this expedient option allowed to collect responses relating to more than one organisation per individual, and therefore retrieve a greater amount of data. On the other hand, it also significantly enlarged the dataset, as the raw dataset contained almost 600 cases and 1280 variables.

After cleaning and restructuring the dataset, a total number of 569 cases were obtained, which means that the average of organisations selected per respondent is 2.5.

(25)

3.4 The Sample

A total of 310 respondents started the survey, however, after removing those with a progress status of < 99%, 228 respondents remained. This indicates a relatively high response rate of 74.3%, compared to the average response rate of 52.7% for studies performed at the individual level (Baruch & Holtom, 2008). From the total of 569 cases, 58.3% (332) is related to the private sector and 41.7% (237) to public sector institutions.

With regard to the demographics, an overview of the sample characteristics is provided in Table 1 as presented on the next page. Furthermore, a graphical depiction of the respondents’ occupation and hours of Internet access per week are provided in Figure 2 and 3.

Table

Sample Characteristics

Item Characteristic n _% _Mean _SD

Gender .54 .49

Women 299 52.5%

Men 259 45.5%

Do not want to disclose 11 1.9%

Age 2.06 1.44 16 - 25 years old 282 49.6% 26 - 35 years old 164 28.8% 36 - 45 years old 12 2.1% 46 - 55 years old 32 5.6% 56 - 65 years old 68 12% 66+ 9 1.6%

Highest Education 2.60 .68

High school diploma 51 9%

University degree (bachelor) 247 43.4% University degree (master) 256 45%

Ph.D degree 10 1.8%

Occupation 3.83 1.99

(26)

Figure 2. Respondent Occupation

Figure 3. Respondent Web Access

According to these graphics, most respondents were either student or private sector employee. With regard to the web access, almost half of the respondents indicated to have more than 21 hours per week access to the Internet.

From the total cases of 237 in the public sector, 34.6% (82) of the organisations accounts for the ‘Belastingdienst’, 30.8% (73) for ‘DUO’ and 18.6% (44) for ‘Mijnoverheid’.

0,7% 46,9% 1,4% 1,9% 35,3% 6,3% 0,5% 3,9% 3%

Respondent Occupation

< 21h 16 - 20h 11 - 15h 6 - 10h 1 - 5h > 1h

Respondent webacces

(hours/week)

< 21h 16 - 20h 11 - 15h 6 - 10h 1 - 5h > 1h

(27)

Furthermore, 8.4% had selected the ‘CJIB’ and, the rest of the 7,5% can be assigned to the other public institutions. The graphical depiction of this distribution can be found in

Appendix C, Figure 6.

The database retrieved from the private sector consisted of 332 cases, from which 218 banks and 144 insurance companies. From the banks, most respondents selected the

following organisations: ING (16.6%), ABN AMRO (14.5%), Rabobank (13.3%). From the insurance providers, Zilverenkruis was used the most often (9.9%), together with CZ (4.2%), Achmea (2.6%) and OHRA (3.3%). A graphical depiction of the data related to the private sector is presented in Appendix C, Figure 7.

3.5 Statistical Process

Firstly, in order to ascertain the efficiency of the measurement model and reduce the number of items, a Confirmatory Factor Analysis (CFA) was performed using AMOS. This theory-driven analysis allows representation of the relationships between the observed variables which are all based upon a subset of smaller unobserved latent variables or factors, accompanied by a random error term (Fox, 2011). In this study, a CFA performance takes precedence over an Explanatory Factor Analysis (EFA) as it aims for (un)standardised solution outcomes and hypotheses testing (Albright & Park, 2009). Subsequent to the CFA, convergent and discriminant validity are tested.

CFA is often referred to as a critical element for the Structural Equation Modeling (SEM) that has been applied, based upon the model build in the CFA (Fox, 2011). Once the model is identified, SEM is used to estimate how well the hypothetical model fits the data used in this case.

The fit implies the extent to which the model accounts for the correlations between the variables in the given dataset (Verdegem & Verleye, 2009). Thence, a good model fit

(28)

indicates the absence of significant discrepancies between the proposed and observed correlations, with regards to the variables in the model. Accordingly, in case of significant discrepancy between the correlations proposed and those observed, a poor model fit is presumed (Hoe, 2008).

While testing the model fit, several widely accepted norms are applied. The normative fit index (NFI), comparative fit index (CFI) and Tucker-Lewis index (TLI), should exceed the value of .9 (Bentler, 1990). The root mean square error of approximation (RMSEA) measures the discrepancy between the observed and proposed covariance (Steiger, 1990). A value less than .08 indicates an acceptable fit, whereas a value below .05 indicates a good fit (Schreiber, Nora, Stage, Barlow & King, 2006). With regard to the sample (n = 596), the chi-square statistic (χ2) is noted, however not used in interpretation as it is almost certain to be rejected in models with large samples (Waterman et al., 2010).

Moreover, the value of CMIN/DF (χ2/df) is assessed in determining the goodness of the hypothetical model fit. Some studies state that the value should be close to 1 for correct models, and not exceed the value of 3 (Byrne, 2001). Contrary to this, other suggest using ratios between 2 and 5 for the fit indication (Marsh & Hovecar, 1985). For the purpose of this study, the χ2/df-ratio of 3 or less is assumed to be a reasonable model fit indicator (Hoe, 2008). Moreover, taking the sensitiveness of the χ2 to the sample size into account, it is recommended to look at the other values, as stated in the previous sections, in the assessment of the model fit.

4. Results

After the process of restructuring the data and reducing the number of cases based on missing data and progress status, a final dataset with 596 usable cases remained. The sample

(29)

was checked for normality; no issues were identified. To be able to compare the data of both sectors, a dummy was created for both public (=0) and private (=1).

4.1 Testing for Reliability

In order to assess the reliability of the assumed measurement constructs, the internal consistency of the measurements used is calculated. Firstly the single-item measurement scales were removed: Awareness, Infrastructure, Cost, Technical aspects, Customer Friendliness, Content and Perceived Quality. These variables will be later on added to the model as observed variables. After this, 30-7=23 items remained, spread over the five presumed constructs. Internal consistency is assessed by the calculation of the Cronbach’s alpha per measurement. Additionally, the method of Carmines and Zeller (1974) is applied to determine the reliability. According to these researchers, if at least 50% of the item-total- correlations are between 0.3 and 0.7, the scale is assumed reliable (Nunally, 1994). Apart from the single-item scales, no additional items were deleted by applying this rule. An extensive overview of the factors, factor loading, variance, mean and standard deviations can be found in Appendix E. The calculated Cronbach’s alpha per measurement are presented in Table 1

Construct Reliability

Construct Cronbach's alpha

Satisfaction (SAT) 0.876

Trust (TP) 0.899

Reliability (REL) 0.809

Usability (UE) 0.865

Availability (AV) 0.763

4.2 Confirmatory Factor Analysis

In order to confirm these scales, ascertain the efficiency of the measurement model and reduce the number of items, a Confirmatory Factor Analysis (CFA) was performed. After

(30)

running the model several times, both items UE3 and UE4 are deleted, due to low internal factor loading and high residual covariance. A graphical depiction of the final CFA model is presented in Figure 4.

Figure 4. Confirmatory Factor Analysis

4.3 Reliability and Validity

In this section, the validity of the constructs, as identified in the CFA, is calculated. A widely accepted threshold for the construct reliability is CR > .05 (Hair et al., 1998). All the constructs pass this threshold, and therefore its measurement consistency is confirmed (Sebjan & Tominc, 2014). With the computation of the average variance extracted (AVE),

(31)

the convergent validity per construct is assessed. All the values of AVE are above the

threshold of 0.5 and all values of CR > AVE. Hence, the convergent validity of the constructs used in the measurement model is confirmed.

The discriminant validity is partly confirmed, as all constructs passed the threshold of MSV < AVE, except for UE.In Appendix E, table E2, an overview of the factor correlations and construct validity is presented. This table shows that the internal construct correlation of UE (.743) is lower than the correlation between the constructs of UE and SAT (.771). In other words, this means that the construct of UE looks too much like SAT.

After running the CFA model several times, different items are removed attempting to decrease the correlation between the two constructs. However, despite efforts, apart from removing UE3 and UE4, the correlation could not be brought down to a lower level. As all the other constructs do pass the threshold, it can be concluded that discriminant validity is partly confirmed (Fornell & Locker, 1981).

Table 2 Construct Validity Construct CR AVE MSV Results of convergent validity CR > AVE AVE > .5 Results of discriminant validity MSV < AVE UE .831 .552 .594 Yes No

SAT .879 .601 .594 Yes Yes

TP .902 .569 .251 Yes Yes

AV .774 .635 .494 Yes Yes

REL .819 .602 .514 Yes Yes

After the assessment of the model’s validity, the goodness of the model fit is assessed. A detailed summary of the final CFA model can be found in Table 4. The thresholds, as elaborated on in the previous chapter, are applied in evaluating the model fit.

(32)

Table 3

Model Fit Summary of CFA

Model Threshold

value Results

Degrees of freedom DF 179

Chi-square min value (χ2) CMIN 661.56

Chi-Square min value / Degrees of freedom CMIN / DF 3.696 < 3 Moderate

Probability level P 0 p < .001 Significant

Normative Fit Index NFI .902 >.9 Yes

Relative Fit Index RFI .873 >.9 Moderate

Incremental Fit Index IFI .927 >.9 Yes

Tucker-Lewis Index TLI .904 >.9 Yes

Comparative Fit Index CFI .926 >.9 Yes

Root mean square error of approximation RMSEA .069 <.05 to .08 Yes

As stated in the model summary, (χ2/df) > 3. However, since the χ2 is sensitive to sample size, an acceptable fit is assumed. The only index lower than its threshold is the RFI, though, the value is close to its threshold of .09. Furthermore, the RMSEA is an important indicator of the model fit, which has an acceptable value of .069. Overall, it can be concluded that this model, as developed in de CFA, indicates a good, parsimonious fit. Meaning that the model has good explanatory, predictive power and that the factorial structure fits the data properly (Sebjan & Tominc, 2014). Thus, the analytical process can proceed to the next step: Structural Equation Modeling.

4.4 Structural Equation Modeling

Subsequent to the CFA, Structural Equation Modeling is used to finalise the mode and test the proposed hypotheses. This method allows analysis of the paths amongst the latent variables in the model, to identify dependency relationships in the multivariate dataset (McDonald & Ho, 2002).

The observed variables; awareness, infrastructure, cost, perceived quality, technical aspects, customer friendliness and content, were added to the CFA model . Furthermore, in

(33)

order to include the moderating effect of sector, separate variables (independent variable x sector variable) were created and added to the model as well. A graphical depiction of the drawn model can be found in the figure below.

Figure 5. Structural Equation Model – Final Conceptual Model

The overall fit measures indicated that the model was acceptable: χ2 = 1472.145, df = 483, χ2/df = 3.048 (p=.000). NFI = .895, RFI = .822, IFI = .927, TLI = .873, CFI = .925. Finally, the RMSEA of this model is .06. Despite the fact that not all the indices in pass the threshold of >.9, the values are close; respectively NFI = .895, RFI, .822 and TLI .873, the χ2/df and RMSEA indicate a better fit than the preliminary CFA model.

Conclusively, 70% of the variance in customer satisfaction is explained by the model as created in the SEM. A table with these squared multiple correlations can be found in

(34)

Appendix E, Table E4. The structural model, including the covariances drawn can be found in Appendix F, Figure 10.

4.5 Hypotheses Testing A

With the list of the regression weights, retrieved from the output of the SEM, the first hypotheses are tested. Firstly, the significance is assessed, and in case of significance (at either the 95% or 99% confidence interval) the direction and the strength of the direct relationship is analysed. An overview of the results can be found below, in Table 5. Table 4

Hypothesis Testing (Part A)

B SE B β p Result

H1a Trust positively affects the customer _{satisfaction towards digital services.} .084 .042 .082* .048 Supported H2a Reliability of digital services positively _{affects the customer satisfaction.} .116 .122 .077 .399 Not

supported H3a Usability positively affects the customer

satisfaction of digital services. .785 .162 .633** <.001 Supported H4a Availability positively affects the _{customer satisfaction of digital services.} .364 .098 .331** <.001 Supported

* p < .05 **p < .01

Accordingly, the significant relationship between trust and customer satisfaction was confirmed. However, the strength of this positive relationship is very weak (.082 <.2). Contrary to the first hypothesis, the direct relationship between reliability and customer satisfaction was insignificant, and thus not supported. Furthermore, the third hypothesis on usability is supported, indicating a strong positive relationship (β>.5). This means that the latent variable of usability is a strong indicator of customer satisfaction: for every standard deviation unit increase in usability, customer satisfaction increases by .633. The last

(35)

hypothesis, relating to availability, is supported as well. The results confirm a moderate positive relationship (β>0.3).

Furthermore, the regression weights of the variables combining latent variables with sector, and the observed variables can be found in Appendix E, Table E3. From these results the conclusion can be drawn that the moderating effect of sector is not significant (p>.05) (p=.243). This means that customer satisfaction is not significantly different for both sectors.

With regard to the observed variables, the results indicate a few significant

relationships. Availability is presumed to be a stronger predictor for customer satisfaction in the private sector (β = .543, compared to β =.364), whereas the significant relationship between cost and customer satisfaction indicates a weak, negative relationship (p<.05)

(p=.037, β=-.098, β>.09). Secondly, the significant relationship of cost demonstrates that cost is a stronger predictor of customer (dis)satisfaction in the private sector (β = -.127). The significant negative relationship for content indicates that content is a stronger predictor for (dis)satisfaction in the private sector (p<.01, p=.01, β=-.110). Moreover, the variable of technical aspects is only a (weak) predictor of customer satisfaction in the private sector (p<.01, p=.01, β=.140).

Overall, it can be concluded that 70% of the variance in customer satisfaction can be explained according to the proposed model. The squared multiple correlations of the

structural model can be found in Appendix E, table E4.

4.6 Multiple group Analysis

As the SEM model has been built upon data retrieved from both the private and public sector, a multiple group analysis is used to test the equality of the factor loadings for both sector groups. Firstly, a chi-square difference test is performed. A nonsignificant p-value

(36)

statistic, at the 95% confidence interval (p>.05), indicates invariance at the group level, whereas model measurement variance is assumed in case of significance.

In order to compare the model for multiple groups, the interaction variables are removed, as well as the covariances drawn with the endogenous variables. After that, two groups are created in Amos, using the dummy made in the SPSS data with Public = 0 (n = 237) and Private = 1 (n = 332). With the tool ‘Multiple-Group Analysis’ in Amos, two models are created: one with unconstrained estimates, and the other with constrained estimates (including measurement weights, measurement intercepts and structural weights). By treating each path equally (relationship/influence), and running both models, the model fit summary indicates whether the models are different at group level. If the chi-square

difference test indicates significance at the 99% interval level (p<.001), that the model is significantly different at the group level. The model fit summary can be found in Table 6 as depicted below.

Table 5

Nested Model Comparisons (Assuming model Unconstrained to be correct)

Model DF CMIN P NFI IFI RFI TLI

Model 1 - Structural

weights 48 196.098 0 .019 .021 -.008 -.008

The next step is the assessment of each path individually. This is done by constraining the same path in both models individually, while all the other paths are estimated freely. After this, the model fit estimates are analysed and, the p-value is assessed. An overview of the χ2-difference tests per path is presented in Table 7.

(37)

Table 6

χ2-difference test

Latent variables DF CMIN P NFI IFI RFI TLI

Trust  SAT 1 4.505 .034 .000 .000 .000 .000 Reliability  SAT 1 1.086 .297 .000 .000 -.001 -.001 Usability  SAT 1 0.025 .874 .000 .000 -.001 -.001 Availability  SAT 1 18.677 .000 .002 .002 .002 .002 Observed Variables Awareness  SAT 1 349.691 .000 .034 .037 .040 .044 Infrastructure  SAT 1 241.534 .000 .024 .025 .028 .030 Cost  SAT 1 195.238 .000 .019 .020 .022 .024

Perceived Qual  SAT 1 39.401 .000 .004 .004 .004 .004

Tech aspects  SAT 1 448.320 .000 .044 .047 .052 .056

Customer Fr  SAT 1 588.099 .000 .058 .062 .068 .074

Content  SAT 1 563.000 .000 .055 .059 .065 .071

The results indicate significance for both trust and availability, at the 95% (trust) and 99% (availability) confidence interval. Furthermore, the table indicates that all the paths of the observed variables are significant at the 99% confidence interval (p<.001). However, it is not clear how these relationships are different among the group models. Therefore, the factor loading of each path will be compared in both structural weight models. The results of this model comparison are being depicted in Table 8 below.

Table 7

Path factor loading Public Private

Latent Variables Trust .07 .09 Availability .20 .15 Observed Variables Awareness .03 .02 Infrastructure .09 .10 Cost -.05 -.04 Perceived Quality -.04 -.04 Technical Aspects -.03 -.03 Customer friendliness .07 .06 Content .07 .07

(38)

4.7 Hypothesis testing B

According to Table 8 the latent variables reliability and usability are not significantly different at the model level (p>.05). Therefore, both hypothesis H2b and H3b are rejected.

In contradiction, the observed variables trust and availability indicate significance. The relationship of trust on customer satisfaction is significant at the 95% confidence level (p<.05, p=.034), indicating that the relationship of trust on satisfaction is significantly

different per sector. The loading values of the constrained paths in both models demonstrate a stronger positive effect in the private sector (.09), compared to the public sector (.07). Thus, hypothesis H1b is supported.

Furthermore, availability is significant at the 99% confidence interval (p<.001). According to the path estimates, the positive effect of availability is stronger in the public sector (.20) than in the private sector (.15). Therefore, hypothesis H4b is supported as well.

According to the results of the performed multiple group analysis, all the relationships between the observed variables and customer satisfaction indicate significance at the group level (p<.001). Further analysis reveals sector differences in five of the seven observed variables. The positive effect of awareness was stronger in the public sector (.03) than in the private sector (.02), as well as the effect of customer friendliness in the public sector (.07) compared to the private sector (.06). However, the effect of infrastructure on customer satisfaction is higher in the private sector (.10), in comparison with the public sector (.09).

5. Discussion

The primary goal of this study was to contribute to the literature by researching the relationship between the attributes of trust, reliability, usability and availability on customer satisfaction in the industry of financial e-services, with the moderating effect of sector. This

(39)

section elaborates on the main findings, hypotheses tested and the conclusions that can be drawn, followed by the limitations and potential implications of this study.

First and foremost, the main findings of this research suggest a hypothetical model that combines all the aforementioned relationships into a validated structural model, which can be held accountable for 70% of the variance in customer satisfaction. In line with the theoretical background, this study has demonstrated direct positive relationships for the latent variables of trust, usability and availability. From these three components, usability was determined to be the strongest predictor, followed by availability with moderate power and trust with a weak positive relationship. This positive, albeit weak, relationship is in agreement with empirical evidence found in public-related studies (Alaweneh et al., 2013).

By contrast, reliability was found to be an insignificant predictor of customer satisfaction. The cause of this discrepancy might not lie in the fact that the measurement items used for this variable were mainly derived from public-sector-related e-service studies. If that had been the case, the variable should have demonstrated a significant relationship in the public sector model as a result of the multiple group analysis. Since no such relationship was found, a more likely explanation could be that the utilised measurement items potentially relate more to the concept of service quality, and therefore serve a mediating role in e-service satisfaction. This prudent inference hinges on the literature anent reliability in relation to the e-service quality domain (Zavareh et al., 2012).

The expected overall moderating effect of sector on customer satisfaction was nonsignificant, implying that customer satisfaction was not significantly different per sector. Since there were no similar or comparable studies to be found that tested the moderating effect of sector on customer satisfaction attributes in the same way and in such a context, the results might provide a valuable contribution to the literature of customer satisfaction in both the public and private e-services spheres. Due to the disciplines of the two sectors becoming

(40)

increasingly intertwined, the identification of the significant attributes, confirmed on both sector levels, can provide guidance in further determining the attributes that are essential to e-services in this context.

Still, these increasing similarities do not take away from the fact that some significant dissimilarities have been revealed by the performance of the multiple group analysis. During the analysis, each path was treated equally, which enabled identification of (in)variance at the model level. Support was found for two of the four hypotheses. The significant effect of trust (H1) was confirmed to exert a stronger positive effect on customer satisfaction in public e-sector services compared to private e-sector e-services. This result is in line with the theoretical background, supposing that factors such as privacy concerns, fear of security attacks and theft of personal information appear to possess a greater influence on satisfaction in the public sector than in the private sector, and thus, causes this variance (Alawneh et al., 2013).

Likewise, the stronger positive effect of availability (H2) in the public sector compared to the private sector was confirmed as well. In accordance with previous work concerning e-service availability in these sectors, this difference could be explained by the fact that users of e-services are much more limited than users in the private sector. Too often, there are few to no alternatives available in the public sector, creating a situation where users are forced to cope with the way the service is delivered, one way or another (Lindgren & Jansson, 2013).

Aside from the confirmed hypotheses relating to trust and availability, no empirical evidence was found to support the remaining hypotheses concerning reliability and usability. The lack of support for the reliability dimension (H4) might not be surprising, considering the nonsignificant direct relationship found between reliability and customer satisfaction.

(41)

The results revealed a disconfirmation of the expected stronger effect usability (H3) was presumed to have on customer satisfaction in the public sector compared to the private sector. An explanation for not finding this anticipated effect may be found in the user-centric view adopted in this research. Although federal institutions ought to focus on delivering equallyaccessible and comprehensive services that are usable for each individual, this does not automatically mean that it is also a vital service element from a user point of view. Additionally, it must also be taken into consideration that some believe in a so-called ‘trade-off’ between the levels of security and usability. In line with this theory, some assume that, in order to create the desired level of security, one must yield regarding the level of usability (Cranor & Garfinkel, 2004).

Building further on the disconfirmed dimension of usability; discriminant validity could not be claimed due to the items of usability, which correlated more strongly with the items of satisfaction than with its own set of indicator variables (Shui, Pervan, Bove & Beatty, 2011; Zait & Bertea, 2011). These attributes were supposed to be unrelated, but the results of the divergent validity test revealed an overlap between the two scales, indicating that these share variance, and thus, are likely to measure the same thing (Sebjan & Tomic, 2014). As a result, the partial acceptance of discriminant validity in the current discourse might serve as a suggestive starting point for future research.

With regard to the sector models developed, some noteworthy relationships amongst the observed variables were found, as each single-item measurement indicated significant difference at the sector level. Whereas the level of infrastructure possesses a slightly higher positive influence in the private sector than in the public, the single-item measurement of perceived quality and technical aspects were invariant, meaning that these were not different at the sector level. Hence, the latent variables of awareness, customer friendliness and content

(42)

were revealed to be stronger positive predictors for customer satisfaction in the private domain than in the public.

However, the interpretation of these results should be approached with caution, as it was not possible to compute the internal consistency reliability of these single-item scales. Furthermore, the construct validity is assumed to be weaker for single-item scales, as these do not pass the widely known rule that at least three items should be used per measurement (Hoeppner, Kelly, Urbanoski & Slaymaker, 2011). Aside from the validity and reliability concerns, the revealed variable variance on the sector level forms fertile ground for further research.

5.1 Implications

Depending on the dimension(s) in focus, several practical implications can be derived from the validated model proposed in this study. The study could serve as either a starting point or guidance in both practical and theoretical approaches.

First, from a theoretical point of view, the insights gained from this research can be used to provide an overall understanding of the key e-service dimensions. By adopting a user-centric perspective, the service-specific attributes affecting customer satisfaction can be better understood.

Whilst the proposed model may not have incorporated all the attributes affecting e-satisfaction, it calls for further investigation in order to be extended and transformed into a cross-sectoral measurement instrument that captures the full complex concept of customer satisfaction.

The latter brings forth the most contributing aspect of this study, due to how it bridges the gap between the established theories in the public and private domain. The identification of three cross-sector dimensions, which have been proven to play a significant role in the

(43)

development of customer satisfaction, can aid and support both practitioners and academics in redesigning, adjusting or sharpening financial e-services.

Furthermore, of a more pragmatic interest, the thorough understanding obtained through this study can prove of practical use in assisting and supporting planners, strategists and developers as they reshape and adjust services in practice. The empirical data might support the efforts of those aiming to draw management attention towards a more fruitful and effective focus on the development of certain attributes (Alawneh et al., 2013). The

strengthening effect of raising awareness based on empirical evidence will enable practitioners to rethink and direct their service offerings in terms of effectiveness and efficiency (Verdegem & Verleye, 2009).

Foremost, the progressively growing number of organisations, where the sector prefix is difficult to determine, are liable to an environment where the innovation decisions largely depend on the pre- or absence of budget surpluses and administrative savings (Borins, 2014; Lindgren & Jansson, 2013; Verdegem & Verleye, 2009). By acknowledging and focussing on these dimensions, decision making can be advanced regarding potential output and

efficiency. Ergo, endeavours to balance the input resources with the possible output scenarios will be much more rewarding due to the increased accuracy of those estimates.

In addition, not only can profound initiatives be estimated more precisely, but it also allows for a more detailed preparation and scenario-based anticipation, allowing resources to be allocated more efficiently. The dependency of innovative opportunities or profound corrective adjustments on the state of the federal budgets will decrease, and opportunities can be exploited by a more demand-driven approach. By a more correct estimation of the

anticipated scenario aimed for, utilisation from both the user and the provider side can be maximised.

(44)

This study’s relevance for private sector organisations can be explained following the same line of reasoning. However, those responsible for private sector e-services will be guided by different incentives. With the lurking threat of losing customers, private

organisations are focussed on retaining existing customers, as acquiring new ones is much more expensive (Gallo, 2014).

In the cross-sector comparison, the presumed attribute of trust is a stronger predictor for satisfaction than in the public sector. However, looking only at the private model

parameters, availability is identified as the strongest indicator. When one attempts to enhance and improve the overall customer satisfaction, investing in availability is assumed to exert the greatest influence. However, practitioners should bring this into perspective by determining the current state of the dimension. Furthermore, attention should be devoted to extending the measurement of the single-item scales used in this research, in order to further sharpen and finalise the model proposed in this study.

5.2 Limitations and future research opportunities

As the previous section discusses, the results can facilitate and support the

prioritisation of performance areas for further improvement. However, caution is required while interpreting the results of this study.

5.2.1 Sample & recruitment

Before all else, the study is limited in its generalisability due to the composition of the utilised sample. Because of cost-effectiveness and time constraints, convenience sampling was applied in the form of snowball and network sampling. However, recruiting via the personal network and social media has resulted in a sample with low external validity, and hence, low generalisability. Despite the equal divide in men/women and sector division, more than three-quarters of the respondents are under the age of 35, and almost half of the group

(45)

consists of students, followed by more than one-third of the group being private sector employees. The unequal distribution of the sample imposes concerns over representativeness and repetition of data, concerns often perceived as general disadvantages of snowball and network sampling (Biernacki & Waldorf, 1981). Besides the commonalities in terms of demographics, it should also be taken into consideration that, due to these commonalities, the respondents might also be similar in their beliefs and interest (Heckathorn, 1997). Therefore, exploitation of the researchers’ personal network might have reinforced the retrieval of repetitive data.

In order to increase the generalisation and accuracy of the research findings, future studies should keep track of the composition of the group of respondents during the sampling phase, allowing for direction to be adjusted as needed.

5.2.2 Research method

With regard to the research method adopted in this study, limitations arise due to the fact that the research conducted is not longitudinal. Despite how the performance of

longitudinal study would go beyond the scope of this research, it would have also allowed for examining the reliability of the single-item measurements. The proposed model, which already explains 70% of the variance in the central dimension of customer satisfaction, could be sharpened and extended by including more items concerning the observed variables while also validating the whole model with longitudinal data. The motivation behind the

recommendation of retrieving repeated observations lies in the fact that this would not only enhance the validity and reliability but could also serve as a central clarification in both sector domains, where many definitions, models and measurements have been proposed. In this way, such future research would add to a solid foundation in the literature.