Service Quality, Customer Satisfaction
and Brand Loyalty in the Swedish
Subscription Video On
Demand-Industry
BACHELOR DEGREE PROJECT THESIS WITHIN: Business & Administration NUMBER OF CREDITS: 15 ECTS
PROGRAMME OF STUDY: Marketing Management AUTHOR: Oliver Berg, Elliot Strand, Viktor Sandell JÖNKÖPING May 2019
A mixed methods study on what factors of service quality affect customer satisfaction and brand loyalty within the subscription based
i
Bachelor Degree Project in Business Administration
Title: Service Quality, Customer Satisfaction and Brand Loyalty in the Swedish Subscription Video On Demand-Industry
Authors: Oliver Berg (940624), Elliot Strand (950713), & Viktor Sandell (970708) Tutor: MaxMikael Wilde Björling
Date: 2019-05-17
Key terms: Subscription video on demand (SVOD), Service Quality, Customer Satisfaction, Brand Loyalty
Abstract
The main objective of this research is to measure the relation between different dimensions, or constructs of service quality with customer satisfaction and brand loyalty; thereby identifying the most important factors of service quality that affect customer satisfaction and brand loyalty, in the context of subscription video on demand (SVOD) services in the Swedish market. This study utilised a mixed methods approach divided into two stages, whereof the first involved multiple focus groups for collecting qualitative data. The second stage consisted of quantitative data collection through online questionnaires from 122 valid respondents, ranging from the age of 18 to 29 by a non-probability sampling method. SPSS was used to analyse the data with three statistical analysis methods: bivariate correlations, for establishing evidence of construct validity; a reliability analysis to test Cronbach alpha for internal consistency; and multiple regression analysis to examine the relationship between independent and dependent variables. The research findings indicate that there is a positive, significant relationship between customer satisfaction and certain constructs of service quality. Also, findings indicate that the statistical relationship between the constructs of service quality and brand loyalty are partially mediated by customer satisfaction. As a result, companies operating in the Swedish SVOD-industry need to focus on improving the significant factors of service quality, as they may increase customer satisfaction of 18 to 29-year-old customers, and further create long-term relationships as a way of gaining competitive advantage.
Keywords: Subscription video on demand (SVOD), Service Quality, Customer Satisfaction, Brand Loyalty.
ii
Acknowledgements
The authors of this thesis would like to acknowledge everyone who have been involved and contributed to the research project. First of all, the authors would like to express a special thanks of gratitude to their tutor, MaxMikael Wilde Björling, who have been helpful in providing guidance and support throughout the research project.
Secondly, the authors would also like to extend their gratitude to the individuals who participated in the focus group sessions. Your insights and opinions expressed have been invaluable for developing the online questionnaire. Also, many thanks to all respondents who completed the online questionnaire. Without your answers, there would not have been any quantitative data to analyse.
Third, the authors would like to express their appreciation to the teaching team at Jönköping International Business School for providing the authors with clear guidelines and instructions throughout the research project.
Last but not least, the authors would finally like to thank Jonas Persson for sharing his knowledge and technical expertise in questionnaire development in Qualtrics, and data analysis in SPSS.
This thesis would not have been completed without the concerned parties’ enormous help and support. Thank you indeed!
Jönköping, May 17, 2018
iii
Table of Contents
1. Introduction... 1
1.1 Background ... 2 1.2 Problem ... 4 1.3 Purpose ... 5 1.4 Delimitations ... 52. Frame of Reference... 6
2.1 Literature review ... 62.1.1 The Subscription Video On Demand-industry... 6
2.1.2 Generation Y&Z ... 7
2.1.3 Customer Satisfaction ... 7
2.1.4 Brand Loyalty ... 8
2.1.5 Service Quality ... 8
2.2 Conceptual Model & Hypotheses ... 10
3. Methodology ... 11
3.1 Research Context ... 11
3.2 Research Approach ... 11
3.3 Research Design ... 12
3.3.1 Stage 1-Exploratory Qualitative Research ... 12
3.3.2 Stage 2-Descriptive Quantitative Research... 13
3.4 Population & Sample ... 14
3.4.1 Sample Size ... 14
3.4.2 Sampling Method ... 15
3.4.3 Sampling Error ... 15
3.5 Data Collection Methods ... 16
3.5.1 Focus Groups for Qualitative Data Collection ... 16
3.5.2 Online Questionnaire for Quantitative Data Collection ... 18
3.5.3 Questionnaire Development ... 18
3.5.4 Conceptualisation ... 19
3.5.5 Questionnaire Design ... 19
3.5.6 Questionnaire Translation ... 21
3.5.7 Questionnaire Testing & Revision ... 21
3.5.8 Quantitative Data Collection ... 21
3.6 Data Analysis Methods ... 21
3.6.1 Correlational Research Design & Multiple Regression ... 21
3.6.2 Multiple R & Adjusted R2 ... 22
3.6.5 Standardized Beta Coefficient ... 23
3.6.3 Multicollinearity... 23 3.7.1 Reliability ... 23 3.7.2 Construct Validity ... 24 3.7.3 Convergent Validity ... 24 3.7.4 Discriminant Validity ... 25 3.8 Ethical Considerations ... 25
iv
4. Measurement Instruments ... 26
4.1 Constructs of Service Quality ... 27
Efficiency… ... 27
System Availability ... 28
Privacy..….. ... 28
Fulfilment - Library of Content ... 29
Recommendation System ... 30
Quality of Experience ... 31
Pricing…… ... 31
4.2 Additional Variables ... 32
4.2.1 Overall Customer Satisfaction ... 32
4.2.2 Brand Loyalty ... 32
4.3 Classification Variables ... 32
5. Statistical Analysis & Results ... 33
5.1 Reliability & Validity... 34
5.1.1 Cronbach Alpha ... 34
5.2 Convergent & Discriminant Validity ... 34
5.2.1 Convergent Validity ... 35
Efficiency… ... 35
System Availability ... 35
Privacy..….. ... 35
Recommendation System ... 36
Fulfilment - Library of Content ... 36
Quality of Experience ... 36
Pricing…… ... 37
5.2.2 Discriminant Validity ... 37
5.3 Multiple Regression Analyses ... 38
Efficiency… ... 38
System Availability ... 39
Privacy..….. ... 40
Fulfilment - Library of Content ... 41
Recommendation System ... 42 Quality of Experience ... 43 Pricing…… ... 44 5.3.1 Hypotheses Testing ... 45 Hypothesis 1 ... 46 Hypothesis 2 ... 47 Hypothesis 3 ... 48 Hypothesis 4 ... 49
6. Conclusion ... 51
7. Discussion ... 53
7.1 Theoretical Implications ... 56 7.2 Practical Implications ... 568. Limitations... 58
8.1 Suggestions for Further Research ... 59
v
10. Appendices... 65
10.1 Questionnaire ... 65 10.1.1 Information Sheet ... 65 10.1.2 Form of Consent... 66 10.1.3 Questionnaire in English ... 67 10.1.4 Questionnaire in Swedish... 70 10.1.5 Absolute Frequency ... 74 10.2 Convergent Validity ... 87 10.2.1 Efficiency ... 88 10.2.2 System availability ... 88 10.2.3 Privacy ... 89 10.2.4 Recommendation system ... 8910.2.5 Fulfilment - Library of Content ... 90
10.2.6 Quality of Experience ... 91
10.2.7 Pricing ... 91
10.3 Discriminant Validity ... 92
10.4 Model Summaries and Coefficients ... 93
10.4.1 Efficiency Construct ... 93
10.4.2 System Availability Construct ... 93
10.4.3 Privacy Construct ... 94
10.4.4 Fulfilment-Library of Content Construct ... 94
10.4.5 Recommendation System Construct ... 95
10.4.6 Quality of Experience Construct ... 95
10.4.7 Pricing Construct... 96 10.5 Hypotheses Testing ... 97 10.5.1 Hypothesis 1 ... 97 10.5.2 Hypothesis 2 ... 98 10.5.3 Hypothesis 3 ... 99 10.5.4 Hypothesis 4 ... 100
1
1. Introduction
________________________________________________________________________________
In this section, an introduction followed by a background on the research topic is given. A
discussion of the problem and the purpose of this research is later covered. The research question is presented, along with the research objectives. Moreover, the last section discusses the
delimitations of the research.
________________________________________________________________________________
About a decade ago, smartphones by the standard as we know them by today did not exist. Three decades earlier, most people did not even own a personal computer. Nowadays, people are spending their time looking at bright screens, whether it is a smartphone screen, a computer screen or a television screen. The progression of technology is moving at a faster pace than ever before and the advances in technology have paved the road for new business opportunities to emerge. During this development, the opportunity to exploit the systems arose, and piracy became increasingly popular by consumers of online entertainment (Poort, Quintais, van der Ende, Yagafarova, & Hageraats, 2018). Subjected to many challenges, consumers’ willingness to pay for entertainment decreased, and the danger of piracy harmed businesses worldwide, facing economic losses estimated to over $200 billion in 2005, as a result of counterfeit and pirated goods (Organisation for Economic Co-Operation and Development, 2007). In recent years, rapid advances in technology have enabled users to access increased bandwidth and speed in their internet connections. Such improvements in combination with modern operating systems and optimized personal computers have created new, lucrative business opportunities.
Through several business ideas that have risen, subscription based video on demand streaming is an example of a service that has made its mark on the modern media industry. The phenomenon is more commonly known as subscription video on demand (SVOD). The services include all forms of video services where users have unrestricted access to a frequently updated variety of movies and series, at a fixed monthly rate, and where they have the ability to play, pause, fast forward and rewind the content as they please (Grece, 2014). New industries are argued to be important for economic growth, as they may transform mature industries and drive innovation, which accurately describes what has happened within the entertainment industry (Feldman & Tavassoli, 2014). Furthermore, for businesses to flourish, understanding potential customers is of high importance in order to generate a
2
superior offering. Absolute determinants of customer satisfaction and brand loyalty within the industry of SVOD are yet to be established, due to the limited research conducted in the field. The positive relationship between the two concepts, however, has been made apparent by previous research (Oliver, 1999; Ahmed, Chandio, & Quresh, 2015).
1.1 Background
Two highly dominant marketing topics of today are customer satisfaction and brand loyalty. The actuality of the topic has led to extensive research in the field, and vast advances for both practice and academia (Schirmer, Ringle, Gudergan, & Feistel, 2016). Satisfaction can be defined as a person’s enjoyment or disappointment resulting from a performance specific expectation, in comparison to the expectancy disconfirmation, which implies diminishment of prior expectations (Oliver, 1980; Kotler, 2000). Also, some researchers argue that disconfirmation of expectations has an effect on perceived quality, together with various other factors, such as product features and attributes (Bolton & Drew, 1991; Parasuraman, Zeithaml, & Berry, 1988). Perceived quality is also proposed to be similar to attitude, and that the process shaping perceived quality is similar to the journey of attitude creation. (Gotlieb, Grewal, & Brown, 1994; Dodds, Monroe, & Grewal, 1991; Parasuraman et al., 1988).
Furthermore, brand loyalty is based on the fundamentals of customers seeking the best value, in which customer satisfaction is a predominant driver (Reichheld, 1994; Anderson & Sullivan, 1993). The link between satisfaction and loyalty has been evident in previous research, though reports have also indicated that satisfaction alone is not always necessarily a recipe for loyalty (Reichheld & Sasser, 1990). The two concepts can be examined through the utilization of the Net Promoter Score (NPS), which is used in management as a loyalty metric to predict firm growth (Keiningham, Aksoy, Cooil, Wallin Andreassen, & Williams, 2008). Even though the NPS has been widely criticized for its simplicity (Kristensen & Westlund, 2004; Morgan & Rego, 2004; Kristensen & Eskildsen, 2014), it has still captured a lot of interest and is frequently applied in marketing research (Kristensen & Eskildsen, 2014).
Aided by technological enhancements, the media industry has recognised the capability of evolving the internet into a video outlet. The vast increase in online media has brought to question the viability of cable television and further motivated new companies to arise, as well as transition, from traditional media to online platforms. This has allowed for consumers to access media content considerably quicker. The fast access, along with the simplicity of modern media sharing aided by internet
3
development may, according to Cha and Chan-Olmstead (2012), have been contributing to the debated cannibalising effect on cable television. As mentioned by Trowbridge (2013), “A paradigmatic shift has been occurring in terms of how film and media content at large is distributed, spread, and shared between people and organisations...” (p. 228). The presence of a shift has also been noticeable in the Swedish market (Mydigheten för Press, Radio och TV, 2017). For many individuals in Sweden under the age of 30, screens have been a part of their daily life since childhood, along with the internet. Individuals in Sweden, younger than 30 years of age, are among those who watch cable TV the least amount of time each day, while being the heaviest users of SVOD-services (Myndigheten för press, radio och TV, 2017).
The Swedish market consists of both subscription-based video on demand services as well as services that rely on advertisements as their main source of revenue. In a study conducted by Audience Project (2017), 59% of the Swedish population use the internet in order to access TV shows or movies on a weekly basis. The most popular sources for online media in Sweden are YouTube, SVT Play, Netflix, TV4 Play, Viaplay, Cmore, and HBO Nordic, whereof YouTube, SVT Play, and TV4 Play do not require a subscription. Instead, SVT-play is free for users to watch, whereas YouTube and TV4 Play rely on advertisements. Contrarily, Netflix, Viaplay, Cmore, and HBO Nordic provide subscribers with on demand content, in exchange for a monthly fee.
Since April 2018, new rules have made it possible for consumers to stream online content all across the EU (European Parliament, 2018). Naturally, this includes videos from the previously mentioned service providers. The reasoning behind the decision is partially to discourage piracy, as some content was geographically limited to only some countries within the EU before the new regulation. The regulation aims to counter copyright related issues, as copyrights were previously acquired and enforced on a country-to-country basis. Together with new roaming regulations, travelling consumers within the EU now have the same access to the content they usually watch at home (European Parliament, 2018).
4
1.2 Problem
In recent years, a dramatic switch from analogue home entertainment towards internet-based behaviour has been apparent, which has allowed for video streaming services to thrive (Rodríguez, Rosa, Costa, Abrahão, & Bressan, 2014). A survey conducted by Swedish quality index (2018), showed that 70% of Swedish video on demand subscribers believe that they would remain loyal to their current streaming service providers for at least one year. This entails a 30% gap in consumers who believe that they would not remain loyal to their current service provider. The reasons for why the doubtful 30% question their loyalty remains to be determined. Furthermore, the reason for why the Swedish population within the age category 18-29 is of interest to this study, is since people of these demographics are among those who are watching cable TV for the least amount of time, and spend the most time using SVOD-services instead (Mydigheten för Press, Radio och TV, 2017).
As new business opportunities emerge through the change in consumer preferences, the matter of investigating factors that allow businesses to thrive in the focal industry becomes critical. In order for businesses in the subscription-based video streaming industry to grow, returning customers are of high importance. As proven by Reichheld and Sasser (1990), customer loyalty may lead to increased profits by 25 to 85 percent. In order to maintain subscribers, companies need to understand which factors are satisfying their customers to the extent where they perceive the service to provide a higher value than what they are paying. To date, research regarding this issue remains limited.
5
1.3 Purpose
To address the mentioned gaps in current literature, this study aims to explore what factors of service quality affect customer satisfaction and brand loyalty in the SVOD-industry. This is important as the SVOD-industry is subject to change in both consumer taste and technology. Thus, SVOD-service providers must focus their efforts toward factors of service quality that are important drivers of customer satisfaction, which in turn, may facilitate the creation of long-term relationships. In order to do so, companies must be able to measure users’ perceptions of service quality (Juluri, Tamarapalli, & Medhi, 2016). Accordingly, the question that this study will aim to answer is:
“What is the relation between service quality, customer satisfaction, and brand loyalty in the Swedish SVOD-service industry, amongst 18 to 29 year-olds?”
In order to answer the proposed research question, the following objectives are presented:
➢ To identify what factors of service quality affect customer satisfaction and brand loyalty with regards to SVOD-services.
➢ To determine the importance of the derived factors of service quality in relation to customer satisfaction and brand loyalty.
➢ To examine the mediating effect of customer satisfaction between service quality and brand loyalty.
1.4 Delimitations
This study seeks to solely focus on services which exclusively offer video streaming, more specifically, on demand video streaming services that rely on subscription fees. Thus, other streaming services that do not solely rely on subscriptions are disregarded. As there has been a finite amount of studies conducted with focus on the Swedish market, this study aims to make a contribution to the scarcely researched field.
In Sweden, the major market shareholders are Netflix, Viaplay, C More and HBO Nordic (Mydigheten för Press, Radio och TV, 2017), and will for this reason be included in this study. This entails disregarding services such as TV4 Play, and Dplay, as they can be used without a subscription, and the concept of loyalty could be interpreted differently depending on the revenue model utilised by the service provider. This may therefore contribute to misleading results.
6
2. Frame of Reference
________________________________________________________________________________
The following section will review the work of literature on the subscription video on demand-industry, generation Y and Z, customer satisfaction, brand loyalty, and service quality. This is followed by a presentation of a conceptual model and the developed research hypotheses.
________________________________________________________________________________
2.1 Literature review
2.1.1 The Subscription Video On Demand-industry
With increased quantity of user ratings from online streaming services, more studies in recent years have focused on the motivation for why users view content from online media services instead of cable television (Bondad-Brown, Rice, & Pearce, 2012). Studies have also covered the various impacts that piracy has had on the industry (Hui & Png, 2003; Borja, Dieringer, & Daw, 2014), whether the cannibalization of online video platforms on cable television indeed takes place (Cha & Chan-Olmsted, 2012), and to what extent changes in consumer behaviour can be caused by fluctuations in video quality (Krishnan & Sitaraman, 2013). However, research on SVOD-services have mainly focused on what factors impact an individual when deciding whether to choose online streaming over cable TV (Lee, Nagpal, Ruane, & Lim, 2018).
The SVOD industry has been evolving naturally, as new technologies become available, and consumer preferences and demands change. This further emphasises the importance of providing relevant content for their viewers through well-crafted search engines and recommendation systems (Hasan, Jha, & Liu, 2018). All SVOD-services investigated in this study utilise a recommendation system and a search engine. The latter is a tool for users to locate specific information sources, however, requires the user to know a significant part of the information which they are searching for, in order to be successful in finding the correct result (Hasan et al., 2018). Recommendation systems display suggestions for users about the next activity to indulge in, which are derived from the users preferences, history, and several other factors (Park, Choi, Kim, & Kim, 2011). Thus, recommendation systems are not limited to the memory of the user (Hasan et al., 2018). A well-crafted recommendation system works towards attracting consumers to continue watching, and remain on the site by recommending relevant content to the user (Li, Zheng, Yang, & Li, 2014).
7
According to Häubl and Trifts (2000), a recommendation system possesses the capability to significantly increase user’s experience and exposure of the particular platform.
2.1.2 Generation Y&Z
It is agreed that individuals born around 1980, up until the mid, or late ‘90s belong to Generation Y (Nagy, 2012; Shatto & Erwin, 2016). However, there is no agreed definition of the age category Generation Z, as some treat this age category as logically following Generation Y, including individuals born from the mid1990’s (Shatto & Erwin, 2016; Southgate, 2017; Montana & Petit, 2008). Meanwhile, Caroline Geck (2006) treat individuals born in or after the year of 1990 as Generation Z, as these share more characteristics with regards to internet knowledge and habits, since they were all born into a world where internet existed. As this study aims to investigate the age category of 18 to 29 year-olds, it can be agreed that the target population of this study consist of young individuals belonging to Generation Y, as well as the oldest ones of Generation Z. With this in mind, the target population is not represented by the entirety of generation Y and Z. Thus, some characteristics may not be shared between individuals belonging to the older tier of Generation Y, or the younger tier of Generation Z.
Moreover, individuals of Generation Y are considerably more skilled in multitasking than previous generations, while they generally are more technologically knowledgeable (Nagy, 2012). Also, they often have difficulties in concentrating on one issue for a long period of time, which is even more apparent when considering Generation Z (Nagy, 2012; Shatto & Erwin, 2016). According to Shatto and Erwin (2016), Generation Z is also the first generation to watch programming or streaming services when they find it convenient, as smart devices and sophisticated services allow for them to watch content anywhere, and at all times.
2.1.3 Customer Satisfaction
Customer satisfaction has for decades been a highly researched topic within marketing, and shown to be a viable predictor of future consumer purchasing intentions (McQuitty, Finn, & Wiley, 2000; Cardozo, 1965; Oliver 1980; Bloemer & Kasper, 1995). Marit G. Gundersen, Morten Heide, and Ulf H. Olsson (1996) define satisfaction as “a postconsumption evaluative judgment concerning a specific product or service” (p.74). Oliver (1980) argues that the determinants of satisfaction include factors such as consumers pre-purchase expectations and post-purchase evaluation. This, in turn, means that disconfirmation resulting from the consumption of a good, should ultimately decrease as consumers’ knowledge of the product experience increases, hence a decrease in pre-purchase expectations
8
(McQuitty et al., 2000). Customer satisfaction is argued to increase a firm’s profitability through the development of customer retention (Hennig-Thurau & Klee, 1997). Furthermore, Patterson, Johnson, and Spreng (1996) have discovered customer satisfaction as the primary predictor of repurchase intent.
2.1.4 Brand Loyalty
The concept of brand loyalty is highly connected to customer satisfaction, as it often is the ultimate goal towards which businesses strive, and where customer satisfaction is one of the predominant drivers, mainly together with brand trust (Oliver, 1999; Ahmed et al., 2015). However, there are contradicting studies which argue that the certainty of the relation between the two concepts is not as straightforward as suggested. Instead, factors such as demographics, education, and age also play major roles in the process of attaining loyal customers (Kumar, Pozza, & Ganesh, 2013; Keiningham, Aksoy, Cooil, & Hsu, 2007; Homburg & Giering, 2001). Although there is no fine line regarding what a loyal customer is, Newman and Werbel (1973) define loyal customers as those who repurchased a brand and only considered that particular brand without engaging in any brand related information search.
In a study conducted by Hellier, Carr, Geursen, and Rickard (2003), with support from empirical evidence, customer satisfaction was found to have a direct positive effect on repurchase intention (Anderson & Sullivan, 1993; Bolton 1998; Cronin Jr. & Taylor, 1992; Fornell, 1992; Oliver, 1980). Gupta, Lehmann and Stuart, in their article “Valuing Customers” (2004), argue that an improvement in customer retention by 1% may lead to an increase in firm value with up to 6,75% if managed correctly. The article, however, does not focus on SVOD-services, but rather measures the value of subscriber retention in general.
2.1.5 Service Quality
In previous studies, results have shown that service quality is a significant predictor of customer satisfaction (Gorondutse & Hilman, 2014). Service quality is a measure of how well the service is performed, according to a customer’s subjective assessment, however, it is not only an evaluation of a services’ outcome as it also involves evaluations in the way that the service is delivered (Parasuraman, Zeithaml, & Berry, 1985). Based on insights from exploratory research, Parasuraman et al. (1985) have developed a model which summarises ten dimensions of service quality as consumers perceive it. In later studies, however, a strong correlation among the ten dimensions was found. The ten dimensions were, therefore, combined into five dimensions of reliability,
9
responsiveness, assurance, empathy, and tangibles which constituted as the basis of a tool known as SERVQUAL, which is used for measuring service quality (Parasuraman, Zeithaml, & Berry, 1996).
While the SERVQUAL instrument has been extensively used in various contexts for measuring service quality, the rise of the internet as a channel for selling goods and services demanded a refinement of the SERVQUAL instrument (Parasuraman, Zeithaml, & Malhotra, 2005). Accordingly, Electronic Service Quality (E-SERVQUAL) was developed and designed to measure the service quality of electronic commerce websites. It was developed in 2005 by A. Parasuraman together with Valarie A. Zeithaml and Arvind Malhotra. The scale consists of 22 items categorised into four dimensions: efficiency, system availability, fulfilment and privacy (Parasuraman et al., 2005).
Several other scales such as WebQual (Loiacono, Watson, & Goodhue, 2007), WebQual (Barnes & Vidgen, 2002), and SITE-QUAL (Yoo & Donthu, 2001) have been developed for evaluating and measuring website quality. The WebQual scale developed by Barnes and Vidgen (2002) has been criticised for lacking aspects of the purchasing process, as it only measured the consumer’s website experience until the point of purchase (Parasuraman et al., 2005). Also, the WebQual designed by Loiacono, Watson and Goodhue (2007) was, in the same article criticised for generating information for website designers instead of measuring customer experiences, as well as for not including the fulfilment dimension. Parasuraman et al. (2005) argue that the customer’s evaluation of the quality of a website involves both experiences during website interaction, as well as service related experiences following interaction with the website. Therefore, E-SERVQUAL is argued to be a superior satisfaction measurement, as it takes into account each stage of a customer’s interaction with a website (Parasuraman et al., 2005).
Moreover, a significant amount of research has also facilitated the development of metrics that attempt to measure the quality of service, or quality of experience perceived by the user (Juluri, Tamarapalli, & Medhi, 2016; Su et al., 2016; Rodríguez et al., 2014; Torres Vega, Perra & Liotta, 2018; Maia, Yehia, de Errico, 2014). Furthermore, research has shown a link between quality of service and repurchase intent, and how it is mediated through customer satisfaction (Brady & Robertson, 2001).
10
2.2 Conceptual Model & Hypotheses
The proposed conceptual model (figure 1.) for this study assumes that the best way to explain the effect of the independent variable, service quality, on the dependent variable, brand loyalty, is by using a third variable, customer satisfaction, as a mediator. That is, instead of the independent variable causing the dependent variable directly, the dependent variable is caused by a mediator, which in turn is caused by the independent variable. The conceptual model for this research, therefore, assumes that service quality, consisting of several dimensions, influences customer satisfaction, which in turn influences brand loyalty. The constructs of service quality were derived partly from the multiple-item scale of E-SERVQUAL (Parasuraman, Zeithaml, & Malhotra, 2005), studies by Hasan, Jha and Liu (2018), Krishnan and Sitaraman (2013), Maia, Yehia and Errico (2014) and MUX (2017), as well as from qualitative, primary findings gathered from focus groups. The final constructs included efficiency, system availability, fulfilment-library of content, privacy, quality of experience, pricing and recommendation system.
Figure 1.
From the conceptual model, the following hypotheses were developed:
H1: At least one construct of service quality is significantly, positively associated with customer
satisfaction
H2: At least one construct of service quality is significantly, positively associated with brand loyalty
H3: Customer satisfaction is significantly, positively associated with brand loyalty
11
3. Methodology
________________________________________________________________________________
The following section presents the research context, the research approach, and the research design, where the two stages of this study are explained. This is followed by a description of the target population, sample size, and sampling method. Lastly, the methods for data collection and analysis are presented and discussed, as well as ethical considerations.
________________________________________________________________________________
3.1 Research Context
The study aims to research the Swedish SVOD sector, due to the previously mentioned changes with regards to consumer behaviour in the industry (Mydigheten för Press, Radio och TV, 2017; Svenskt Kvalitetsindex, 2018). As both domestic and global players compete for market share in Sweden, understanding how consumers value factors of quality is of high importance.
3.2 Research Approach
In quantitative studies, a deductive research approach is widely applied when testing a conceptual or theoretical structure, and when attempting to establish, confirm, or validate relationships, in order to make generalisations that add value to theory (Churchill & Brown, 2004; Collis & Hussey, 2014). A deductive approach was therefore considered to be an appropriate, logical process for exploring the phenomena from a top-down approach, where the findings would provide indications and generalisations, and which in turn could contribute to theory. This was utilised through developing a conceptual model, deriving constructs from existing research which served as foundations for the remaining parts of data collection and analysis.
The study was executed in two steps, where both qualitative and quantitative research methods were utilised. This course of action is known as mixed methods research and aims to draw from the strengths as well as minimize the weaknesses associated with the two approaches (Creswell, 2013). For this research, the utilisation of mixed research methods allowed the researchers to explore the problem, and obtain a deeper understanding before conducting the descriptive stage of the research.
12
3.3 Research Design
Research designs are generally classified into three broad categories: descriptive, exploratory and causal-comparative (Burns, Veeck, & Bush, 2017). The descriptive research approach is a method that focuses on describing the nature of a particular condition as it exists during the time of the study. It involves an observational basis where attributes of a particular phenomenon are identified, or where the correlation between phenomena are explored (Leedy & Ormrod, 2001). Exploratory research is unstructured and informal research, and is generally conducted when the researcher has limited knowledge about the research problem, and where there is a need for additional and more recent information. Causal-comparative research is on the other hand undertaken to measure the causality in relationships between variables (Burns et al., 2017), and will therefore not be included in this study.
Based on the research question, this study was conducted in two stages: an exploratory qualitative research stage, followed by a descriptive quantitative research stage. The exploratory research was undertaken in the outset of the research project to efficiently review previous research through library and online sources. Also, it assisted in the collection of current information about the general nature of the research problem, which was necessary in order to conduct a descriptive research design. The results of the exploratory approach could then be quantified and described with more detail, thereby reflecting the application of a descriptive approach. Conducting research in this manner allowed for the exploratory qualitative phase to serve as a foundation for the descriptive quantitative phase, as well as providing first-hand knowledge for the research questions.
3.3.1 Stage 1-Exploratory Qualitative Research
In the exploratory qualitative stage data was collected from focus groups, which is a useful technique for gaining individual insights on the research problem and to generate fresh ideas from a limited sample of respondents (Burns et al., 2017). Thus, a total number of three focus groups were conducted to ensure that the results were consistent and rigorous. The findings from the focus groups were then used to guide the second phase of the research where a questionnaire was developed and distributed.
In the development of the questionnaire, findings from the focus groups were combined with the findings obtained from academic papers and journals. The secondary sources were used in order to take advantage of questions appearing in questionnaires of prior research and to ensure that the requirements of face validity were met to a reasonable extent. Face validity, in this case, ensures that the questions measure what they are supposed to through high relevance and that desired topics are covered (Fallowfield, 1995; Rowley, 2014). Hence, the purpose of the focus groups was to generate
13
fresh ideas and to supplement findings obtained from research papers in order to manifest a degree of contextualisation.
3.3.2 Stage 2-Descriptive Quantitative Research
For the quantitative descriptive stage, a cross-sectional sample survey was conducted using structured computer-administered questionnaires. A cross-sectional survey design is one of the most common designs applied in research (Olsen & St. George, 2004). Its purpose is to measure units from a representative sample of the population, only at one point in time (Salkind, 2010, Burns et al., 2017). Based on the work of dos Santos Silva (1999), cross-sectional surveys can be appropriate for measuring a populations’ attitudes, beliefs, practices, and knowledge concerning a certain phenomenon. Even though it is not possible to obtain information about changes in behaviour over a period of time, cross-sectional studies offer cost efficiency, as the method does not require a lot of time and can be used to examine several outcomes (Salkind, 2010).
Moreover, the purpose of using a structured computer-administered questionnaire was due to its major advantages: ease of administration, standardisation and the reduction of interview evaluation apprehension, i.e. when respondents feel that they have to provide answers that are “right” or “desirable”, and also to facilitate the process of analysing and keeping the study within budget (Burns et al., 2017).
To summarise, this study can be considered to use a combination of exploratory and descriptive research design. In the exploratory stage, an extensive literature review and multiple focus groups were conducted. While in the descriptive stage, a cross-sectional computer-administered questionnaire was used to collect quantitative data.
14
3.4 Population & Sample
A population can be defined as the total number of objects in a dataset, which shares a specific characteristic and can be analysed through collecting and analysing data. In most cases, collecting data from an entire population is simply not possible, while in other cases, vastly resource demanding. Samples of the entire population are therefore used, which can be described as a representable portion, or a subset of the population. When studying a sample, it is important to ensure that it is representable of the entire population, which is determined by the suitability of the used sampling method (Burns et al., 2017). However, a sample can always be regarded as inaccurate. Thus, the aim should be to eliminate the significance of the inaccuracy, which can be accomplished by increasing the sample size (Burns et al., 2017).
The population targeted for this study is Swedish individuals, both male and female, within the age gap of 18 to 29 year-olds. This target population amounts to almost 1.4 million individuals (Valmyndigheten, 2018). This population is amongst the heaviest users of the four SVOD-services included in this study (Mydigheten för Press, Radio och TV, 2017), and constitutes around 18,4% of the total Swedish population over 18 years of age. (Valmyndigheten, 2018). Out of the total Swedish population, more than 50 percent have access to, and use SVOD-services on a daily basis (Mydigheten för Press, Radio och TV, 2017).
3.4.1 Sample Size
Determining a suitable sample size may for any researcher be challenging, as it is generally a settlement between what is perfect in theory and what is manageable in practice (Burns et al., 2017). Even though there are no uniform rules of how a sample size should be calculated, several “rules of thumb” have been brought forward. For example, in the relationship between the number of respondents per variable, Nunnally (1978) recommend a 10:1 ratio, while Stevens (1996) suggests a ratio of 15:1. Green (1991) and Maxwell (2000) independently, found no support for this method and concluded that using this method may lead to a non-sufficient sample size. Maxwell instead suggested between 70 and 100 observations per variable in order to establish correct results. On the other hand, Green (1991), as well as Tabachnick and Fidell (2007), advocated using a method based on 50 observations, plus an additional 8 observations for each independent variable, when testing multiple correlation. Based on the latter, this would entail that this research would require a sample size of 106 respondents, as the measures of service quality consisted of seven constructs, or independent variables. It may be considered naive to believe that all distributed questionnaires will be answered.
15
Thus, 180 potential respondents were asked to answer the survey of which a total of 122 provided complete responses, which translates to a response rate of 67,78%.
3.4.2 Sampling Method
As a master list of the studied population is not possible to obtain, a non-probability sampling method was to be used. A non-probability sampling method utilizes subjective techniques, which entails an unknown probability of participants being selected (Burns et al., 2017). Purposive sampling is a method categorised as non-probability, which implies that the researchers invite certain individuals to participate, based on their shared characteristics, and where the probability of being selected is unknown (Burns et al., 2017).
3.4.3 Sampling Error
It is, however, important to note that due to the sampling method applied in this research, the representativeness of the sample being studied may be questionable. Hence, the ability to make generalisations from the sample to the population is limited. Even though the researchers were confident that the sample selection would yield knowledgeable respondents about the given issue and would be qualified to participate, the inherent bias in the purposive sampling should be considered.
In order to obtain a sample as representative as possible of the target population, despite the chosen sampling method, respondents were drawn from a national homogenous sample on the basis of their age, and existing prior experience of using SVOD-services. Both criteria were of high importance in order to avoid sampling error, since the survey was of post-evaluative nature and that respondents had to fit within the targeted age category. To further improve the representativeness and reduce bias, selecting participants from only one particular subset of the population was avoided. For example, invitations to participate were sent electronically to individuals from diverse parts of the country, both larger cities, as well as a range of smaller towns. Individuals were also selected based on other demographics, such as occupation, as well as parental and relationship status.
16
3.5 Data Collection Methods
In this research, both primary data and secondary sources were used to address the research problem. The secondary sources used constituted of scholarly journal articles that were extensively reviewed, as well as data from governmental institutions. There are several advantages that justify the utilization of secondary data. For example, the data can be inexpensive to collect, obtained quickly, and it may also enhance the primary data by providing supporting evidence (Burns et al., 2017). However, the research may not be up-to-date nor match the researcher’s need. It should also be handled carefully, as previous researchers have analysed and interpreted the data, and subjectiveness may alter the findings.
Primary data refers to information that is being collected for the first time and specifically for the research project (Burns et al., 2017). The primary data collected for this study was done partly through multiple focus groups, where relevant, qualitative data contributed to the second stage of the research process. The findings were, together with theories collected from secondary sources, used to develop a cross-sectional sample survey, which was conducted using structured computer-administered questionnaires. A major drawback with using a survey is that respondents may not answer all questions accurately, due to misinterpretations (Cooper & Schindler, 2011).
3.5.1 Focus Groups for Qualitative Data Collection
A focus group is a frequently used method for collecting qualitative data and consists of a few participants that is guided by a moderator, through an unstructured discussion which can provide the researcher with relevant information for the research problem (Burns et al., 2017).
The objective of the focus groups was to collect consumers’ perceptions, evaluations and feelings towards the four major SVOD-services in Sweden, and to discover plausible drivers of customer satisfaction and brand loyalty. The discussions, in turn, would also be helpful for developing dimensions and question items for the online questionnaire. The ideal size of a traditional focus group is considered to be six to twelve people (Burns et al., 2017). Although there is no correct answer to how many focus groups should be conducted, small projects generally consist of three to four focus groups in order to ensure that results are consistent and rigorous (Burns et al., 2017).
Considering the selection of participants for the focus groups, purposive sampling was used. Purposive sampling refers to when the selection of individuals is based on the judgement of whom the researcher feels represent the population (Burns et al., 2017). As the purpose was to discover
17
plausible drivers of customer satisfaction and loyalty in SVOD-services, it was necessary that the participants were consumers of SVOD-services, and purposive sampling we therefore considered appropriate.
In order to guide the discussion and maintain focus on the topic, a semi-structured interview technique was applied. Such a technique can uncover valuable descriptive data regarding participants’ perspectives, and personal experiences (Schensul & LeCompte, 2013). Therefore, this technique was considered appropriate for the development of variables and items of service quality, that previous research failed to recognise, and which may contribute to customer satisfaction and brand loyalty.
In the finishing stages of each focus group, the discussion leader conducted a member check, where the main topics brought up during the discussion were summarized, for the focus group members to approve, explain or rephrase. This was done in order to ensure that the points were interpreted as intended by the participants (Harper & Cole, 2012).
After each session, the transcripts were analysed through a thematic analysis, where the researchers partly were searching for repeated points and themes not examined by previously reviewed literature. Analysing focus groups thematically can result in great advances since the method allows for flexible, yet richly detailed conclusions to be drawn (Braun & Clarke, 2006). However, researchers should utilize this method in a careful manner since flexibility may lead to inconsistency when deriving appropriate themes from the focus groups (Holloway & Todres, 2003).
Moreover, the focus groups were conducted by sending out requests in student groups and to specific individuals who are familiar with the research topic. The sessions took place in group rooms at Jönköping University, as this was considered to be a familiar setting for all participants. Beverages and snacks were provided for the participants. In total, three focus groups of eight participants were carried out over one week. All participants of the three groups were within the targeted age category, however, the youngest individual attending was 19 years old, and the oldest 27. Thus, the entire target population was not represented during the conducted focus groups. Two researchers acted as moderators during the discussions in order to guarantee the coverage of all questions formulated prior to the meetings. The remaining researcher made sure that the discussions were recorded and that main points were transcribed at the time of the discussion. Before initiating conversation, all participants signed a form of consent which clearly stated why their opinions were desired, the purpose of the study, and information regarding their rights. The information sheet and the form of consent are available in appendix (10.1.1; 10.1.2).
18
The discussions were initiated with an explanation of the aim and purpose of the study and why the input from the participants was needed, as well as a brief introduction to the SVOD-industry. Later, a few situational questions were asked to make the participants feel comfortable and further promote conversation, before they were faced with questions regarding their habits, preferences, and feelings with regards to the SVOD-services included in this study.
3.5.2 Online Questionnaire for Quantitative Data Collection
In the quantitative stage of the research, data was collected by the use of an online questionnaire consisting of close ended questions. Questions of this nature can be helpful for increasing the response rate of the questionnaire, as they are quick for respondents to complete, and facilitates the process of coding and analysing (Rowley, 2014). Thus, this research adopted a structured questionnaire with close-ended questions for collecting quantitative data.
3.5.3 Questionnaire Development
The questionnaire development is an important activity in the research process since a poorly developed questionnaire can result in less valid and reliable data. Contrarily, a rigorously designed questionnaire may yield valuable information. Therefore, the questionnaire development used for collecting the quantitative data of this research was linked with relevant literature and guided through the steps presented by Brancato et al. (2006), in “Handbook of Recommended Practices for Questionnaire Development and Testing in the European Statistical System”. According to Brancato et al. (2006), there should be a consistent strategy that can guide the designing and testing of a questionnaire. In questionnaire design and testing, five mains steps need to be covered by the strategy. These steps include conceptualisation, design, testing, revision and data collection (Brancato et al., 2006).
19
3.5.4 Conceptualisation
The first step in designing a questionnaire is to specify the conceptual basis before considering the formulation of questions. In this phase, the objectives and concepts of the research, and the information needed to address it must be determined. In addition, specifying what is to be measured is essential for designing a survey instrument of high standard. Thus, basic concepts such as the target population, sampling, preferable data collection method, and the variables to be measured need to be clarified and defined (Brancato et al., 2006; Harlacher, 2016). Therefore, for this research, the phase included defining the research objectives, developing the conceptual model, and determining measurable variables by conducting an extensive review of relevant literature related to service quality, customer satisfaction, and brand loyalty.
3.5.5 Questionnaire Design
After the conceptual basis was defined, designing a first draft of the questionnaire including wording of concrete questions was formulated, and the process of structuring of the questionnaire took place. Based on the research design, data collection method requirements, and the questionnaire content, the questionnaire was organized into three sections, and is available in appendix (10.1.3).
In the first section, potential respondents were given an introduction to the questionnaire. This was followed by a few screening questions to determine whether the respondent qualified for participation. In addition, a few so-called “warmup questions” were used at the beginning of the survey to demonstrate the ease of completing the survey and generating the respondent’s interest (Burns et al., 2017). The warm-up questions concerned whether the respondent had any experience with SVOD-services, if they had access to an SVOD-service account, and which service they were most familiar with.
In the second section, a transition statement was given to make the respondent aware of a change in format. In this section, the intention was to capture the respondents’ attitudes, opinions, evaluations, impressions, perceptions, feelings, and intentions toward multiple items representing each underlying constructs. In order to do so, rating scale formats were adapted to allow respondents to translate their feelings and opinions in a clear and convenient way. A suitable tool for this purpose is using a Likert scale, which is an interval scale that can be used to capture the direction, and the degree to which respondents agree or disagree with various statements or questions (Burns et al., 2017). All questions of this type used a 5-point Likert scale, in which only the endpoints were labelled. The reasoning for choosing a 5-point scale rather than providing more options, is to obtain a response rate as high as
20
possible. A scale with more options or not being able to answer neutrally, may lead to the respondent becoming frustrated when answering the survey (Revilla, Saris, & Krosnick, 2013). It is also argued by Revilla et al. (2013) that a 5-point scale provides a higher mean quality than a 7-point scale.
The respondents were also given the alternative to answer the questions of this type with an “I do not know” option. This was included in order to facilitate the process of answering the questionnaire truthfully, however, was removed for the final question of each construct, since they measure the overall satisfaction connected to the respective construct. The measure will in this study be referred to as construct satisfaction. Denying the option to answer “I do not know” for these questions will, at minimum, result in receiving the respondents’ overall impression of each construct. This is important if the respondent chooses the “I do not know” option in several other questions from that construct. For some constructs, the labelling was changed to very satisfied to very dissatisfied, as it was better suited, while still measured on the same range scale.
In the third section, respondents were asked to indicate on a 5-point scale, their overall satisfaction regarding the chosen SVOD-service, as well as to what extent the service has met their expectations. An average of these measures was later used as a dependent variable, since a regression analysis cannot be performed with two dependent variables. This measure will be referred to as overall customer satisfaction. The Net Promoter Score (NPS) was also included in the third section and was used in order to measure the user’s loyalty intentions with their chosen SVOD-service.
Lastly, the end of the final section was reserved for a set of classification questions that were used for classifying respondents into groups based on gender, age, income, and subscription length. The reason for why these classification questions were placed at the end of the survey, was since respondents may have refused to answer them if they would have occurred in an earlier section, as they can be considered personal. Placing the classification questions in the final section, therefore, would allow any refusal to happen at the end of the survey process. This placement is also important as it could remove a potential negative tone towards the survey (Burns et al., 2017).
21
3.5.6 Questionnaire Translation
Since the target population consists of only Swedish individuals, aged 18 to 29, the questionnaire was tailored to suit the knowledge and understanding of the potential respondents in regard to language. Therefore, once all questions were developed, they were carefully translated into Swedish to eliminate potential language barriers and minimize the number of “I do not know” -answers, resulting from language capabilities. The translated version of the questionnaire is available in appendix (10.1.4).
3.5.7 Questionnaire Testing & Revision
When a questionnaire has been designed, a constituting problem remains that one cannot be entirely confident whether the respondents understand the questions, and answer them accurately (Cooper & Schindler, 2011). Before distributing the questionnaire, it is recommended to conduct a pre-test of the questionnaire on a set of representative respondents in order to discover any questionnaire errors. By testing the questionnaire in the early process of developing it, words, phrases, instructions, the flow of questions, or other questionnaire aspects that appear to be confusing, difficult to comprehend, or problematic to the respondent, can be identified and revised (Burns et al., 2017).
3.5.8 Quantitative Data Collection
For collecting the quantitative data, respondents were directed to the questionnaire page, generated through Qualtrics, which is a software for creating surveys and recording survey results. The respondents were presented with information about the study, such as the aim of the research, the included streaming services, as well as contact information to the researchers. The collected data was thereafter exported from Qualtrics to SPSS for further analysis.
3.6 Data Analysis Methods
3.6.1 Correlational Research Design & Multiple Regression
The correlational research design is a type of non-experimental research in which the statistical relationship between two or more variables are measured by using correlational statistics (Creswell, 2013). The reason for applying a correlational research design is to determine if two or more variables of interest are related. The statistical analysis of the correlation can be analysed through a multiple regression analysis, which is a useful tool for examining what factors are related to the dependent variable, as well as how the dependent variable is influenced by each of the factors, and to what
22
degree it is influenced (Burns et al., 2017). Thereby, it is possible to inspect the strength of a linear relationship between variables of interest. Multiple regression analysis was therefore considered as an appropriate statistical method for analysing the relation between service quality, customer satisfaction, and brand loyalty.
This was done by first conducting a multiple regression analysis for each of the seven constructs of service quality, in order to determine the amount of variation of each construct that could be explained by all the items within the constructs explained together. In each of the seven multiple regression models, the items underlying its construct were considered as independent variables, whereas the measure of construct satisfaction was considered as the dependent variable.
Four additional multiple regression models were then conducted in order to test the research hypotheses stated earlier. In the first of these four regression models, all measures of construct satisfaction were considered as independent variables, while the overall customer satisfaction was considered as the dependent variable. This was done in order to test hypothesis 1, and determine if any service quality constructs were significantly, positively associated with overall customer satisfaction. In the second test, the independent variables were unaltered, however, brand loyalty was considered as the dependent variable. This was done in order to test hypothesis 2, and determine if service quality was positively associated with brand loyalty. This was followed by a third regression model to test hypothesis 3, and examine whether overall customer satisfaction was positively associated with brand loyalty. Therefore, overall customer satisfaction was considered as an independent variable, whereas brand loyalty was considered as the dependent variable. In the fourth and last regression model, all measures of construct satisfaction, together with overall customer satisfaction were considered as independent variables, whereas brand loyalty was considered as the dependent variable. This procedure allowed the researchers to test hypothesis 4, and examine the relation between service quality, overall customer satisfaction, and brand loyalty.
3.6.2 Multiple R & Adjusted R2
A useful measure for determining the strength of a linear relationship between variables is the coefficient of determination (R2), also known as multiple R. It ranges from 0 to 1 and refers to the
amount of variation in the dependent variable that can be explained by all the dimensions of the independent variable(s). The higher the fraction, the more the dependent variable can be explained by the combined dimensions of the independent variables (Burns, et al., 2017). However, in sampling distribution, multiple R has a tendency to overestimate the true value in the population as the sample size decreases (Tabachnick & Fidell, 2007). In order to provide a better estimate of the true
23
population, adjustments must be made to the multiple R. Hence, the adjusted R square is a modified version of the multiple R which provides an adjustment of this value. Therefore, with respect to the relatively small sample size used in this research, the adjusted R square may be a better indicator to report, concerning the amount of variation in the dependent variable associated or explained by all the independent variables explained together.
3.6.5 Standardized Beta Coefficient
If the sample correlation is statistically significant, the strength of the relationship can be determined based on the size of the correlation, and therefore also the importance of the independent variables with regard to each other. However, as these variables are regularly measured by the use of different units, they must therefore be standardised by normalising the independent variables through obtaining a standardised beta coefficient. A larger absolute value of a standardised beta coefficient indicates a greater relative importance of the value when considering the dependent variable (Burns et al., 2017). A negative number indicates a negative correlation, which entails that investigated variables move in opposite directions. Thus, correlation only assumes a linear relationship between the variables and does not imply causation (Simon & Goes, 2012).
3.6.3 Multicollinearity
Multicollinearity is another important measure to examine as it refers to the existence of either moderate or strong correlations among the independent variables. If multicollinearity is present, the results of the independence assumption of multiple regression analysis will be negatively affected. Therefore, testing and potentially removing multicollinearity is crucial. A way to remove multicollinearity is by using the variance inflation factor (VIF), which determines if multicollinearity is present. VIF is a single number which is generally considered to reject multicollinearity if the number is lower than 10, and therefore if the VIF of a variable is greater than 10, it should be removed and the multiple regression test must be revisited (Burns et al., 2017).
3.7.1 Reliability
When selecting scales to be included in a study, one concern is to test the internal consistency, or reliability of each construct. That is, the extent to which items underlying a construct, are measuring the same construct. In order to measure the reliability of each construct, the Cronbach’s alpha test was run in the analysis. This is a common indicator for internal consistency were a Cronbach’s alpha value equal to or above .7 is required for establishing reliability (Pallant, 2007).
24
3.7.2 Construct Validity
According to Peter and Churchill (1986), the interpretation of the calculated relationship and its strength between latent variables is only relevant if construct validity can be demonstrated. Because construct validity is an assessment of how well a study’s measurement model measures what is it supposed to measure, researchers must apply various subcategories of construct validity, including convergent & discriminant validity, in order to assess study results (Henseler, Ringle, & Sarstedt, 2014).
3.7.3 Convergent Validity
Convergent validity pertains the extent to which measures that theoretically should be measuring the same construct, are in fact more correlated with each other than with measures of other constructs. Evidence of correspondence between measures of constructs can be established by conducting a bivariate correlation test, where the correlation coefficient (represented by the letter r for Pearson’s correlation) can be interpreted as representing an effect size, which is an indicator of the correlation strength between two variables. A correlation coefficient value of approximately .5 is an indicator for evidence of a strong correlation, whereas a value of approximately .3 indicates a moderate correlation, and a value of approximately .2 indicates a weak correlation (Swank & Mullen, 2017). Hence, this assumes that the greater the correlation coefficient between measures of the construct, the better the evidence of correlation. On the other hand, if two items are completely correlated, they would have a correlation coefficient of 1, which would suggest that the two questions have been answered identically. This is not an optimal result, as one question could be seen as overabundant if there are no changes to Cronbach’s alpha after removing one of the questions.
It is, however, essential that the bivariate correlation coefficient is statistically significant, or meaningful. Otherwise, no evidence of convergent validity can be established. To test for statistical significance, the p-value is examined and compared to the level of significance for the analysis. In addition, the hypothesis related to the correlational analysis needs to be tested (Swank & Mullen, 2017).
25
3.7.4 Discriminant Validity
Contrary to convergent validity, discriminant validity reflects the extent to which measures of constructs that theoretically are supposed to be dissimilar, are in fact observed to have a low or weak correlation between one another. To establish evidence of discriminant validity, a common norm is to seek for low correlation between measures, or items that are not supposed to measure the same construct (Swank & Mullen, 2017; Henseler, Ringle & Sarstedt, 2014). As with convergent validity, a bivariate correlation test was conducted in order to establish evidence of discriminant validity.
3.8 Ethical Considerations
In any research study, it is important to implement appropriate ethical principles in order to provide protection for human subjects (Orb, Eisenhauer, & Wynaden, 2001). In this study, potential participants of the focus groups were approached individually and were adequately informed about the research purpose, the research process, and the methods of data collection. It was explained that the participants had the freedom of choice to participate, decline, or withdraw from the study at any time during the process. After a thorough explanation of the research, potential participants were provided with an information sheet concerning the study. Participants were given the appropriate time to carefully read through the study purpose, think of any questions or concerns, and to decide whether they would like to participate or not.
The identities of the participants were not disclosed in any section of the study, thus the anonymity and confidentiality of the participants was maintained. Also, privacy and confidentiality of the focus group environment was managed cautiously over the course of the focus group sessions, data analysis, and dissemination of the research findings. The same applied for the online survey, where respondents were provided with a description of the study, an indication of what kind of questions would be asked, that respondents had the opportunity to decline or withdraw from the online survey at any time, and that confidentiality and anonymity of the respondents was reassured.
