The digital entertainment industry:
Motivators and barriers in the shift to digital goods
Master Thesis
Suzie Lednor
5875943
Msc Business Administration
Track: Entrepreneurship and Management in the Creative Industries
Supervisor: Dr. J.J. Ebbers
2nd reader: Dr. F.B. Situmeang
2 Statement of originality
This document is written by Suzanne Inez Lednor, who declares to take full responsibility for the contents of this document. I declare that the text and the work presented is original and that no sources other than those mentioned in the text and its references have been used creating it
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
3 Abstract
In today’s world, many products are increasingly becoming available in a digital form. Entertainments goods, such as books, games and music, can be bought in both a physical and digital manifestation, with no difference in content. The aim of this research is to identify fundamental predictors of digital purchasing behaviour. A review of the current academic status of factors contributing to online and offline purchasing identify experience, product risk, financial risk, age, convenience and service to be of influence for online and offline shopping. No resreach has been done yet investigating the drivers for purchasing digital products. These factors are re-tested for offline and online purchases, as well as their influence on digital purchases. For digital purchases specifically perceived usefulness and ease of use are also tested, following the Technology Acceptance Model. An online survey was posted via a social media channel of a very large video game developer and distributor. This led to 160 usable responses. Negative binomial regression analysis was conducted in order to test 17 hypotheses. Experience was found to be the most
influential predictor in all three settings. Service was confirmed to be positively related to physical purchases. Age was found to be positively related, contrary to the hypothesis, to digital purchases. Besides the confirmation and rejection of the hypotheses, certain assumed relations based on the literature were not found. This indicates that the established predictors of purchasing decisions in both an offline and online setting are possibly out-dated, and need to be re-identified. This research brings a unique contribution to the academic world as being one of the first to investigate the factors of influence for purchasing digital goods, and the effect this has on more established purchasing channels. Future research should follow this direction, identifying the factors of influence for digital purchases, and re-testing the factors of influence for offline and online purchases.
4 Table of Contents
1. Introduction 5
2. Literature review 9
2.1 Frameworks: TRA and TAM 9
2.2 Motivators and barriers 9
2.2.1 Experience 11 2.2.2 Risk perception 12 2.2.3 Age 13 2.2.4 Convenience 14 2.2.5 Service 16 2.3 Conceptual model 17 3. Research Design 19 3.1 Research Setting 19
3.2 Consoles and games 19
3.3 Operationalization 20 3.3.1 Dependent variables 20 3.3.2 Independent variables 20 3.3.3 Control variables 23 3.4 Data collection 23 4. Results 25 4.1 Descriptives 25 4.2 Correlations 27 4.3 Regression analysis 29 4.4 Robustness checks 31 5. Discussion 33
5.1 Hypotheses and implications 33
5.2 Limitations 35
5.3 Future research 36
6. Conclusion 38
Bibliography 29
5 1. Introduction
'Ecommerce sales in Europe will increase by 18.4% in 2015'
This quote illustrates the vastness and rapid growth of the online market (Internationale E-commerce studie, 2015). In The Netherlands alone, online revenues surpassed €6 billion in 2014, and was expected to grow almost 17% in 2015. At the same time, offline revenues decreased by 1.4% in 2014 (Internationale E-commerce-studie 2015, 2016). The channel-purchasing decision of consumers will depend on what they value the most. In 2004, Rohm and Swaminathan started identifying different types of online shoppers (Rohm & Swaminathan, 2004). Measuring six different construct, they identified four typical online shoppers: the convenience shopper, variety seeker, balanced buyer and offline shopper. These four different types were driven by different motivators, which influenced why they preferred to shop either offline or online. As time and the online shopping environment progressed, more and more factors came into play. A further developed overview was created by Zhou et al (2007) called the ‘Online Shopping Acceptance Model’. Building on the growing amount of literature, they reviewed a variety of variables and the supporting evidence for many constructs such as gender, age, normative beliefs, shopping
orientation, shopping motivation and web apprehensiveness (Zhou, Dai & Zhang, 2007). This showed a new set of characteristics for online shoppers, and a different take on existing types. For instance, internet experience was found in earlier researches to be of significant influence on online shopping behaviour, while later research found that internet experience ‘did not necessarily
influence their online shopping behaviour’ (Zhou, Dai & Zhang, 2007, p. 47). One of the explanations given for this is that as the internet becomes more ingrained in our daily lives, everyone gains experience and thus it is no longer an adequate predictor for online shopping behaviour. Illustrated by the quote at the head of this section, more and more people are shopping online, and are apparently valuing the benefits associated with online shopping. This raises the question, what drives consumers today to shop online? Besides that, there a more developments in the realm of online shopping, which should get attention from the academic community. The technological developments in recent years have not only transformed the online market, but also created a new alternative.
In addition to the digitization of the shopping process, some goods have also become
available in digital form. Products no longer need to have a physical manifestation for people to buy or own them. Books can be read on an e-reader, music can be downloaded as an mp3, movies and shows can be watched online, newspapers can be read on your tablet. The costs of these digital goods are usually much lower than the cost of the physical goods. However, the cost of the
6 hardware needed to store and play some of the goods can be quite high, and also can come with certain limitations. In short, each type of product has its advantages and disadvantages, and these will have a different effect on different people. So far, hardly any consumer research has been done concerning digital entertainment goods.
A short clarification of the options and how they are related might be useful in order to explain why this is of interest. When a digitized alternative is available to a consumer, the channel through which consumers can buy their products can vary in two ways: offline or online. The product can be bought as a physical or digital good. This results in four different possible settings for consumers. You can buy a video game in a toy store (an offline purchase of a physical good), order the game online through an online retailer such as Bol.com (online purchase of a physical good) or download the digital game via the digital library of the particular console store, like Xbox Live (online purchase of digital product). It can be argued that there is also a category of offline purchasing of digital products, as for instance with a gift-certificate for Xbox Live. However, these products are most likely to be bought as gifts, and not for the purchasing consumer, and thus the rationale is different. Figure 1 shows a two-by-two matrix of the possibilities, with video games taken as an example.
A big part of of online shopping is related to the perceived risks. Different kinds of risk have been established in having an effect on online shopping (Gupta, Su & Walter, 2004; Zhou et al, 2007; Dai, Forsythe & Kwon, 2014). Most researchers also find that risk perception goes down as experience increases (Zhou, Dai & Zhang 2007; Dai, Forsythe & Kwon, 2014). As was found by Zhou et al (2007) in relation to internet experience, the normalcy of a channel can lead to the dissolution of a predictive effect. The online revenue continues to grow (rapidly), and this alone signifies both an academic and practical incentive to continuously test for the motivators and perceived risks related to shopping behaviour. As is true with online shopping, the perception of risk and benefits of digitized products may change over time. As more products become available in a digital form, it is of interest to both the academic community and the marketing departments of organization to know if, and what, motivators have an effect when buying a product that is available in three forms.
The aim of this research is two-folded. On the one hand, it seeks to establish the motivators and barriers involved with the purchasing of digitized goods. These motivators, such as age, experience, risks and convenience are derived from existing literature regarding purchasing
behaviour in offline and online settings (Rohm & Swaminathan, 2004; Zhou et al, 2007, Kollman et al, 2012; Dai, Forsythe & Kwon, 2014) . Besides these concepts, a framework designed especially to predict human behaviour when it comes to new technology is tested in this setting. The
7 that are involved with the adoption of a new technology. Specifically, ease of use and perceived usefulness should have a positive relation to the amount of purchased digitized products. As a new technology comes up, the relation between ease of use and perceived usefulness should be
demonstrated if the TAM continues to be an effective ‘technology acceptance model’.
Purchasing channel
Type of product
Online Offline
Digital Digital download
through Xbox Live X Physical Physical video game
through bol.com
Physical video game at toy store
Figure 1 - Purchasing options for consumers when including digital entertainment goods
Besides establishing a model for purchasing digitized products, the effect of a new alternative on older established models is also scrutinized. In a transforming market it cannot be assumed that established relationships stay the same when a new alternative is added to the choices for consumers. Thus, this research brings an unique contribution by comparing the effect of a new alternative on previously found relationships. In short, the question is not only what motivators are associated with buying digitized products, but also what effect this has on previously established relationships between motivators and offline and online shopping. Currently, no other research makes a comparison between these three settings. This research provides a valuable contribution to consumer research by determining what constructs play an important role when buying digitized products, and the effect of this alternative on older channels.
As this study shows, the rapid transformation and digitization of markets and goods changes the way people look at things, and the risk and benefits they perceive. The practical implications of this are even more challenging. Designers and marketeers must also be able to adapt to the evolving perceptions in order to keep up with the wishes of the consumers.
This research is structured as followed: In chapter 2, the review of the academic literature regarding motivators and barriers in offline and online shopping are discussed. The research design and operationalization of the concepts are discussed in chapter 3. Chapter 4 will present the results of the research, and chapter 5 will discuss these results and the implications of the findings. The
8 limitations and directions for future research will also be presented. Chapter 6 will give a small summary and conclusion of the research.
9 2. Literature review
Much research has been done regarding factors that influence consumers’ purchasing decisions, both in an offline and online setting. This review will focus on the research that has been done on variables that influence either settings, and then try to predict the effect of that variable on the purchasing process of digitized products. This serves not only the purpose of identifying the factors of influence when purchasing digitized products, but will also re-test if previously identified factors are still of significant influence. The factors contributing to physical purchases may change as more alternatives arise, and as the novelty of online shopping wears off the perceived risks associated with online shopping might change. Attitudes and behaviours are likely to change when a technology becomes more mainstream. Many people have been documented stating that the cell phone was an invention that would not become very popular... Nowadays most people can't live without one! Before looking at specific motivators and barriers when making a purchasing decision, it makes sense to first look at the human behaviour regarding decision making and new technology.
2.1 Frameworks: TRA and TAM
There are two general frameworks that have been well established in the academic literature regarding human decision making, and the emergence of a new technology. With regards to future decision making Ajzen and Fishbein’s Theory of Reasoned Action (TRA) is one of the most well-known (Ajzen and Fishbein, 1975; Madden et al, 1992). This theory is used to predict and explain behaviour, and designed to be applicable to virtually any human behaviour (Chen, Gillenson and Sherrell, 2002). It focuses on user’s attitudes, intentions, social norms and actual behaviour (Davis, Bagozzi and Warshaw, 1989; Godin, 2003). It has been tested in different settings, ranging from abortion to choosing political candidates (Ajzen and Madden, 1986). It also gained support as it was the first theory to find a strong link between verbal attitudes and actual behaviour (Ajzen, 2012).
The main take away of the theory for this research is that measuring ones attitudes towards something can predict behavioural intention. The behavioural intention can then be used as a valid indicator to predict the behaviour that will take place in the future. As Madden et al found in their 1992 research, when the behaviour is under volitional control, the theory of reasoned action is most suitable for predicting behaviour (compared to another famous theory of Ajzen, the theory of planned behaviour) (Madden et al, 1992). When trying to figure out what contributes to a certain behaviour, it is therefore justified to use people's own prediction of their future behaviour as a directory. When consumers speak of their intentions to engage in certain types of shopping behaviour in the future, it can be assumed that this is a decent predictor of the consumer's future behaviour.
10 Using the Theory of Reasoned Action, Davis created the “Technology Acceptance Model” (TAM), which ‘is an adaption of TRA specifically tailored for modelling user acceptance of information systems’ (Davis, 1986; Davis, Bagozzi and Warshaw, 1989). The TAM combines perceived usefulness and ease of use to predict to what extent a new technology gets adopted (Davis, Bagozzi and Warshaw, 1989; Chen, Gillenson and Sherrell, 2002, Zhou, Dai and Zhang, 2007). The research of Davis, Bagozzi and Warshaw (1989) found three things: ‘people’s computer use can be predicted reasonably well from their intentions, perceived usefulness is a major
determinant of people's intention to use computers and ease of use is a significant secondary determinant’ (Davis, Bagozzi and Warshaw, 1989). The usefulness of TAM is not only limited to computer use, but can be applicable to almost any information or electronic system. For instance, the TAM has been used for a variety of things, including predicting mobile learning by university students (Park, Nam and Cha, 2012) and the intended usage of health care systems (Pai and Huang, 2011).
As stated, the two main components are ease of use and perceived usefulness. Ease of use measures to what extent a person thinks the new technology would be easy to master. In the research by Davis, Bagozzi and Warshaw (1989), they investigated MBA's students attitude and usage of a word processing software, WriteOne (which was fairly new at that time). The question involved where such as 'learning how to operate WriteOne would be easy for me' and 'I would find it easy to get WriteOne to do what I want' in order to determine the skill mastery involved.
Perceived usefulness determines whether mastering the new skill (regardless of the effort involved) would actually be an improvement when compared to the available alternatives. These two factors influence a person's attitude towards a certain computer system. As noted in the previous paragraph, based on the 'theory of reasoned action' research by Ajzen and Fishbein, one's attitude is essential in predicting one's behaviour.
When the purchasing of digitized products increases it would be of value to measure people's perceived usefulness and the ease of use associated with the product. Or put the other way around, does an increase in perceived usefulness and ease of use lead to an increase in the amount of digital purchases thought to happen in the future? This leads to the first two hypotheses:
H1: There is a positive relationship between an individual's perceived usefulness of buying digitized products and the expected amount of digitized products purchased.
H2: There is a positive relationship between an individual's perceived ease of use and the expected amount of digitized products purchased.
11 The TAM is often used as a basis in research, but is also limited by the small number of variables (Vijayasarathy 2004; Ha and Stoel, 2009). Therefore further exploration of constructs is needed, based on the current state of the academic literature. Possibly these variables cause an increase in offline shopping, and cause the equal decrease in online shopping. The constructs selected are based on their influence on either an offline or online context. Based on the amount of similarities between the offline and online context, it is then hypothesized what the influence of the variable will be on the purchasing of digitized goods. With internet and digital products becoming more mainstream, it is feasible that the motivators to buy products in a certain way can change over time. This research will also test whether the most common variables associated with offline and online shopping still are of influence.
2.2 Motivators and barriers
As stated above there can be many factors influencing a purchasing decision. Not much research has been done regarding the switching tendencies of consumers from physical to digital goods, therefore it is a good idea to start with the most fundamental concepts. Experience is taken as a starting point, and leads to different concepts that can serve as either a motivator or a barrier towards digital purchases. As online shopping evolves and new alternatives arise, the relationship between an established concept and an online and physical setting might change. When a relation is assumed a hypothesis will be formulated in order to re-test a construct, and to find the influence of that construct on digital purchases.
2.2.1 Experience
Experience, in this instance, refers to the numbers of occurrences one has had with the different purchasing options, and not necessarily the quality of the occurrence. In general, people like to stick to what they know when they have found a way to do things that suits them. In a 2002 study, Mauldin and Arunachalam found that comfort with the internet was of significant influence on the level of intent to purchase something online (Mauldin and Arunachalam, 2002). Bhatnagar and Ghose (2004) found that as internet experience went up, product risk perception went down, which could lead to an increase in purchases. Forsythe et al (2006) stated that a one-time visitor will turn into repeat customer if the experience of online shopping delivers value. Zhou et al (2007) found more research both confirming and disputing the effect of experience. They also stated that when experience reaches a certain level, it is possible that it loses its predictive power. In recent years, not much more research has been done regarding the relationship between experience and online shopping. As the online market continues to grow it is valid to retest the relationship that have previously been confirmed. Since digitized products are still fairly new, the hypothesis will
12 assume that experience is a strong indicator for repeat purchases. As experience goes up, the
amount of digitized products will go up. Because online shopping has been around for quite some time, it is hypothesized that this relationship will be less strong than the relationship between experience and amount of digitized products purchases, but still significant. By retesting the relationship between experience and online shopping, it will provide a new insight towards the claim of Zhou et al (2007) that experience becomes less or not significant over time, due to the ingrained practice in daily life. Due to this reason, no significant relationship is assumed between offline shopping and experience. For online shopping, a significant relationship is still assumed, since most research did point in that direction (Mauldin and Arunachalam, 202; Bhatnagar and Ghose 2004; Zhou et al, 2007).
H3a: There is a positive relationship between experience and the expected amount of digitized products purchased.
H3b: There is a positive relationship between experience and the expected amount of online products purchased
H3c: The relationship between experience and expected digitized products purchased will be stronger (more positive) than for experience and expected online products purchased.
2.2.2 Risk perceptions
The relation between experience and purchasing behaviour seems to be tied in with certain risk-perceptions (Zhou et al, 2007) . When investigating risk-perception several distinctions could be made. In a research regarding channel-switching tendencies from offline to online purchases, Gupta, Su and Walter (2004) found that channel-risk perceptions were negatively related to channel-switching tendencies. As the perceived risk increases, people become less likely to move away of the current and adequate mode.
In more recent research, Dai, Forsythe and Kwon (2014) differentiated between three different types: product risk, financial risk and privacy risk. Product risk and financial risk have been found to also have an effect in an offline shopping context. Dai, Forsythe and Kwon (2014) were among the first researchers to differentiate between products, in this case apparel and digital music. For digital music, product and financial risk had a significant effect. Also, the authors recommend that risk perception should be re-tested in other digital categories. This research will take up that recommendation.
Al-Rawad et al (2015) also found that financial risk and functional risk (which is the same as product risk) were of influence on online shopping behaviour. Both Al-Rawad et al (2015) and Dai et al (2014) found that previous online shopping experience reduced the risk perception. It can
13 be hypothesized that the more one gets acquainted with a new technology, the risk perception will go down (unless the end result is inherently risky as with certain medicinal or food products). The evidence for the connection between risk perception and shopping behaviour is strong. What is to be tested in this research is firstly whether or not both types of risk perception are also a significant influencer when purchasing digitized products, since this is a new type of shopping. Since the research testing the relationship between risk perceptions and online shopping was still recent and unambiguous, the relationship is assumed to be valid. The two main risks perceived when shopping online are product risk and financial risk. Since purchasing digital goods is a new phenomenon, it is hypothesized that there will be a bigger perception of risk. Therefore, relations hypothesized are as follows:
H4a: There is a negative relationship between product risk perception and the expected amount of digitized products purchased.
H4b: The relationship between product risk perception and the expected amount of digitized products purchased will be stronger (more negative) compared to product risk perception and the expected amount of online products purchased.
H5a: There is a negative relationship between financial risk perception and the expected amount of digitized products purchased.
H5b: The relationship between financial risk perception and the expected amount of digitized products purchased will be stronger (more negative) compared to product risk perception and the expected amount of online products purchased.
2.2.3 Age
Inevitably, experience is also related to age. In the earlier research of Rohm and
Swaminathan in 2004, they found no significant effect for age (Rohm and Swaminathan, 2004). However, when Zhou et al (2007) were constructing their ‘Online Shopping Acceptance Model, they reported mixed findings with respect to age (Zhou, Dai & Zhang, 2007). More recently, in the research of Kirk et al, the acceptance and adoption of a new technology was investigated, with digital immigrants and digital natives being a factor (Kirk, Chiagouris and Gopalakrishna, 2012).
Digital immigrants are born before 1980, and digital natives after 1980, meaning they grew up with internet and all the technology surrounding it. In the research of Kirk et al (2012), they compared consumers’ satisfaction with regard to digitized products. Their first conclusion was that age did not moderate the level of acceptance of this technology. All respondents were more likely to accept the interactive book than the static book. However, they did find that the expected
14 satisfaction of digital immigrants was not increased. Lian and Yen (2014) specifically researched the possibilities of the growing group of e-consumers in a higher age category. One of their conclusions was that older people (over 50 in this case) still perceive more barriers towards online shopping compared to younger people (Lian and Yen, 2014). This would mean that the age group who purchased digital products would be significantly younger.
Following from these results it can be hypothesized that even though older consumers might still be a little hesitant towards online shopping and digitized consumption, the group is growing. It remains difficult to pinpoint the age where the acceptance and usage of new technologies and products become stronger. It seem plausible that the relationship between age and offline shopping grows stronger over time, meaning the older people get, the more likely it is they will shop offline. In a reverse effect, it can be hypothesized that younger people are more likely to shop online. Combining the results of Kirk et al (2012) and Lian and Yen (2014), the relationship between age and digitized products can be hypothesized to follow the same path as age and online shopping. Younger people will value the benefits of digitized products more and will see less barriers, whereas older people will not experience increased benefits but more barriers, and will therefore rely more on physical goods and offline shopping. When looking at age there seems no reason to assume that there will be a difference for online and digitized purchases. Therefore, the following hypotheses are set up:
H6a: There is a negative relationship between age and expected digitized products purchased.
H6b: There a positive relationship between age and expected physical products purchased.
2.2.4 Convenience
In the research of Rohm and Swaminathan (2004), the online shopping types that emerged were convenience shoppers, variety seekers, balanced buyers and store-oriented shoppers. When looking at what factors could be extracted to the "typology of digitized products shoppers" convenience was most appropriate. Variety seeking was linked to the possible purchasing of a variety of products. This is not applicable to digitized products, as the product is always within one category. To clarify, in the instance of purchasing groceries you can buy food, medicine and
cleaning supplies. When buying a digitized product there is no variety in the category of things you buy, or a difference of the content. An e-book is still the same book whether you buy it for an Amazon Kindle or another brand. You cannot use the same e-store to buy different products, such as using Xbox Live to buy e-books. Another type in Rohm and Swaminathan’s research was the store-oriented shopper. A store-oriented shopper prefers the act of walking around in a store and the
15 social interaction accompanied with this. This component will be discussed in section 2.2.5, albeit in a slightly different interpretation. The balanced buyer was attracted equally by variety seeking, convenience and store orientation, with no strong preference. While this could be a possible conclusion of this research, no hypothesis about this was made from this notion. Convenience was the type that was further explored.
Convenience was mostly seen as time and effort saving by Rohm and Swaminathan. This resulted in a stronger preference for online shopping compared to offline shopping. Chaing and Dholakia (2003) also found that time saving was a significant driver to shopping online. Kollman et al (2012) reaffirmed that when consumers value convenience, they would be more likely to look for, and purchase, products online. A distinction was often made between time and effort saving, and immediate possession. While Rohm and Swaminathan (2004) found that immediate possession was a motivator for selecting offline shopping over online shopping, Beauchamp and Ponder (2010) did not find this to be true six years later. In their research, online shoppers had a higher perception of four different types of convenience compared to offline shoppers. This also included possession convenience (as well as acces, research and transaction conevenience), although an adequate explanation was missing as to why. This research will retest the relationship between convenience and offline and online shopping, in order to gain more clarity and evidence regarding Beauchamp and Ponders' findings. All previous researches demonstrated that online shopping was seen as more convenient than offline shopping. This results in the following two hypothesis for offline and online shopping:
H7a: There is a negative relationship between convenience and the expected amount of physical products purchased.
H7b: There is a positive relationship between convenience and the expecte amount of online products purchased.
In the case of purchasing digitized goods, time and effort saving and immediate possession are combined. Time and effort is diminished as the consumer can look for the product online, while immediate consumption is also available, with a negligible download time for most products. Logically it could be stated that it combines best of both worlds. Therefore the relation between the valuation of convenience and the purchasing of digitized goods is thought to be positive.
H7c: There is a positive relationship between the valuation of convenience and purchasing digitized products in the future.
16
products will be stronger (more positive) compared to the relation between the valuation of convenience and purchasing digitized products online.
2.2.5 Service
An important element in all shopping endeavours is the service provided by the shop. This type of social interaction is likely one of the attractions for the store-oriented shoppers described by Rohm and Swaminathan (2004) mentioned in the previous section. As Kollman et al (2012) state, the costs and benefits that lead to motivational factors with regards to shopping behaviour are associated with convenience, risk and service. Their research found that the desire for service has the potential to ‘cannibalize’ customer away from online shopping, more so than risk aversion (Kollman et al, 2012). The risk associated with online shopping has dropped over time, the request for online store service has remained as strong.
Motaya-Weiss, Voss and Grewal (2003) found evidence that as the perception of service quality in online services rises, the amount that the online channel was used also rises. Accordingly, when the perception of quality went down, the usage of the online channel was also diminished. The same effect was seen as the service perception of an alternative medium went up. As they concluded, an increase in online activity could be stimulated by diminishing the quality of service in an offline setting. In order to put this conclusion into practice on must first know how the service is perceived in one’s business. This forms another basis for measuring the perception of service in relation to different the different purchasing options available.
Ha and Stoel (2009) also found that customer service has a direct impact on online shopping intention. While websites are getting better at providing more service online, such as zoom
possibilities on the products and chats with store employees, the most and easiest interaction can be found in an offline setting. For digital products no such help exists. The effort to find product information on a digital product online is the same as for finding physical products online.
It also stands to reason that people can mix different purchasing channels when searching for a product. Pauwels et al (2011) investigated the effect of a store’s website introduction and the revenue of offline sales. They found that in the long run having a website does not increase offline revenues. Apparently people's online search does not directly translate into an increase in offline purchases for the same store. It would seem likely that consumers wishing to get help on their purchasing decision will chose to purchase their product in an offline setting, and people who search online buy their product online. The valuation of service could therefore be an important factor in deciding which purchasing channels to choose. A low valuation (or need) of service will then result in more online and digital purchases, and a higher valuation of service would result into more physical product purchased. As no extra service is provided when buying digitized products
17 compared to shopping online, there is no reason to assume that the relationship would be stronger in either direction. This translates to the next hypotheses:
H8a: There is a positive relationship between service and physical products purchased. H8b: There is a negative relationship between service and digital products purchased.
2.3 Conceptual Model
Many more concepts can be found to be of influence in a consumer's purchasing decision. The main aim of this research is to find the factors that drive or deter people from purchasing digital goods. These constructs were, as seen above, retrieved from established, academic research.
However, the validity of these constructs will also be tested by measuring once again their
(established) influence in a physical and online setting. Because all constructs will be tested within the three different settings, this research is limited to the aforementioned constructs of experience, product risk, financial risk, age, and convenience and service. In the case of purchasing digital goods, perceived usefulness and ease of use have been added, following the Technology
Acceptance Model. The conceptual model gives a schematic overview of the hypothesized direct relationships to each setting. In order to keep the model as clear as possible, the outcome variable is named ‘amount of products bought’. The different colours of the lines indicate which setting it is related to. All blue lines indicate the relationship between a construct and the expected amount of games bought in a digital setting. All red lines indicate the relationship between a construct and the expected amount of games bought in an online setting. All purple lines indicate the relationship between a construct and the amount of physical purchases. Besides the direct hypotheses that are reflected in the conceptual model, there are also 4 hypotheses regarding a ranking of the strength between a variable and multiple settings, which can not be found in the model.
18
19 3 Research Design
3.1 Research setting
The research was conducted with the help of one of the world’s largest video game producer and distributor. The video game industry is an adequate and appropriate field for this research for various reasons. First of all, it is a huge industry, which came up quite rapidly. Video games became popular in the 1970’s, and by 2004 the video game revenues overtook box office receipts (Alpert, 2007). More Americans now play video games than go to the movies (Marchand and Hennig Thurau, 2013). Another interesting feature of the video game industry is that, as opposed to the movie and music industry, it has much less had to deal with illegal pirating of digital versions. Although it somewhat exists, it does not have the same size as in other entertainment industries (Marchand and Hennig Thurau, 2013). Lastly, there is a major trend to be found in the shift from physical to digital games being purchased. From 2009 to 2011 ‘the ratio of physically distributed games dropped from 80% to 69% and digitally distributed games rose accordingly’ (Marchand and Hennig Thurau, 2013). This combination of factors make the video game industry an ideal setting to test the motivators of individuals when facing the choice of buying a physical or digital product.
3.2 Consoles and games
This research focuses on console video games. A console is a type of hardware that is
needed to store and play video games. The most sold consoles are the PlayStations by Sony, and the Xbox by Microsoft. Approximately every 6 years a new ‘generation’ of consoles is released. The older generation consoles that still have a big installed base in the Benelux are the Xbox-360 (released in 2005) and the PlayStation 3 (released 2007 in Europe). Currently there are 1.9 million PlayStation 3 and 1.1 million Xbox-360 (per September 2015). The next generation consists of the PlayStation 4 and the Xbox One. The distribution is very much skewed, with 574.000 PlayStation 4 and 99.000 Xbox One (per September 2015). This was in part caused by a worldwide shortage of the new Xbox console and a delayed introduction to the Benelux market, while the PlayStaion 4 was released at roughly the same time. The main difference between the two generations is the storage capacity and the speed. The first PlayStation 3 had a storage capacity of 20 GB, the PlayStation 4 has a capacity of 500 GB. The size of video games can vary quite a bit, but are between 25 and 50 GB. The average price of a PlayStation 4 in the Benelux in 2015 was €427, for the Xbox One this was €421 (measured August 2015).
Physical video games are CDs that can be inserted into, saved to and played on or via the console. Digital games are downloaded from a virtual cloud, and then played via the console. The technical specifications of digital and physical games are beyond the scope of this research, and are
20 also irrelevant here. PC games are excluded from this research due to the vast differences in
technical capabilities. Comparing the two really is like comparing apples and pears, and much discussion can be found on different websites concerning the 'superiority' of PC or console game(r)s. The focus on console over PC was in part influenced by the knowledge already in
possession of the video game distributor, meaning they had quite a bit of information regarding the behaviour of PC gamers, but not yet of console gamers. The other part is that a console is bought specifically to play games, and while it can be used for other things it is not as varied in its use as a PC.
Most video games are available on all types of consoles. Sometimes a distributor makes a deal with a console producer (for instance Microsoft) to release a game exclusively on one platform (“Xbox exclusive titles”). Newer games are sometimes only released for the latest generation of consoles.
3.3 Operationalization 3.3.1 Dependent variables
This research has three different dependent variables. The first dependent variable measures how many video games respondents expect to buy in the next 6 months in a physical store, i.e. a CD in a toy store. The second dependent variable measures how many video games a respondents expects to buy in the next 6 months via an online store, i.e. bol.com. The third dependent variable is related to the purchasing of digitized product. It measures how many games a respondent expects to buy via Xbox Live or PlayStation Network in the next 6 months. The amount of games could be indicated by filling in any number, with no restrictions to a minimum or maximum.
3.3.2 Independent variables
The independent variables tested in all three models were experience, product risk, financial risk, age, convenience and service. Age was simply measured by asking their age, with no
classification of age groups. Experience was measured by asking the respondent how many games he/she had bought in the last 6 months in all respective settings.
Product risk, financial risk and convenience were all measured in relation to all settings. The respondent got three “blocks” of questions. The first block was regarding the purchasing of physical products. This meant that all question started with ‘When I’m purchasing a video game in a
physical store…..’. In the second block, all questions started with ‘When I’m purchasing a video game via an online retailer…’. The third block began all question with ‘When I’m purchasing a video game via PSN or Xbox Live….’. The question was then completed by inserting an item constructed by previous research.
21
Product risk
Product risk was measured by 3 items constructed by Dai, Forsythe and Kwon (2014). The first item was ‘It is difficult for me to judge the product adequately’. In result, the first question for physical purchases read as ‘When I’m purchasing a video game in a physical store it is difficult for me to judge the product adequately’. The respondent could then express his/her opinion by clicking 8 different options, ranging from ‘Strongly Disagree’ to ‘Strongly Agree’, with an 8th option of not relevant. The 7-point Likert scale was used for all the mentioned constructs in this section. The second item measuring product risk was ‘It is difficult for me to compare the quality of similar products’. The third item for product risk was ‘The product purchased may not perform as
expected’. Measuring the Cronbach’s alpha, product risk scored 0.699 in a physical setting, 0.813 in an online setting and 0.815 in a digital setting. While the alpha in a physical setting is just below the normally upholded value of 0.7, no item was deleted in order to be able to still compare the means in the different setting. The strong Cronbach’s alpha in the other settings would be compromised by also having the same item deleted. This scale thus consists of three items.
Financial risk
Financial risk was also measured by three items derived from the research of Dai, Forsythe and Kwon (2014). The first item read ‘I’m concerned my bank records may not be secure’. Again, depending on the “block” a respondent was in, the first part of this sentence was ‘When I’m purchasing a video game in a physical store/via an online store/ via PSN or Xbox Live…’. The second item measuring financial risk was ‘I am concerned I may not receive the product’. The third item measuring financial risk was ‘I am concerned I may buy the product at a lower price
somewhere else’. The Cronbach’s alpha for financial risk was 0.23 in a physical setting, 0.674 in an online setting and 0.380 in a digital setting. The alpha was improved for all scales by deleting the third item resulting in respective scores of 0.75, 0.74 and 0.68. The scale then included two items. While having a scale constructed of more than two items is preferable and will result in a better indicator, it is not uncommon in survey with larger sample to end up with two-item scales due to poor quality items (Eisinga, 2013). Important to note it that Eisinga does not advocate the use of two-item scale, but recognizes the circumstances under which it can happen and then recommends the use of a Spearman-Brown coefficient for testing the reliability (Eisinga, 2013). The two items scale leads to a Spearman-Brown coefficient of 0.77 in a physical setting, 0.76 in an online setting and 0.69 in a digital setting.
Convenience
22 The wording of the items was changed slightly in order to fit the previously explained blocks. In the original research of Rohm and Swaminathan the first item was worded as ‘The internet is a
convenient way of shopping’. In the online survey posted by this research the first item read as ‘When I’m purchasing a video game in a physical store I find this a convenient way of shopping’. The second item was a reversed coded item, ‘The internet is often frustrating’. This item was also posed by replacing ‘The internet’ by all three settings. The third item for convenience was ‘I save a lot of time by shopping this way’. “This way” referred to the context being questioned at that part of the survey. The scale reliability for convenience scored very low Cronbach’s alpha’s in all three setting when all items were included. By deletion of the second item, the alpha was increased significantly. As a result this scale also consisted of a two items and so the Spearman-Brown coefficient was calculated. This yielded a result of 0.526 in a physical setting, 0.631 in an online setting and 0.748 in a digital setting. While the level of alpha is still quite below the value of 0.7 for the physical setting, this scale was kept due to stronger alphas in an online and digital setting, in order to be able to compare the variable over the three settings.
Service
The items measuring service were not adapted to each setting. While this would have been consistent with the previous adapted items, the items did not lend themselves for this kind of transformation while still being logical. The first item, as used by Kollman et al (2012), was ‘I consider it important to be able to touch and try out product before purchase’. This would have been completely nonsensical if the questioned started with ‘When I’m purchasing a video game via PSN or Xbox Live…’. Therefore all three items were posed at the end of the survey, without referring to a context. Item 1 was simply ‘It is important that I can see and touch a game before I buy it’, item 2 was ‘I am prepared to pay extra for more advice and service’ and item 3 was ‘It is important to get guidance while buying a game’. When testing reliability with all three items the Cronbach’s alpha was 0.648, which would go up to 0.707 when item 3 was deleted. The Spearman-Brown coefficient was 0.717 for this two item scale, thus resulting in the exclusion of item 3.
Ease of use
Following the Technology Acceptance Model, ease of use and perceived usefulness were added in the digital purchases model. For the adaption to the current research the original
technology used by Davis, Bagozzi and Warshaw (1989) was simply replaced by the technology applicable. Ease of use was measured by 4 items, as set up by Davis, Bagozzi and Warshaw (1989). Item 1 was ‘It would be easy to get what I want from PSN/Xbox Live’. Item 2 was ‘It would be easy to become skilful in PSN/Xbox Live’. Item 3 was ‘It would be easy to buy a game with
23 PSN/Xbox Live’ and item 4 was ‘PSN/Xbox Live is easy to use’. Ease of use had a Cronbach’s alpha of 0.699 with 4 items, which was upped to 0.758 with three items. Due to the large increase (>0.1) and still having three items left in ‘PSN/Xbox Live is easy to use’ was deleted as a scale item even though the initial alpha was close to 0.7.
Perceived usefulness
Perceived usefulness was also measured by four items. Item 1 was ‘Using PSN/Xbox Live would improve my purchasing experience’, item 2 was ‘I would buy more games with PSN/Xbox Live’, item 3 was ‘I save time by using PSN/Xbox Live’ and item 4 was ‘It would be useful to use PSN/Xbox Live’. Perceived usefulness had a Cronbach’s alpha of 0.799, thus being above regular sufficiency levels.
An overview of the eventual scales and corresponding Cronbach's alpha/Spearman-Brown coefficient can be found in table 3.1.
3.3.3 Control variables
The control variables measured were gender, occupation and income. The effect of gender was most recently found by Pascual-Miguel et al (2015) to have a significant effect on online shopping. Hernandez, Jiménez and José Martin (2011) had found that age, gender and income are no longer of significant value, but that experience matters most. Due to the relation between experience and age, age was considered an independent value and income a control variable. Income is then again linked to occupation, which was also added as a control variable. Occupation was also added in the survey for reasons not related to this research, and it was not harmful to include this in regression model. Occupation listed seven categories; high-school student, secondary education student (MBO/HBO/WO), working part-time, working full-time, unemployed, retired and other. None of the respondents filled in retired or other, and thus the categories were excluded from further analysis. Income was constructed into categories based on the most frequent observed income in The Netherlands in 2014 (CPB, 2015). The most observed income was €2500 per month. The categories constructed ranged from less than €625 per month, between €625 and €1250 per month, between €1251 and €2500 per month, between €2501 and €3750 per month, between €3751 and €5000 per month and more than €5000 per month. All categories were used in the survey.
3.4 Data collection
The survey was posted on the Benelux Twitter account of the global producer and
24 of choice. The first post led to 25 respondents. After 19 days a second post was published on
Twitter, this time with a bigger emphasis on the possibility to win a free game. Different employees of the company also shared a link on different Facebook-pages not belonging to the company. The survey was available in total for 23 days, resulting in 161 responses. One response was discarded due to an apparent lack of understanding and level of seriousness, leaving 160 responses to be analysed. An overview of the items measured in the survey can be found in the appendix
The survey also contained question that were not drafted for the purpose of this research. The example given previously was for instance occupation, because it says something about the demographics of video game players. This information was requested by the organization, to provide additional insights for their own understanding. While these questions and answers are not founded in academic theorization, they might be useful in the discussion of the results, or point out new directions for future research. The aim is not to include this data, as no hypotheses rely on this information.
Eventual scales used*
Physical Online Digital
Product risk 0.699 0.813 0.815
It is difficult for me to judge the product quality adequately. It is difficult for me to compare the quality of similar products. The product purchased my not perform as expected.
Financial risk 0.75 0.74 0.68
I'm concerned my bank records may not be secure. I am concerned I may not receive the product.
Convenience 0.526 0.631 0.748
I feel this is an easy way of buying a game. I feel this is a time saving way of buying a game.
Service** 0.648
It is important that I can see and touch a game before I buy it. I am prepared to pay extra for more advies and service.
Ease of use** 0.758
It would be easy to get what I want from PSN/Xbox Live It would be easy to become skilfull in PSN/Xbox Live It would be easy to buy a game with PSN/Xbox Live
Perceived usefulness** 0.799
Using PSN/Xbox Live would improve my purchasing experience I would buy more games with PSN/Xbox Live
I save time by using PSN/Xbox Live It would be useful to use PSN/Xbox Live
*Not asked in a specific setting
** Two-items scales were measured with a Spearman-Brown coefficient
25 4 Results
Since this research contains quite a number of hypotheses, an overview can be found in table 4.1. This also includes whether or not the hypotheses were accepted or rejected, based on the
statistical analysis that was conducted. A checkmark indicates that the relationship was found and that the hypothesis was accepted. The steps and results of this analysis can be found in the rest of this chapter.
4.1 Descriptives Dependent variables
The three dependent variables were expected physical purchases, expected online purchases and expected digital purchases. This was measured by simply asking the respondents how many games they expected to buy in each setting in the next 6 months. For physical purchases the range varied from 0 to 15. When testing for outliers the case with 15 expected purchases was considered an outlier. As is stated by Field (2009) to be an acceptable practice, the outlier was replaced by the mean + two times the standard deviation. This resulted in a score of 6, and the new range was then between 0 and 10. The mean then became 1.65, with a standard deviation of 1.826. For online purchases three outliers were found, with a score of 25, 15 and 10. Using the same method as before the outliers were transformed. The range then went from 0 to 7, with a mean of 1.61 and a standard deviation of 1.615. For digital purchases four outliers were found, with scores of 50, 15, 12 and 10. Again these were transformed using Field's recommended technique of the mean + two standard deviations. This resulted in a mean of 1.16 and a standard deviation of 1.574.
Independent variables
The difference in means and standard deviations of the same variables in different models tells us something about the distribution of the answers. All items could be answered on a 7 point Likert scale, with an 8th option of 'not relevant', in case respondents did not want to or did not know how to answer the question. All responses of 'not relevant' were excluded from the calculations.
The mean of the physical product risk scale was 2.50, with a minimum range of 1 (strongly disagree) and a maximum of 6 (Agree). The standard deviation is 1.206. Online product risk had a mean of 2.61, with a standard deviation of 1.382. Respondents used the full range of 1 to 7. Product risk had a mean of 2.87, with a standard deviation of 1.505 in a digital setting, with a scale only ranging from 1 to 6.
26 Hypotheses
1. There is a positive relationship between an individual's perceived usefulness of buying digitized products and the expected amount of digitized products purchased. × 2. There is a positive relationship between an individual's perceived ease of use and the
expected amount of digitized products purchased. ×
3a. There is a positive relationship between experience and the expected amount of digitized products purchased.
ü 3b. There is a positive relationship between experience and the expected amount of
online products purchased. ü
3c. The relationship between experience and the expected amount of digitized products purchased will be stronger (more positive) than for experience and the expected amount of online products purchased.
ü
4a. There is a negative relationship between product risk perception and the expected
amount of digitized products purchased. ×
4b. The relationship between product risk perception and the expected amount of digitized products purchased will be stronger (more negative) compared to product risk
perception and the expected amount of online products purchased. × 5a. There is a negative relationship between financial risk perception and the expected
amount of digitized products purchased. ×
5b. The relationship between financial risk perception and the expected amount of digitized products purchased will be stronger (more negative) compared to product risk
perception and the expected amount of online products purchased. × 6a. There is a negative relationship between age and the expected amount of digitized
products purchased. ×
6b. There a positive relationship between age and the expected amount of physical
products purchased. ×
7a. There is a positive relationship between convenience and the expected amount of digital products purchased.
× 7b. There is a positive relationship between convenience and the expected amount of
online products purchased. ×
7c. There is a negative relationship between convenience and the expected amount of
physical products purchased. ×
7d. The relationship between convenience and the expected amount of digitized products purchased will be stronger (more positive) compared to the relation between convenience and the expected amount of online products purchased.
× 8a. There is a negative relationship between service and the expected amount of digital
products purchased. ×
8b. There is a positive relationship between service and the expected amount of physical
products purchased. ü
27 For financial risk in the physical setting the range also varied from 1 to 6, with a mean of 1.62, and a standard deviation of 0.957. For online purchases financial risk had a mean of 3.29, with a standard deviation of 1.676. In a digital setting financial risk scored a mean of 2.58 with a
standard deviation of 1.449.
For convenience the mean was 3.79 with a standard deviation of 1.449 in a physical setting. In an online setting convenience scored a mean of 5.20, with a standard deviation of 1.521. Digital convenience had a mean of 5.06, with a standard deviation of 1.596.
Service scored a mean of 3.14, with a standard deviation of 1.426. Ease of use had a mean of 4.51, with a standard deviation of 1.405, while perceived usefulness scored a mean of 3.55 and a standard deviation of 1.459.
Control variables
In terms of gender, 92% of the respondents were male. While it doesn’t seem likely that this is representative of the total console gaming population in the Benelux, exact data on the gender distribution for console gaming is unavailable. More generally speaking, the internet Advertising Bureau stated in 2014 that 52% of all gamers in the UK were women, but this included the vastly growing popularity of mobile gaming (The Guardian, 2014). The biggest occupational group consisted of full-time workers (73 respondents, 46%), and most people earned between €1251 and €2500 a month (33%). The second biggest occupational group were respondents in their secondary education (46 respondents, 29%), which also explains the large group of people earning less than €625 a month (32%). The PlayStation 4 was the most commonly owned console (59%), followed by the Xbox One (16%) and the PlayStation 3 (13%). This is also not in line with the general distribution of consoles, but can be explained by the fact that active gamers have switched to the new generation of consoles, whereas older generations that don't game as much anymore still have the older generation consoles.
4.2 Correlations
Table 4.2 shows the correlation matrix for all control, independent and dependent variables. There are some significant correlations worth mentioning. When looking at the dependent
valuables, expected physical products purchased has significant correlations with physical
convenience and physical experience. Interestingly, there is also a positive correlation found with online financial risk, and a negative relation with convenience. As the perceived online product risk goes up, the expected amount of physical games bought in the next 6 months go up. When the valuation for online convenience goes down, the amount of expected games to be bought in a
28 physical store goes up. Also a strong correlation can be found between the amount of games bought in the last 6 months in a store, and the expected amount of games that will be bought in the next 6 months
When looking at the dependent variable of expected online purchases, negative correlations can be found for both online product risk and online financial risk. A decrease in the perceived risks is correlated to an increase in the expected amount of games to be bought online. Again, a strong correlation is found between the amount of games bought online in the last 6 months and the amount of games expected to be bought in the next 6 months. A negative correlation is found between both physical and digital convenience and expected online games bought, meaning that as the valuation of physical and digital convenience goes down, the amount of games bought online goes up. Lastly, another negative correlation can be found with service. As the valuation of service decreases, an increase can be seen for the expected amount of games bought online.
With respect to expected amount of digital games to be purchased in the next 6 months, significant correlations can be found for product risk, financial risk, convenience, experience, service, age, perceived usefulness and ease of use. Interestingly, the correlation between age and expected amount of digital games to be bought is positive, whereas a negative relation is
hypothesized. Other significant correlations were found for occupation and income.
All models were also tested for multicollinearity. For the model with expected physical purchases as a dependent variable, tolerance ranged between 0.323 and 0.835, and the VIF between 1.197 and 3.094. As stated by Field (2009), cause for concern arises when the tolerance level drops below 0.2 and the VIF is above 10. In the model with online expected purchases as the dependent variable, tolerance ranged between 0.323 and 0.828, and the VIF between 1.207 and 3.093. For digital purchases as the dependent variable the tolerance ranged between .338 and 0.868 and the VIF between 1.152 and 2.818. Thus, multicollinearity was not found to be of existence.
29
Table 4.2 Correlation matrix of all variables
4.3 Regression analysis
A negative binomial regression analysis was conducted in order to test the full model with all variables. A negative binomial regression was most appropriate due to several reasons. First of all, the dependent variable contained count data, as it reflected the expected number of games to be purchased in the next six months. Secondly, the data contained a high numbers of zero’s, and was positively skewed. Third, the variance was much higher than the mean, indicating overdispersion. In the case of overdispersion a negative binomial regression is more appropriate than a Poisson regression, which can also be used for count-data.
When conducting a negative binomial regression using SPSS, you have the option to specify the overdispersion parameter. The default setting is always set at 1. When running the negative binomial regression with the regular overdispersion setting for physical purchases, the model was significant with at p < 0.001, with a LR-chi² of 54.451. No control variables or independent variables were found to have a significant effect, except for experience. When the overdispersion parameter is set to estimate the value of the overdispersion parameter of this particular distribution, two issues arise, indicating a uncertainty in the validity of the fit of the model. The first issue was regarding the number of step-halvings undertaken, and was resolved by increasing the number of step-halvings from 5 to 100. The second issue is related to the singularity of the Hessian matrix. Singularity in the Hessian matrix can only be caused by two things: multicollinearity or an exceeding number of variables compared to the number of observations (Gill & King, 2004). It appears that the singularity is caused by an assumption of multicollinearity between expected purchases and past purchases. However, when testing for multicollinearity the relation was not found to be significant, with a Spearman correlation coefficient of 0.735 at p < 0.001, and a
30 tolerance level of 0.86 and VIF level of 1.6. As stated before, tolerance levels bellows 0.2 and VIF levels above 10 are cause for concern, which is not the cause. Therefore the warning regarding the uncertainty of the validity of the model fit can be discarded. When using the estimated
overdispersion parameter, the LR-chi² increases to 115.376 at p < 0.001. Also, more variables were found to have a significant influence.
Table 4.3 gives an overview of the results of the regression analysis. Dummy variables were created for occupation (high-school student as a reference category) and income (less than €625 per month as a reference category). Model 1 had expected physical purchases as the dependent variable. The LR chi-ratio showed that the model was significantly better than an intercept-only model at p < 0.001. The control variables were gender, occupation and income. None of the control variables had a significant effect on the outcome variable. When looking at the main effect, experience, product risk, financial risk, age convenience and service were entered into the model. Financial risk had a significant positive effect on physical expected purchases, with β = 0.269 at p < 0.01. Service had a significant effect with β = 0.132 at p < 0.05. Experience was the strongest variable yielding
significant results, with β = 0.513 at p < 0.001.
Model 2 had the expected online purchases as a dependent variable. This model was also significant, at p < 0.01, and a LR-chi² of 59.894. Only experience had a significant relationship to the dependent variable, with β = 0.278 at p < 0.001. All other variables, testing both main effects and control variables were non-significant.
Model 3 was related to the expected digital downloads in the next 6 months. As independent variables perceived usefulness and ease of use were added. The total model was significant at p < 0.001. Significant results were found for age with β = 0.050 at p < 0.05 and experience with β = 0.412 at p < 0.001.
As experience was found to have a significant effect in all three models, the hypotheses related to this will be discussed together. Hypothesis 3a and 3b claimed the positive relationship between experience and the amount of games to be bought in the next 6 months both digitally and online. These hypotheses can be confirmed, at p < 0.001 and β = 0.278 for online purchased games and β = 0.412 for digital games. Hypothesis 3b stated that the relation would be stronger between experience and expected digital games compared to experience and expected online games. While they were both significant at p <0.001, the β of expected digital games is almost 1.5 times bigger than the β of expected online bought games. This means that the slope of the line is steeper, and the effect of experience on the to be expected amount of games is indeed bigger. Therefore, hypothesis 3c is also confirmed.
When looking at model 1, service has a significant effect on expected physical products purchased, confirming hypothesis 8b. However, the evidence is quite weak, with p < 0.05 and β of
31 0.132. The other effect found was a positive significant relationship between financial risk and expected physical products purchased, with a β of 0.269 at p < 0.01. This would mean that as the financial risk associated with buying a game in a store goes up, the actual amount of games bought in a store also goes up. This is contrary to previous found relations between risk perceptions and amount of purchased goods in all settings. The implications of this finding will be discussed in the next chapter. The other hypotheses all related to physical purchases, 6b, 7c and 8a, were rejected.
Since no significant relations were found in model 2. As a consequences hypotheses 3b and 7b were rejected.
When looking at model 3, age was found to have a significant positive effect on expected amount of digital games to be bought. This is contrary to hypothesis 6a. Age was normally
distributed, but the range was quite small, from 13 to 46. Why this is of influence will be discussed in the next chapter.
4.4 Robustness checks
A couple of robustness checks were performed. First, the dependent variables were tested or normal distribution. The Kolmogorov-Smirnov test showed that all three variables deviated
significantly from a normal distribution at p < 0.001. Secondly, the regression analysis was performed using untransformed data. When the analysis was performed with the untransformed data, the results and significant effects remained the same. Third, as mentioned before the
parameters for overdispersion were set to be estimated based on the distribution of the count data. Normally the dispersion is set to 1. An estimate close to 0 indicates a Poisson distribution. The estimated dispersion was very close to 0 for all dependent variables, so a Poisson regression was also run to check the results. Again, the results and effects remained the same. Fourth, negative binomial regression analyses were performed using past purchases as the dependent variable. For reasons explained in the next chapter, the number of expected purchases might be overestimated. For online purchases the model was not found to be statistically significant. The model for physical purchases was significant at p < 0.05, and convenience was found to have a significant effect at p < 0.001 with β = 0.207. For digital purchases the model was significant, and convenience and
perceived usefulness were found to have a significant effect. Convenience was significant at p < 0.001 with β = 0.249 and perceived usefulness was significant at p < 0.05 with β = 0.195. The implications of the findings of these new models will be discussed in the next chapter.
32
Table 4.3 Results of the regression analysis
Model 1 Model 2 Model 3
Physical purchases Online purchases Digital purchases Coefficient Standard Error Coefficient Standard Error Coefficient Standard Error Control Variables Gender (male) -0.206 0.287 0.042 0.439 -0.798 0.43
Occupation (Ref: High school student) Secondary education -0.499 0.338 -0.21 0.272 -0.267 0.423 Working, part-time -0.647 0.466 0.134 0.468 -0.727 0.675 Working, full-time -0.436 0.517 -0.174 0.446 -0.753 570 Unemployed 0.115 0.755 -0.199 0.559 -0.658 0.637 Income (Ref < €625) Between € 625 and €1250 0.357 0.282 0.326 0.278 0.006 0.318 Between €1251 and €2500 0.165 0.429 0.526 0.382 0.482 0.469 Between €2501 and €3750 0.1052 0.477 0.392 0.428 0.458 0.503 Between €3751 and €5000 -0.466 0.558 0.58 0.542 0.694 0.574 More than €5000 -0.87 0.725 1.036 0.452 0.763 0.682 Main effects Product Risk -0.115 0.087 -0.044 0.06 -0.033 0.087 Financial Risk 0.269** 0.096 -0.042 0.045 -0.112 0.093 Convenience 0.026 0.069 -0.019 0.047 -0.079 0.07 Service 0.132* 0.06 -0.105 0.055 -0.066 0.0686 Age 0.012 0.02 -0.015 0.015 0.050* 0.02 Experience 0.458*** 0.052 0.278*** 0.041 0.412*** 0.052 Perceived Usefulness - - - - 0.147 0.08 Ease of use - - - - 0.126 0.084 LR chi2 115.355 59.894 110.807 N 133 142 134 * p < 0.05 ** p < 0.01 *** p < 0.001
