Big Data
The effect of awareness on the consumers attitude
By
: R.M. de Koning
Student number
: 10687858
Qualification
: Executive Programme Business Studies, Marketing track
Institution
: Amsterdam Business School, University of Amsterdam
Thesis supervisor
: Dhr. E. Peelen
2
Statement of originality
This document is written by student R.M. de Koning who declares to take full responsibility for the contents of this document.
I declare that the text and work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of the completion of the work, not for the contents
3
Abstract
In a fast changing and digitizing world a new phenomenon has been on the rise, Big Data. Companies,
corporations, institutions are all collecting the digital breadcrumbs the everyday consumer leaves behind with almost everything they do. The current generation consumers is fueling this phenomenon. But to really make a difference with this new data we need to know more about how this is going to fit within our current society. We investigate if the consumer is aware of this phenomenon taking place and what effect this might have on their attitude towards Big Data, expand on the current problems third parties are currently facing with collecting data and protecting the consumer and address potential solutions to the problems. We find that although the awareness of the consumer is relatively high it does not impact the attitude in a strong matter. Benefit for the consumer seems to make the difference in the attitude shifts of the consumer, where a beneficial situation renders the consumer indifferent and in a situation where the consumer does not seem to benefit directly awareness as such seems to have an small impact on the attitude.
4
TABLE OF CONTENTS
1.
Introduction ... 5
1.1
Big Data ... 6
1.2
Awareness ... 7
2.
Privacy concerns ... 9
2.1
Privacy ... 9
2.2
Transparency ... 10
2.3
Privacy intrusion ... 11
3.
Big data and legal issues ... 12
3.1
Current legal protection ... 12
3.2
Addressing potential solutions... 13
4.
Application possibilities of Big Data in the field of marketing ... 14
4.2
Data growth ... 15
4.3
Marketing applications ... 15
4.4
Is this what the consumer wants? ... 16
5. Methodology ... 17
5.2
Method and data ... 18
5.3
Analysis and common method variance ... 19
6. Results ... 19
6.1
Measurements ... 19
6.2
Descriptive statistics and correlations ... 21
6.3
Hypothesis testing ... 24
7. Discussion ... 31
7.1
Conclusion ... 34
7.2
Limitations and recommendations for future research ... 35
8.
References ... 36
5
1.
Introduction
We live in a fast changing world. Merely 32 years ago ARPANET decided to switch their network of computers from NCP to TCP/IP (networkprotcol), a step that would change the world forever and as least as much as the industrial revolution did. With this switch, the internet as we know it today was born. Now roughly 30 years later we are on the verge of stepping into a worldwide ‘Big Brother’ scenario. We call this phenomenon ‘Big Data’. At its core Big Data is not primarily a business or research revolution, but a social one. In the past decade we allowed machines to act as intermediaries in almost every aspect in our existence. When we communicate with friends, entertain ourselves, drive, exercise, go to the doctor, read a book – a computer transmitting data is there. The seemingly meaningless, incidental bits of data that we shed are turning the concept of privacy into an anarchism, despite half-hearted regulations to protect the ‘personally identifiable information’ (Seife, C. 2015). But are consumers of this generation aware of the trail of breadcrumbs they leave behind about themselves, and if so, how do they feel about this? The attitude of the consumer regarding the collection of data about them, might play a large role in how this will play out in the near future. Are consumers with a high(er) level awareness more inclined to try and protect their privacy and do other aspects moderate their attitude?
This paper addresses three sections in specific: awareness, privacy and the attitude of the consumer. The awareness concept in this paper refers to the current awareness consumers have regarding the phenomenon Big Data and all its perils. The privacy issue is one of the most prominent topics regarding the gathering of personal data for marketing applications. The legality of this and in which way(s) the data and the individual is being protected is intertwined. The applications possibilities for Big Data are endless, however this papers focuses on four marketing concepts in general and the influence awareness might have on them. The main question revolves around awareness, does this influence the consumers attitude and if this effect is moderated by transparency.
6
1.1
Big Data
Francis Diebold was in 2003 probably the first one to use the term Big Data for the emerging explosion data growth. Kirk Borne (2014) has written a paper on the so called 10 V’s of Big Data. He has mentioned these 10 V’s as challenges to put Big Data to use. These 10 V’s are: ‘Volume, variety, velocity, veracity, validity, value, variability, venue, vocabulary and vagueness. All these aspects contribute to the challenges and definition of Big Data. Boyd and Crawford (2012) define Big Data as a cultural, technological and scholarly phenomenon that encompasses a closely intertwined amalgam of technology, analysis and mythology. Hence there are many different definitions of this phenomenon, they differ throughout different markets, usage and corporations. Although there is no
commonly accepted definition of Big Data, we can say that it is data that can be defined by some combination of the following five characteristics: Volume (where the amount of data is to be stored) Variety (where the data consists of) Velocity (where the data is produced at high rates), Value (where the data has perceived or
quantifiable benefit to the enterprise using it) and Veracity (where the correctness of the data can be assessed). (Gordon K. 2013). To visualize the concept of Big Data, Doug Laney (2001) came up with a 3-V model of Big Data (figure 1). This model is still a good abstract visualization of the Big Data phenomenon. It defines the three dimensions of Big Data in a 3D management. Volume, Variety and Velocity are identified as the 3 main constructs of Big Data.
7
If we combine all the constructs which the current academic literature links to the concept of Big Data we can state that the term Big Data is the name of the dramatic increase in data growth and the possibilities that come with this. Possibilities like new insights and the availability of this data for decisions based on a large informational base. Data can be reused in a smart way in which they can be the ingredients for new innovations and services. (Steenbruggen. J, et al, 2015). Boyd and Crawford state in 2011 that there is little doubt that the quantities of data now available are indeed large but that this is not the most relevant characteristic. Big Data is notable not because of its size but because of its relationality to other data.
1.2
Awareness
When Edward Snowden made himself known as the whistleblower that had exposed N.S.A. practices of routine surveillance to the news media, he described in detail the ‘architecture of oppression; that enabled him and many other N.S.A.-contractors to intercept the metadata of three billion phone calls and interactions recorded by Facebook, Google, Apple and other tech companies. This came as more than a wakeup call for consumers who have gradually come to accept the ‘sharing’ of personal information – everything from marital status to colds, sports, interests, hobbies, eating habits to music tastes – via social network sites or apps as the new norm. (Van Dijk, J. 2014). However consumer awareness remains relatively low despite the accusations Edward Snowden has made towards the government. Everyone who makes use of the facets of the digital revolution, leaves behind a digital fingerprint. A small piece of information about themselves. Technology has come so far to gather all these parts of information left behind on the internet on various platforms and combines them for multiple purposes. Data is generated from so many places namely; social media, RFID codes, satellites, digital sensors, Internet of Things, Energy meters, mobile phones, vehicles, social media and so on (Moorthy, Janakiraman et al, 2015). The current generations are almost not aware of this phenomenon taking place. This was partly shown in a study conducted by Timothy R. Greaff and Susan Harmon in 2002. They conducted a survey to investigate why consumers thought supermarkets were using discount loyalty cards. The results showed that only a mere 16.5 percent of the customers gave responses related to database marketing and collection of information about customer buying habits. One might assume that with the increase in usage of Big Data, upcoming generations will be more aware. Even the current ‘younger’ generations will become increasingly aware of this phenomenon. When consumers become better aware of this phenomenon, the questions remains how consumers will react to that. We are currently living in a changing society, one where the boundaries of privacy and sharing are becoming vaguer than ever. Users of social media are putting pieces of personal information on the web, information which previous generations maybe would think twice about to share. Hence when the current consumer becomes completely aware of the Big Data phenomenon, will they accept the fact that corporations are gathering and using personal information about them for other purposes than originally intended by the personal who made the information available. Considering that the consumer is the one that made this information available in the first
8
place, will they realize that the consumer is also the one who can deflect the gathering of information about them. There has been no research yet on how consumers might counter the gathering of information about them or how much effort they are willing to put in it. This paper explores the possibility of awareness having an effect on the attitude of the consumer. To test this a conceptual model has been developed which is the basis for six different hypotheses. The first four hypotheses will test the direct effect awareness might have on the attitude of the consumer related to four different marketing concepts which are already being put to use by different
organizations. The first hypothesis has been formulated to explore the possibility of awareness having an effect on the attitude of the consumer regarding the usage of the their public available data. Awareness as such will consist out of two components. The first refers to whether the consumer is aware of the fact that he or she leaves behind traces of digital information about themselves on the internet. The latter refers to how aware the consumer is regarding the fact that companies, organizations, institutions (i.e. third parties) are actively searching and collecting these pieces of data. The first attitude concept relates to the usage and collecting of data that has been made publicly available by the consumer. Think of pictures that have been placed in a public folder on Facebook or Instagram, but certainly also information about interests (likes), places the consumer has been (check-ins) or work experience (Linked-In).
H1: The level of awareness the consumer has about Big Data has a direct effect on the attitude of the consumer regarding the usage and collection of public available personal data by third parties The second attitude concept relates to a more personal matter. One of the usages of the personal information is that third parties are able to filter your specific interests. For example, one might search frequently for outdoor shops (google), likes outdoor athletes on Facebook and books an outdoor vacation via an online bookings site. A third party who combines this information can conclude that their outdoor contraband might be of interest of you. Hence the next advertisement e-mail the consumer gets from this third party, contains only the products related to the consumers interests. We call this phenomenon ‘personalized advertising’.
H2: The level of awareness the consumer has about Big Data has a direct effect on the attitude of the consumer regarding the usage and collection of their personal data for personalized advertising by third parties
The following hypotheses are more generally formulated. Third parties can generally use Big Data for two reasons, commercial goals and the collective good. In this paper we define the commercial goals of a third party as
everything which the third party will benefit from. This could be increasing sales, turnover and profit margins but also improving service or being able to update their product assortment more tailored to their customers’ needs.
H3: The level of awareness the consumer has about Big Data has a direct effect on the attitude of the consumer regarding the usage and collection of their personal data for commercial goals of the
9
The fourth hypothesis explores the possibility that awareness might have an influence on the attitude of the consumer when their data is being put to use for the collective good. Imagine a governmental institution collecting personal data of the consumer to see if the public facilities of a city are still sufficient or the government using personal data of the consumer to search for possible threats to society in general, like terrorism. The data that is being collected is so diverse that the possibilities are endless.
H4: The level of awareness the consumer has about Big Data has a direct effect on the attitude of the consumer regarding the usage and collection of their personal data for the collective good
2.
Privacy concerns
2.1
Privacy
With the emergence of Big Data it seems that we need to reconsider the aspect of privacy. The desire for privacy in sense of protection or escape from other human beings, emerges when an individual becomes subject to social obligations that that individual cannot meet or does not want to meet (Moore, B. 1998). This definition of privacy is partly at play here. The part where Moore defines the sense of protection or escape is at use, however the part where it refers to social obligation is not. The privacy concerns we are facing in the age of Big Data is one of not being able to hide certain personal aspects when we want to. The amount of user-generated content which is being uploaded to the web is expanding rapidly. The growing proliferation and capabilities of mobile devices is creating a deluge of social media which can effect out privacy. Due to the vast amount of data being uploaded every day it is next to impossible to be aware of everything which effects the consumer (Mattew, S; Szongott, C; Henne, B; Von Voigt, G. 2012). Hence even if a consumer would stop using technology all together right now the amount of information already left behind or even the information other consumers are still providing about this individual makes privacy to a certain extend hard to maintain. Categorization of privacy issues can be split up in to two classes. Firstly, home-grown problems: someone uploads a piece of compromising media of himself with insufficient protection or forethought which causes damage to his own privacy. A prime example is uploading a picture into a public folder or timeline (Facebook) instead of a private folder. Secondly there are Big Data problems created by others. This takes place when other users upload content in which the person is harmed is not involved in the upload of this media. (Mattew, S; Szongott, C; Henne, B; Von Voigt, G. 2012). Also actively stalking the consumer by following their every move on the web and using this information for commercial purposes. There is still no clear guideline or law on what information corporations can and cannot use, however the European privacy laws depicts that the consumers data needs to be protected. For example application developers need to make sure that the faith of the consumers data is still in their own hands, by making their apps only use the data needed for the app and their security is up to date (CPB jaarverslag, 2013).
10
2.2
Transparency
The White House (United States) has recently released two reports addressing the public policy implications of Big Data’s proliferations. Both reports make the case that instead of slowing the accumulation of data or placing barriers on its use, we should focus on policy initiatives and legal frameworks that foster innovation, promote the exchange of information, and support public policy goals, while at the same time limiting harm to individuals and society. (Brian, M.G; Heater, E.S; Jennifer G. 2014).
So while the US government is stating that we should actually embrace the transparency of the Big Data phenomenon another study conducted in the United States by Nir Kshetri in 2014 showed results of consumer’s awareness about these privacy issues. He summarized findings of surveys conducted with businesses and consumers regarding their perceptions and responses to Big Data. This survey showed mixed results in a survey held under 600 IT and business professionals only 38% of the respondents raised concerns for data security and privacy issues, 30% of the respondents in a national survey among 2254 US adults said that they turned off their location tracking on their mobile phone out of fear that someone might access this information and in a Global Conumer Sentiment Survey conducted by BCG in 2013 75% of the respondents opted that the privacy of personal data was a top issue. Only 7% of the respondents of the latter survey were willing to allow their information to be used for purposes other than initially intended. Transparency of sorts might have a moderating effect on the attitude consumers have toward the gathering of personal information. When a consumers is informed about the collection and the end goal for the collection of this data there might be a shift in attitude. Hence the last two hypotheses explore the possibility of transparency moderating the effect of awareness on the attitude on the consumer. Transparency of sort is again split up in two components. The first is the level of transparency the consumer experiences from the third party that is collecting their data. This could be in the form of a pop-up on a website which refers to the privacy statement or it might be a consent form which the consumer actually needs to accept.
H5: The level of transparency the consumer experiences from the third party that is collecting their personal data moderates the effect awareness has on the attitude of the consumer
The last hypothesis has been formulated to explore the possibility of the attitude of the consumer towards transparency moderates the effect of awareness on the attitude of the consumer towards a specific marketing concept. Thus how does the consumer feel about transparency and does this have a moderating effect? Does the consumer feel that there is a need for transparency from the third party out and does this feeling moderates the effect awareness has on the attitude of the consumer.
H6: The attitude of the consumer towards transparency regarding the collection and usage of personal data moderates the effect awareness has on the attitude of the consumer
11
2.3
Privacy intrusion
Data scientists are using the ‘Big Data flow’ to find out patterns in human behavior. With special regard to the marketing industry which could benefit highly from this kind of information. They try to explore shopping patterns, health status, sleep cycles, moving patterns, online consumption or even the connection to individuals
(friendships, relations etc.) (Zwitter. A, 2014). One independent survey of adult internet users found that two-thirds of the users object to online tracking by advertisers. Respondents especially disliked behavioral advertising, in which commercial websites tailor ads based on a individuals web behavior (Bollier, D., & Firestone, C. M. 2010). The type of channel this information is being gathered through and the use of this information might partly define if consumers experience this as privacy intrusion. Data scientist try to anonymize this data by removing certain aspects which can link the information to the original owner. However the greater part of the information is used to define these kind of patterns in groups with certain aspects. This so called de-individualization (the removing or leaving certain elements out the equation that allows data to be connected to one or multiple specific persons) is just one aspect of the anonymization. So to what extend will data scientists remove elements out of the data set? To strip data from all elements pertaining to any sort of group belongingness would mean to strip the data from its content. This data can be used to target people or groups in a way to persuade them to behave in a certain way, by for example targeted marketing. If certain aspects about conditions and preferences of a group are known this can be used to employ incentives to encourage certain behavior. For example if we know that group 1 has a preference for chocolate (preference) and the majority of the same group has not yet decided at which supermarket to do their groceries (condition) one could provide the preference for this group in the domain of the condition in a specific way to create a conditionality. (When shopping at supermarket X your receive free chocolate). With Big Data the ability to discover hidden correlations increases. Hence the consumer becomes more susceptible for certain influences they might not even know of (Zwitter. A, 2014). Considering the fact that these kind of tactics are being used already does not influence the fact that to a certain extend information has not been available for these kind of purposes. With the emergence of Big Data, correlations within group and/or individuals can be found and put to use without the consumer realizing this. However norms and behavior regarding privacy differs across cultures and time. But if privacy behaviors are culture- and context dependent, however the dilemma of what to share and what to keep private is universal across societies and human culture. The task of navigating the
boundaries we set and the consequences of mismanaging them, have grown more complex within the information age we currently live in, to the point that our natural instincts seem not nearly adequate (Acquisti, A; Brandimarte, L ;Loewenstein, G, 2015).
12
3.
Big data and legal issues
3.1
Current legal protection
The extraordinary technological developments of recent years have created concerns regarding possible harm to rights and interests which the law seeks to protect, such as the right to privacy, freedom of information, the right of access to the courts, the right to reputation and copyrights. The law and technology are often viewed as competitors. Often the law is perceived as lagging behind technology, and indeed – in factual terms, the law by its nature responds to technological developments. (Saxby, S. 2012). Hence the law will always be a step behind technological developments. Big Data challenges the Fair Information Practices (FIPs), which is the basis for all current privacy law. Probably the most influential privacy law at this moment is the European Union Data Protection Directive (DPD). The DPD sets out the basis for privacy regulations regarding ‘personal data’
(Rubenstein. S. I., 2013). Ira Rubenstein confronts the DPD with 3 short comings, the first is that the DPD relies too heavy on the aspect of informed choice, second is that the DPD requires data minimization, however only few corporations have been forced to redesign their systems to allow users to use them anonymously. The third shortcoming is also mentioned at the beginning of this paragraph; the DPD failed to keep up with the globalization of this phenomenon. Big Data challenges the DPD and its foundation by enabling re-identification of the data subjects using non-personal data, which weakens the anonymization as an effective strategy.
As an example; Facebook has conducted an experiment under its users, they changed the newsfeed of almost 700,000 users. They divided the users into two groups, the one group got reduced positive newsfeeds (more negative) and the other group got reduced negative newsfeeds (more positive). The outcome is in this case irrelevant, the discussion that followed was that this was done without prior consent. But what is new here: surely companies have been doing this kind of research to manipulate people’s behavior for some time, and academic researchers such as psychologists have been carrying out experiments to change people’s behavior, however in the latter part they did receive prior consent. Ralph Schroeder (2014) wrote a paper arguing that this ‘there is nothing new’ way of thinking about the usage of Big Data is defect. According to Ralph Schroeder the most pressing issue with this way of thinking is how the more powerful knowledge is used in everyday lives. So the question is actually to what extend is the information provided by consumers freely usable considering the fact that it is freely
accessible. This statement makes you think how the current legal system works for Big Data. Consider the relatively new ‘ cookie’ law. This forces websites to let users know that they place small files on their computer which tracks their activities on that specific website. This law is actually confirming Ralph Schroeder’ argument that although the activities of the consumer on the website are easy and free to track, they are not so much free to make use of that information without prior consent.
13
3.2
Addressing potential solutions
In the age of Big Data it might be unreasonable and impractical to ask third parties to get consent from everyone they gather data from, for example from everyone who likes a certain advertisement on Facebook. However it is just as problematic to claim your data is ethical simply because it is publicly accessible. We need to look at different possibilities to protect the consumer. Policymakers can do so in two basic ways, the first one is to empower the individual and the second option is to protect the individual. Empowering puts the consumer in control of their own information, while protecting requires consent from the individual. The European legal regime regarding privacy in the area of behavior target marketing aims for empowering the consumer, the needed consent for the use of cookies is an example of this (Borgesius. Z.F., 2014). However considering the limited potential of informed consent as a protection measure, Borgesuis (2014) argues that a combination of the empowerment tool and the protection tool is the way to go, with an emphasis on the protection tool. One of the main points Steve Saxby (2012) is pointing out in his article is the need to adopt a holistic perspective and systematic ‘across the board’ action in order to successfully cope with the main problems posed by new technologies. This means all players should join in a collective effort to adopt a comprehensive and systematic regime based on the following principles:
1 Encouraging customers of technology to be more aware of and protective of their rights; disabling Facebook’s facial recognition feature is a good example.
2 Encouraging providers of technology services and website owners to conduct themselves ethically and ensure the greatest possible security of databases containing private information with the result that consumers will be more forthcoming about providing information.
3 Updating legislation – conventions, directives, state legislation – to meet the technological developments challenge. Legal rules should be redefined in a broad manner, leaving the court’s judicial discretion to interpret the law.
4 Providing the courts with wide discretion – the courts should play a key role in any attempt to protect the fundamental rights entrenched in the law
Another idea proposed by Tene and Polenetsky (2013) is a shared wealth strategy. This strategy is based on the assumption that data controllers provide individuals access to their data in a usable format and with this allowing them to take advantage of their own data by using applications to analyze it. This way the users are just as able to draw conclusions from the data as the corporations who are collecting it. If we look at the current state of the DPD and privacy law which should be protecting the users and the data we can conclude that, indeed the law is lagging behind. At this point the data protection laws are not fully covering and protecting the consumer and control on this point is almost none existent. However the problem is duly noted by different institutions and progress is being made. At this point corporations still have a wild card in collecting and analyzing personal data which have undergone the process of anonymization.
14
However various studies have proven that very little (extra) data is needed to re-identify individuals, claims about big data anonymity are highly contestable even spurious (Wigan, R. M; Clarke, R., 2013). John Clippinger, co-director of the Law Lab at the Harvard University Berkman Center for Internet and Society, while not everyone will pore through all their personal data, the point is to set down a set of architectural principles for how data will be handled and how that handling will be disclosed (Bollier, D., Firestone, C. M., 2010).
4.
Application possibilities of Big Data in the field of marketing
The application possibilities of Big Data are endless. They can be company specific but also used in a more generalized way. A few examples of big data applications are: finding correlations across multiple disparate data sources, predicting consumer behavior, predicting product or service sales, predicting fraud or financial risk, analyzing social network comments for consumer sentiment or even identifying computer security risks. Some researchers believe that Big Data will revolutionize the industry in the same way the internet did this. Prospects of value generation are estimated above € 250 billion for just Europe’s public sector administration (Dobbs, R., Manyika, J. Roxburgh, C., Lund, S, 2011). At its core Big Data marketing centers on one thing only: driving value by engaging customers more effectively (Arthur, L. 2013). Lisa Arthur (2013) also describes aspects of Big Data marketing which she calls Intergrated Marketing Management (IMM). She feels that IMM can deliver internal value and external value by the usage of big data. According to Lisa it is a simple solution to maintaining all your marketing content across a global organization and she created a model consisting out of 5 steps to visualize how big data can help your organization benefit. The first step (1) consist out of the marketing operations where the marketing team can better plan, create and execute marketing campaigns by using the most up-to-date
information about how target markets are engaging. The second step (2) is what she calls digital marketing where you can reach your customers when and where it is the most convenient using digital channels. The third step (3) is where the customers ( sales and partners) receive the right message at the right time for the most optimal result (optimize your ROI). The fourth step (4) is the same as the third one but in this case aimed for B2C customers. The fifth (5) and final step is the marketing analytics. Using IMM a broader view into how and where money is being utilized during a campaign. We can see that both sides benefits (i.e. consumers and producers) from the usage of big data.
15
4.2
Data growth
The big data aim is to improve decisions and competitiveness for companies and public administration, which will create a significant growth of the world economy. Big Data will help to better listen to consumers, better
understand their ways of using services and hence the offer (Jamiy Fatima El et al. 2014). In a report McKinsey wrote in 2011 they took a shot at grasping the growth and storage of data. In a table they were able to show that the overall data storage took an exponential growth from 2000. The volume of data of this kind is measured in exabytes (1 exabyte = 1000 petabyte = 1000000 terabyte). The amount of stored data went from 50 exabytes in 2000 to 300 exabytes in 2007. The type of data stored differs as well. In the same report written by McKinsey they show a table in which the different types of data per sector are being measured. In the report we can see that the penetration of text/numbers is still the highest in most sectors but that video, image and audio have the highest penetration in the sectors: communication and media, government and education. IDC (2012) is performing a longitudinal study starting, in which they take a shot at the changes and growth of the digital universe. With data collected in 2005 and extending to 2020, the sixth annual study shows that from 2005 to 2020 the amount of data worldwide will grow with a factor 300, from 130 exabytes to 40.000 exabytes which equals 5200 gigabytes for every individual worldwide. From now until 2020 the amount of data will double every two years.
4.3
Marketing applications
Further on in the report written by McKinsey they identified 5 key sectors within the United States that can make important contributions to the world economy by using Big Data: Healthcare, Public Sector, Retail, Manufacturing and Telecommunications. For this current research the area of retail and public sector are the most important due to the overlap with marketing. In the retail industry leaders are becoming more sophisticated in slicing and dicing the Big Data they collected through multiple channels. The report mentions 6 Big Data levers within the marketing section. Cross-selling; current and most sophisticated cross-selling techniques uses all the data known of the customer. Think of demographics, purchase history, preferences in online stores, real-time locations and this is just a grasp of the possibilities they combine to offer the consumer tailor made offers. Location-based marketing; a type of marketing that makes use of the large share of consumers who uses smartphones or devices which are able to communicate their real-time location. Think of mobile phones but also laptops, PDA’s etc. It targets customers who are nearby or already in the store to provide them with information they might find interesting.
In-store behavior analysis; by tracking consumer’s behavior patterns in-store, retailers are able to optimize store
lay-out, product mix or shelf positioning. Customer micro-segmentation; Big Data analysis has enabled high-end retailers to leverage and track this data on the behavior of individual customers. This increase in sophistication enables retailers to provide custom made offers for individuals rather than the segmentation in which most retailers engage at this point. Sentiment analysis; in the era of social media and increasing online marketing
16
campaigns, sentiment analysis provides retailers the option to gauge a real-time response based on data received from several social media. Consumers are more and more relying on peer sentiment and recommendations of other consumers. Hence a tool to monitor real-time and response to web-based consumer behavior and choices.
Enhancing the multichannel consumer experience; this can be a powerful driver of sales, customer satisfaction
and loyalty. Retailers can integrate promotions and/or pricing for example for shoppers seamlessly. Whether those customers are online, in-store or checking out a catalog.
4.4
Is this what the consumer wants?
Consumers could take this kind of marketing techniques as possibilities to get better information, more
personalized offerings and a better shopping experience. While on the other end is the consumer who might fear intrusion of privacy. These techniques are still in development and are being put to use in a scarce matter. However in the foreseeable future we can imagine this will take a large leap forward, to the point where every retailer will make use of one or multiple techniques.
Big data and the analytics of Big Data are changing the way retailers are making decisions. From the traditional Business Intelligence and Analytics to predictive analytics (Moorthy, Janakiraman et al, 2015). Businesses are keen on this kind of possibilities, it provides them with prospects of higher customer satisfaction and turnover. However consumers might not like the idea of corporations knowing what they want before they do. It sort of takes the aspect of free will out of the equation. So are these ‘grey areas’ combined with high awareness of the consumer acceptable or are consumers not so willingly to accept these predictive analytics and are they more inclined to not provide the information needed to make use of these applications?
5.
Methodology
This paper’s aim is to explore the possibility of the awareness levels of the consumer regarding the collection, usage and analysis of Big Data have an influence on the attitude of the consumer. Next to the level of awareness this paper will try to find support for the moderation of the level transparency companies provide the consumers about the collection of their personal data as well. One could hypothesize that with the Big Data phenomenon on the rise the level of awareness of the consumer will increase as well. Does this awareness have a direct influence on the attitude the consumers have towards four marketing related concepts and does the level of transparency moderate this effect? The design of this study can be captured by the conceptual model (Figure 2) where we can see that the level of awareness (IV) might have a direct effect on the attitude of the consumer regarding the usage of public available data (DV), personalized advertising (DV), usage of the personal data for commercial goals (DV) and usage of the data for the collective good (DV). Furthermore begs the question if the level of transparency (moderator) companies provide about the fact that they collect and usage data moderates this effect on the attitude of the consumer and if the need for transparency the consumer feels (moderator), moderates this effect on the attitude of the consumer as well. To cover all the aspects of the conceptual model the following research question has been formulated:
‘Does the level of awareness the consumer has about the usage and collection of personal data by third parties have a direct effect on the attitude of the consumer regarding the concepts; usage of public available data, usage of the data for personalized advertising, usage of the data for commercial goals, usage of the data for the collective good and does the level of transparency the third party provides the consumer, and the attitude of the consumer towards transparency moderate this effect?’
5.2
Method and data
This study explores the attitude of consumers regarding Big Data, where the most pressing question is, if the level of awareness consumers have about big data gathering, has a direct effect on the attitude towards 4 marketing related big data usages. To find empirical support for the proposed hypotheses in the literature review a new dataset has been created, using an online survey. The survey consisted out of 55 questions measuring the different aspects of the conceptual model. Research for an existing scale which could be used turned up nothing. Hence a new survey has been created. Consumers were presented with statements which would measure their attitude regarding that topic. The questions have been formulated generally to test the consumers awareness and which influence the moderators have on the pending attitude towards the four different attitudes. The survey has been put online via Qualtrics where consumers were able to fill it out via their personal computer, laptop, tablet or mobile phone. Considering the subject the researcher felt an online platform was the most suitable way to set out the survey. Requests to fill out the survey where posted on the personal Facebook page of the researcher and via e-mail. No restrictions other than the minimal age of 18 where required to fill out the survey. No other
demographical statistics then age, gender and highest education where asked to keep the participant anonymous. According to Podsakoff (2003) precautions before the usage of a survey need to be taken. Two of those have been taken into consideration where 1) pretesting the survey has been done within the direct environment of the researcher and 2) ensuring anonymity of the participant. However Podsakoff (2003) also advices including an introduction to the survey, stipulating its purpose. After careful consideration I decided not to include an
introduction as this might influence the current awareness of the participant and hence biasing the results. A total of 94 responses was recorded. A dropout rate of 10% was recorded which reduced the number of useable surveys to 87. After a careful review of the completed surveys another 3 were deleted. Surveys with a completion time above 20 minutes were deleted and surveys with a completion time under 3 minutes were deleted as well. A total of 84 useable surveys remained. This sample consisted out of 43 males (51%) and 41 females (49%). Out of the 84 respondents 65 (76%) were between the age of 18 and 30, 11 (13%) between 31 and 40, 2 (4%) between 41 and 50 and 5 (7%) between 51-65.
19
5.3
Analysis and common method variance
All constructs have been computed into new variables which are mean centred in order to use them in a regression analysis. The Central Limit Theorem states that when samples are large (N>30) the sampling distribution will take the shape of a normal distribution regardless of the shape of the population from which the sample was drawn (Field, 2013). Hence normality of the data is assumed. A brief look at the QQ plots and skewness and kurtosis descriptives showed that the data points from the constructs were not deviating to much from the line of normal distribution. However kurtosis and skewness descriptives showed some variance. This can be explained by the scale that has been used. Due to the fact that a Likert-scale ranging from 1 to 5 has been used, no data entry’s lower than 1 or higher than 5 can be entered, which explains the lack of ‘tails’. No transformations to the data have been done. A residual analysis has been conducted to ensure no harmful outliers were kept within the dataset.
6
. Results
6.1
Measurements
A 5-point likert scale was used in the survey to measure the influence awareness and the different constructs have on the attitude of the participant. The use of a Likert scale should decrease the time spend by the participant to fill out the survey, which should increase the response rate (Jamieson, S. 2004). The likert scale ranged from (1) ‘very negative’ to (5) ‘very positive’. Different answer options were created to match the questions, however the distance between the answers remained the same, with ‘neutral’ at answer option 3. All the constructs and items can be found in the appendix of this study. The output of 23 counter indicative questions were recoded in order to compute new variables. All the different constructs were tested for sample adequacy with the Kaiser-Meyer-Olkin measure and by looking at the factor loadings. According to Guadagnoli and Velicer (1988) if a factor has 4 or more loadings at least or greater than .6 it is reliable no matter what the sample size. To provide evidence for the measurements and its reliability the different constructs where tested separately for the Cronbach’s Alpha to see if the items were actually measuring their intended measurement.
A principal component analysis with varimax rotation has been conducted on the dataset to determine the constructs with statistical argumentation. According to Field (2013) factors with loading above .4 can be considered to include in the construct. Considering the newly developed scale I am using, a higher factor score seems adequate. Hence all factors within the constructs will have a loading of .6 or higher.
20
Initial analysis showed that a total of 13 components had Eigenvalues above 1, which together explained a total of 73,14% of the variance. However after careful analysis of the loadings on the components only 9 constructs were extracted. The remaining 4 components were discarded due to low factor loadings.
The first construct was labeled Internet activity. This construct was (in contrast to the rest of the constructs) not mean centered. After reviewing the variables belonging to the construct I decided to add the scores of the variables to measure the intensity of the internet activity. Reliability analysis scored low α=.551. The construct
Awareness ‘fingerprint’ was created to measure the awareness of the consumer regarding the fact that they leave
behind traces of digital personal information, a so called fingerprint. Principal component analysis showed 3 factors which loaded highly on one scale. The KMO measure resulted in a .698 which according to Hutcheson and Sofroniou (1999) is a middling score. Reliability analysis resulted in α=.816. Awareness ‘collection’ refers to the awareness of the consumer regarding the fact that third parties are collecting the data they leave behind. The construct consists out of 5 factors which all loaded clearly on one scale. The KMO measure resulted in a
meritorious .825 and α=.856. Attitude towards transparency measures the attitude of consumers regarding the fact that data is being collected without their prior consent. Factor analysis showed that all 4 items loaded clearly on one scale with no factor loadings below .7. The KMO measure resulted in .746 and α=.848. The fifth construct was labeled Level of transparency and measured the level of transparency the consumers experience from companies that collect their digital data. Analysis showed that all factors all loaded clearly on one scale and all had high factor loadings. The test for sample adequacy resulted in a KMO measure of .620 and reliability analysis showed that α=.696. The last 4 constructs are all measuring attitudes of the consumer regarding different aspects of Big Data usage. They operate as depending variables in the conceptual model. Attitude ‘usage personal data’ was created to measure the attitude of consumers regarding the usage of personal data openly, like their profile picture in an advertisement. Factor analysis showed that both items loaded clearly on one scale, the KMO measure was a mere .5 however together the items explained over 81% of the variance. Hence the construct was left intact. Reliability analysis resulted in α=.773. Attitude ‘personalized advertising’ is a construct created to measure the attitude of the consumer regarding the usage of personal data in advertising tailored for their specific needs. The question asked is how do consumers feel about the fact that their personal data is being used to offer them advertisements which fit perfectly within their interests. Analysis showed that all factors had loading above .7 and loaded clearly on one scale. The KMO measure was a robust .885 and α=.898. Attitude ‘commercial use’ measures the attitude consumers have regarding the use of data used for commercial means. Factor analysis showed that all items loaded clearly on one scale and the KMO measure was a good .820. Reliability analysis resulted in a comforting .826. That last construct is labeled Attitude ‘collective good’ and measures the attitude of the consumers regarding the usage of data for the collective good. All factors loaded clearly on one scale with all items above .7. The KMO measure resulted in a mere .5, but is still manageable according to Kaiser (1974). The Eigenvalues matrix showed that both items belonged to one component and explained 78% of the variance. Reliability analysis resulted in α=.728. Hence the construct was left intact.
21
6.2
Descriptive statistics and correlations
The descriptive statistics and intercorrelations for each of the study’s constructs are presented in Table 1 and 2. To check if multicollinearity is an issue, variance inflation factors (VIF) have been calculated. The highest value recorded was 2,170 and the average VIF was 1,100. Both scores indicate no cause for concern regarding
multicollinearity. According to Bowerman & O’connell (1999) if the highest value is below 10 and the average VIF score is not substantially greater than 1 the regression analysis should not be biased. Looking at Table 1 we can observe that the average score of awareness ‘fingerprint’ is 3,93 and the average score of awareness ‘collection’ is 4,40. Both scores indicate that the participants are highly aware of the fact that they leave behind traces of digital information and that third parties are actively collecting them. Level of transparency scores a 2,61 which indicates that the level of transparency experienced by the participants actually varies. We can see that the minimum score is 1, referring to no transparency at all and the maximum score is 5, referring to total transparency. The average levels out almost in the center. Looking at attitude towards transparency we see that the average is 1,96,
indicating that the participants do slightly favor transparency provided by the third party. The reason that a lower score is a more favorable one is due to the formulation of the questions in the survey. Participants were asked how they would feel if for example Google would be collecting their data, every time they used the search engine, without notifying them. Here the lowest score (1) would represent ‘very negative’, hence the consumer does not appreciate that and more transparency would be favorable. The last four rows display the scores of the different attitude constructs. Looking at the attitude ‘public available data’ we can observe an average score of 1,69 indicating a negative tendency towards the usage of public available personal data. Attitude ‘ personalized advertising’ scores 2,64 on average which leans towards a neutral stance. Participants are almost indifferent about the fact that personal data is being used for personalized advertising. Following up on attitude ‘commercial goals’ we can observe the same effect, with an average score of 2,59 participants are almost neutral about the fact that their personal data is being used for commercial goals. The last attitude, referring to usage of personal data for the collective good, has an average of 3,36 indicating that the participants are slightly positive about the usage of their personal data for the collective good.
Variable
N
Minimum Maximum
Mean
SE
Awareness 'fingerprint'
83
1,33
5,00
3,93
0,85
Awareness 'collection'
83
1,80
5,00
4,40
0,63
Level of transparency
83
1,00
5,00
2,61
0,87
Attitude towards transparency
83
1,00
5,00
1,96
0,80
Attitude 'public available data'
83
1,00
5,00
1,69
0,87
Attitude 'personalized advertising'
83
1,00
5,00
2,64
0,94
Attitude 'commercial goals'
83
1,00
4,60
2,59
0,84
Attitude 'collective good'
83
1,00
5,00
3,36
0,90
22
Table 2 represents the correlation matrix. Regarding hypothesis 1 we see no significant relationship between awareness constructs and the attitude towards the usage of public available personal data (r=.169, p>0.05)(r=-.160, r>0.05). This also seems to be the case for hypothesis 2 where awareness ‘fingerprint’ shows a non significant correlation (r=-.188, p>0.05) and as well between awareness ‘collection’ and the construct attitude towards personalized advertising (r=.021, p>0.05). In search for support for hypothesis 3 we observe a positive significant relationship between awareness ‘fingerprint’ and attitude ‘commercial use’ (r=.269, p<0.05). This seems to build some support for hypothesis 3. However the construct awareness ‘collection’ has a non significant correlation with the attitude ‘commercial use’ (r=.159, p>0.05). In the search for support for hypothesis 4 we do not observe any significant relationships between awareness and attitude ‘collective good’ (r=-.007, p>0.05) (r=.023, p>0.05). Regarding hypothesis 5 we do not observe any significant relationships between the level of transparency and the different attitude constructs (r=.054, p>0.05)(r=.029, p>0.05)(r=.040, p>0.05)(r=-.020, p>0.05). However the construct attitude towards transparency, which will be tested for its mediating effect, does show strong significant correlations with the constructs attitude ‘usage personal data’ (r=.466, p<0.01), attitude ‘personalized advertising’ (r=.546, p<0.01), attitude ‘commercial use’ (r=.728, p<0.01) and attitude ‘collective good’ (r=.369, p<0.01). Further regression analysis will need to provide further information about the relationships between these constructs. If we take a look at the correlations outside the scope of the formulated hypotheses we do observe strong significant relationships between the attitude constructs. Attitude ‘usage personal data’ shows strong significant relationships with attitude ‘personalized advertising’ (r=.294, p<0.01) and attitude ‘commercial goals’ (r=.298, p<0.01).
Furthermore attitude ‘personalized advertising’ shows a positive significant relationship with attitude ‘commercial use’ (r=.700, p<0.01) and attitude ‘collective good’ (r=.310, p<0.01). Finally we can see a positive significant relationship between attitude ‘commercial goals’ and attitude ‘collective good’ (r=.490, p<0.01).
Table 2: Means, Standard deviations, Correlations and Scale Scores
*. Correlation is significant at the 0.05 level (2-tailed) **. Correlation is significant at the 0.01 level (2-tailed)
6.3
Hypothesis testing
Hierarchical multiple regression was performed to investigate hypotheses 1 to 4. Each of the regression models has been split up in 4 steps. Where the first 3 steps assesses the effect of single and combined variables on the (attitude). In the last step I assessed the model as a whole. Meaning controlling for age, gender and education. Hypothesis 1 investigates the effect of awareness on the attitude of the consumer regarding the usage of public available data, like a check-in on Facebook or photo’s on Instagram. Looking at table 3, model 1 we observe no significant effect of awareness ‘fingerprint’ on the attitude, this is the same for model 2 where awareness ‘collection’ also has no significant effect on the attitude. However If we look at model 3 we can see that awareness ‘fingerprint’ has a positive significant effect on the attitude of the consumer (β= .246, p<0.05). Which means that being better aware of the fact that we leave behind small pieces of information about ourselves has a positive effect on our attitude towards the usage of this information, which is already out there and thus publicly available. One could reason that the consumer who is more aware might feel that they have put the information out there and thus do not care if the third party is making use of that information. Looking at the same model we can see a negative significant effect of awareness ‘collection’ on the attitude of the consumer (β=-.239, p<0.05). This indicates that the awareness of the consumer regarding the fact that third parties are actively collecting data, influences the attitude of the consumer negatively. Hence the consumer with a higher awareness about the fact that third parties are collecting their data are being poked. It could be that the realization of third parties who are actively searching for the breadcrumbs of information gives the consumer a feeling of privacy intrusion. The full model (model 4) as a whole shows no significant effect on the attitude of the consumer. Hence none of the control variables have a significant effect on the attitude.
Hypothesis 2 explores the effect of awareness on the attitude of the consumer regarding the usage of personal data for personalized advertising. Consider using personal background information to be able to offer you only the things you are interested in, does awareness has an influence on this attitude? Looking at table 4, model 1 we observe no significant effect on the attitude, this holds also for model 2, 3 and even the full model. None of the entered independent variables have a significant effect on the attitude. Looking closer at model 4 (full model) we can see that one of the control variables (gender) does have a significant effect on the attitude (β=-.228, p<0.05). This indicates that being a male or female makes a difference in attitude, meaning that one of the sexes, feels differently about personalized advertising with a higher awareness, compared to the other gender.
Hypothesis 3 has been formulated to explore the possibility of awareness having a direct effect on the attitude of the consumer regarding the usage of the personal data for commercial goals aimed solely for the third party itself. This differs from personalized advertising in the sense that consumers might benefit from personalized advertising as well. Imagine that a third party collects data from his customers, and analyzing this data to be able to increase commercial effectiveness, like for example profit margin or sales per square meter. Looking at table 5, model 1 we can observe a positive significant effect of awareness ‘fingerprint’ on the attitude (β=.269, p<0.05).
25
Indicating that a consumer who is more aware of the fact that they leave behind digital breadcrumbs is more positive towards the usage of their data for commercial goals of the third party. Hence when the consumer knows that the data they leave behind is being used for commercial goals, the consumer finds this more acceptable in regard to being left in the dark. After all the consumer is the one who left this information behind in the first place. Model 2 shows no significant effect of awareness ‘collection’ on the attitude. Hence knowing for a fact that third parties are collecting you data does not influence the consumers attitude. Progressing to model 3 we can again see a significant effect on the attitude. The model as a whole explains 7,8% of the variance of the attitude. Looking at the variables we can see that again the awareness of leaving behind pieces of information has a significant effect on the attitude (β=.243, p<0.05) but that the awareness of third parties collecting these pieces of information lacks in explanatory power in the model and shows no significant effect on the attitude (β=.081, p=n.s.). Finally if we look at the results in model 4 (full model) we can see that the model as a whole explains 21,5% of the variance in the attitude and is significant at the 0.01 level (p=.002). Looking closer we can see that the independent variables do not carry any of the explanatory variance in the model but that the control variables age (β=-.260, p<0.05) and gender (β=-.264, p<0.05) both have a significant effect on the attitude. One could reason that age has an influence in the sense that the changing society leaves different age groups at different perspectives. Following up on the gender one could reason that the one type of gender is more easily offended by exploiting personal data for commercial goals than the other. However with the current data an explanation with statistical evidence is not an option.
Hypothesis 4 has been formulated to examine the effect of awareness on the attitude of the consumer towards usage of the personal data for the collective good. One could imagine that a third party is collecting personal data and using this to improve facilities within a city (which would benefit all the individuals living in it) or a governmental institution is gathering data to detect terrorist activity. Looking at table 6, model 1 we observe no significant effect of awareness ‘fingerprint’ on the attitude. This also holds for model 2, 3 and even model 4 (the full model). Therefore we do not find support for hypothesis 4, since awareness does not seem to have a direct effect on the attitude of the consumer regarding the usage of personal data for the collective good. Again we observe no significant effect on the attitude when the attitude seems to refer to a beneficial situation for the consumer.
To further investigate the role of awareness on the attitude of the consumer I used a model provided in PROCESS (Hayes, 2013). The PROCESS modification to SPSS allows me to test the moderating role of the level of
transparency and the attitude towards transparency on the four different attitude constructs. In tables 7 and 8 we can observe 8 different models. Each model represents the moderating effect on the different attitude constructs. Hypothesis 5 states that awareness has a direct effect on the attitude of the consumer, however this effect is moderated by the level of transparency a third party provides the consumer regarding the their efforts on collecting personal data which is publicly available. As an example one could image that a company who provides their customers with loyalty cards makes it very clear they use this data for other purposes, like determining if the
26
consumer belongs to a certain group with specific shopping patterns or to determine if their product diversity is to large or too small for that specific group. Hence they are transparent towards their customers. Looking at table 7, model 1 we can see that the level of transparency does not significantly moderates the effect of awareness ‘fingerprint’ on the attitude of the consumer regarding the usage of public available personal data (B=.018; p=n.s). The same model shows that level of transparency does not significantly moderates the effect of awareness ‘collection’ on the same attitude (B=.137; p=n.s.). The non-significant result holds for all the models testing the moderating effect of level of transparency on the different attitudes. Transparency as such does not seem to make a difference in the attitude of the consumer in this situation. One could reason that is due to the fact that the data that is being collected is already out there and thus transparency becomes obsolete. Model 2 shows both types of awareness are not significantly moderated by level of transparency on the consumers attitude towards
personalized advertising (B=.044; p=n.s.)(B=-.280; p=n.s.). Model 3 shows both types of awareness are not significantly moderated by the level of transparency on the attitude of the consumer towards the use of personal data for commercial goals (B=.031; p=n.s.)(B=-.144; p=n.s.) and finally model 4 also shows both types of awareness are not significantly moderated by the level of transparency on the attitude of the consumer towards the usage of personal data for the collective good (B=.006; p=n.s.)(B=.043; p=n.s.). The level of transparency provided by the third party does not seem to moderate the effect on the consumers’ attitude. One could reason that a higher awareness provides the consumer with enough transparency already to assess the situation that extra transparency provided by the third party renders insignificant.
Hypothesis 6 states that the direct effect awareness has on the attitude of the consumer is being
moderated by the attitude of the consumer towards transparency. One could image that the consumer who has a negative attitude regarding transparency, i.e. they do not like the fact that third parties are collecting their personal data without making this transparent, has a more negative tendency towards the attitudes of the four marketing related concepts. Looking at table 8, model 1 we can observe no significant moderating effect on the linear connection between awareness ‘fingerprint’ and attitude ‘usage of public available personal data’. However the linear connection between awareness ‘collection’ and attitude ‘usage of public available personal data’ is being significantly moderated by attitude towards transparency (B=-.354, p=<0.05). Where a consumer who has a more negative attitude towards transparency has a more negative attitude towards the usage of public available data. Hence disliking the fact that data is being collected without transparency has a negative influence on how the consumer feels about the usage of their data which is already out there and publicly available. If we take a stab at model 2 we can see that both types of awareness are not significantly being moderated by the attitude towards transparency (B=-.199; p=n.s.)(B=-.067; p=n.s.). Looking at model 3 and 4 we can see that in both models the direct effect is not significantly being moderated by attitude towards transparency (B=.181; p=n.s.)(B=.244; p=n.s.). Concluding we can state that the attitude of the consumer towards transparency, does not seem to moderate the effect awareness has on their attitude. It seems to be the case that how the consumer feels about transparency does not seem to make much of a difference with the exception of the usage of public available data.
27
H4:
Awareness Attitude 'collective good'
Rejected
Awareness Attitude 'usage of public available data'
Accepted
Awareness Attitude 'personalized advertising'
Rejected
H3:
Awareness Attitude 'commercial goals'
Accepted
H1:
H2:
Level of transparency
Awareness Attitude constructs
H5:
Rejected
Attitude towards transparency
Awareness Attitude constructs
H6:
Rejected
One could reason that the public available personal data is more private (photo’s, check-in, preferences) than the other types of data provided in the attitude constructs, which are more general (shopping patterns, interests). Hence usage of the first feels more as privacy intrusion than the latter, where the attitude towards transparency plays a role. A consumer who feels that a third party should not be collecting their data without prior consent, and who is aware of the fact that the third party is doing so, has a negative stance towards the usage of this data. Presented below in figure 3, 4 and 5 are the hypotheses with the accepted or rejected states, based upon the results.
Figure 3: Graphical representation of the hypotheses
Figure 4: Graphical representation of hypothesis 5
Hypothesis 1
Model 1
Model 2
Model 3
Model 4
B
SE
β
p
B
SE
β
p
B
SE
β
p
B
SE
β
p
Independent variables
Awareness 'fingerprint'
.173 .112 .169
.127
.252 .116 .246 .033
.244 .123 .238 .051
Awareness 'collection'
-.221 .152 -.160 .149
-.331 .157 -.239 .038
-.361 .160 -.260 .027
Control variables
Age
-.131 .119 -.120 .277
Gender
-.107 .200 -.061 .594
Education
-.057 .111 -.057 .607
R²
.029
.026
.080
.102
F
2,380
2,121
3,462
1,742
p
.127
.149
.036
.135
Hypothesis 2
Model 1
Model 2
Model 3
Model 4
B
SE
β
p
B
SE
β
p
B
SE
β
p
B
SE
β
p
Independent variables
Awareness 'fingerprint'
.206 .120 .188
.088
.222 .127 .202 .085
.164 .131 .150 .213
Awareness 'collection'
.032 .164
.021 .848
-.065 .172 -.044 .707
-.131 .170 -.088 .444
Control variables
Age
-.167 .127 -.143 .193
Gender
-.426 .212 -.228 .048
Education
-.048 .118 -.045 .684
R²
.035
.000
.037
.109
F
2,978
0,037
1,544
1,891
p
.088
.848
.220
.105
Table 3: Regression results
