National Information Markets
Exploring markets for personal information
University of Amsterdam Master thesis Economics – Behavioural Economics and Game Theory Richard Zevenbergen 10642439 Supervisor: dr. Audrey Xianhua Hu“[...] we began to consider an alternative understanding of the “privacy paradox” that is about far more than people’s cost-benefit analysis or their lack of knowledge. It is about citizens’ belief they have no agency in a central area of democratic society: commerce.” - Turow, Hennessy, & Draper (2015) The tradeoff fallacy
1. Introduction ... 4
2. Model Formulation ... 7
2.1 The Kreps-Scheinkman Model ... 72.2 The National Information Market ... 8
2.3 The NIM Model ... 10
2.4 Application to the market ... 11
3. Behavioural Economics and Privacy: Limits to Rationality ... 13
4. Survey ... 15
4.1 Method ... 154.2 Limitations ... 15
5. Analysis of Results ... 17
5.1 Characteristics of respondents ... 175.2 Variables ... 19
5.3 Comparison to survey carried out by Turow et al ... 20
6. Application of results to the model ... 25
6.2 The Bounded Rationality Model ... 256.3 The Bounded Rationality Model with many individuals ... 27
7. Conclusion ... 28
8. Appendix ... 30
9. Bibliography ... 36
1. Introduction
In 1996 Kenneth Laudon proposed setting up a market in which individuals can sell information about themselves to organizations that wish to collect this information. He wrote his paper at a time when the United States was building the so-called Information Highway; the legislative and infrastructural basis that led to the internet boom of the 1990s. In light of these advancements there was little control over the collection of information and legislation was trailing far behind the privacy issues of the time. According to Laudon at the time “existing laws and their conceptual foundations have become outdated because of changes in technology.” (Laudon 1996, p. 92) Instead of protecting the individual in a period of rapid technological change, Laudon proposed a method to empower the individual instead, namely through a National Information Market. In this market people are the producers of their information which allow limit what is shared about them. The boundaries between when data-collection is acceptable and when it is not can be set by the same laws that govern any intellectual property, thus allowing the individual to persecute those who infringe upon their right to privacy. In short, a National Information Market such as the one that Laudon proposes serves as a tool to empower the individual in a system where “the cost of invading individual privacy is far lower than the true social cost of invading that privacy” (idem, p. 93). Now, twenty years later, the individual’s vulnerability with regards to privacy rights is as salient as ever before. In some parts of the world nearly 90% of the world’s population is online1.
As has become clear in light of the revelations done by Edward Snowden2 in 2013 a lot of the online traffic is being surveyed by governments. This surveillance is committed under the banner of domestic security. However, a lot more information is voluntarily given by individuals to organizations that offer online services. In practice, individuals are sharing a lot more information online than they claim to want to share. In the academic world this is known as the privacy paradox: individuals claim to be wary of organizations that harvest their information, but in practice they give a lot of information away for free. The question that everybody is asking is; if you do not want to share this much information then why are you giving it away? 1 Between 2000 and 2015 the number of internet users online increased by 832.5% (http://www.internetworldstats.com/stats.htm 2016), and we now find ourselves in a situation where nearly half of the world’s population is connected. 2 The revelations affirmed suspicions by privacy organizations, and surprised people who were not familiar with these practices. However, the breaches of privacy have done little to alter behaviour. For more information about 2 The revelations affirmed suspicions by privacy organizations, and surprised people who were not familiar with these practices. However, the breaches of privacy have done little to alter behaviour. For more information about the revelations visit http://www.theguardian.com/world/interactive/2013/nov/01/snowden-nsa-files-surveillance-revelations-decoded#section/1.
Organizations that collect information, such as search engines and social media platforms, argue that the individual is getting a service in return for their information (Turow et al 2015). Therefore this can be considered a form of barter, in which services are being traded for data. A problem still remains though, because individuals share a lot more information with these organizations than they claim to consider fair in return for the service. So then the problem might be that individuals are being misled about the information that is being collected and what is done with that information. Turow et al (2015) show that this is not the case. In fact, according to their research, individuals who are better informed about what happens with their information are more likely to be susceptible to the privacy paradox. The authors conclude that people are powerless to control the dispersion of their information and therefore submit to the fact that they cannot control it. Oulasvirta et al (2012) find that once individuals have revealed something about themselves by accident or through carelessness they will take less care to reveal that thing about themselves in the future3. Perhaps people are still getting used to interacting online and sometimes regret their actions in hindsight.
Regardless of the explanation for this phenomenon it is clear that individuals do not feel that their information is properly protected online. Since Laudon’s proposal for a National Information Market data-collecting technology has advanced considerably while there do not seem to have been any revolutionary improvements in the protection of individuals’ privacy. The reason for this is not that the topic has received little attention, but that there is little consensus. Any proposal that aims to give individuals more control over the dispersion of their information deserves to be assessed critically. In this thesis I aim to investigate the individual shoppers’ relationship with privacy-sensitive information and to determine whether this allows for a functioning National Information Market (NIM). The question that I want to answer is whether it will ever be conceivable to put individuals in the role of a supplier of their own information. This question can be divided into two sub-questions, namely; (1) is the transition from a universal human right to a market commodity a feasible option?; (2) are individuals capable of acting rationally enough to benefit from taking on the role of suppliers of their own information? Answering these questions requires a baseline case to which the results can be compared. I take the NIM model as the baseline case, and I have cast it into the form of an economic model with the help of 3 The experiment involved only twelve participants, but the findings are fascinating nonetheless because of how invasive the setup was and the long period over which people were recorded. Subjects agreed to have microphones and video cameras installed in their homes, and their use of computers, smartphones, wireless networks, TV, DVD, and customer cards was logged. Subjects experienced stress as a result of having to take into account that they were being watched, and some behaviours people wanted to keep hidden but could not, thus resulting in acceptance.
Scheinkman model is and how Laudon described the NIM. Building on these descriptions I combine the two to form the baseline case. After this I will try to determine what happens when people are forced to think about their privacy-sensitive information in economic settings. Some people argue that people already face this decision when using search engines, social media, and loyalty cards in stores, and therefore subjects will be familiar with the situation that I aim to sketch and thus be able to answer from experience. In the final section I will try to determine whether people are willing (and capable) to act as suppliers of their own information.
2. Model Formulation
2.1 The Kreps-Scheinkman Model
The model presented by Kreps and Scheinkman in 1980 is a two-period dynamic game with two firms. In the first period firms simultaneously set quantities, that is, they choose the maximum quantity (qi) that
they will be able to produce with no extra cost in the second period. The cost of installing the required facilities is b(qi). In the second period firms compete on prices (pi). The amount that a firm sells given the
relative prices is determined by the following function
min (qi, D(pi)) if pi < pj
zi = min (qi, max (D(pi)/2, D(pi) - qj)) if pi = pj
min (qi, max (0, D(pj) - qi)) if pi > pj
This shows that firms are limited in meeting demand in the second period by the capacity that they installed in the first period, which alters the results of the normal Bertrand model. So if demand exceeds capacity then the firm will only produce up to capacity and vice-versa. As this shows, firms may supply less than in a perfectly competitive equilibrium and therefore the outcome is not always the competitive outcome of the Bertrand model. Profits are given by πi(pi) = pizi - b(qi) and market demand is given by the inverse demand function P(q), whereby D(p) = P-1(p). Both the inverse demand function and the cost function are assumed to be strictly decreasing, concave4, and twice-continuously differentiable, while the cost function satisfies b(0) = 0 and b’(0) > 0. Also, b’(0) < P(0) so that it is appealing for firms to produce more than nothing. Kreps and Scheinkman show that in this two-stage game the Cournot outcome, where q1 = q2 = q*(b) and p1 = p2 = P(2q*(b)), is the unique equilibrium outcome. Theoretically this result is interesting because it
seems to explain the way in which prices are set in Cournot models. However, this result is also interesting for an analysis of Laudon’s national information market, because in essence that is also a two-stage dynamic game with quantity being set in the first period and prices being set in the second period. 4 “Concavity is sometimes invoked to guarantee that a firm’s optimum is characterized by the first-order condition associated with its profit-maximization problem. It is also used in oligopoly models to guarantee that firms’ reaction functions have the “right” inclination (either upward sloping or downward sloping).” Malueg (1994) p. 15
2.2 The National Information Market
Laudon proposes an information market that allows individuals and organizations to trade in individuals’ personal information, whereby organizations using personal information (OUPIs) pay individuals for the right to use their information for a certain amount of time. According to Laudon such a market gives individuals more control over the dispersion of their personal information.
There are two avenues by which information can be traded in this market. The first is through a National Information Exchange (NIE) that functions just like a stock exchange. Banks and other financial institutions allow individuals to deposit their information in information accounts which consist of two parts; information-assets and information-rights. The former can be any information about the individual, such as name, gender, medical details, etc. The latter delimits what can be done with this information. Groups of information accounts are then split up and bundled into baskets that can be traded on the information exchange.
Prices are determined by interactions between buyers and sellers on the exchange, just like on a regular stock market. However, there is a difference in the way that prices are set initially when compared to regular stock markets. While in an initial public offering an investment bank determines the value of the stock, in this market the price will be based on the OUPIs’ anticipated future revenues that each basket represents5. This means that the buyer calculates the value of the basket and determines a price at which it is willing to buy. The behaviour of the OUPI’s that wish to purchase information is captured in the demand function. Sellers, the LIBs, try to maximise profits when selling the baskets. Their behaviour is therefore defined by their individual optimisation problem. If we stay with the stock market comparison then LIBs must charge a fixed rate over every transaction, just like their counterparts in the stock exchange, the stock-broker. It is also possible for individuals and organizations to engage in private transactions. Private placement sales would be regulated by the government-run National Information Accounts Clearinghouse (NIAC). This is compared to the system in the music industry that determines the fees that artists are entitled to, based on use of their music in advertising, films, etc. For example, in Europe a songwriter is entitled to ⅔ of revenues6. An OUPI that wants to sell information to other companies or that wants to use the information for secondary purposes for a fee can get permission from the individual. If the OUPI wants to sell information it can do this on the NIE and therefore competes with LIBs for the available demand. 5 Laudon does not specifically mention how initial prices are set, but I gather this from the following sentence; “Information-using organizations would offer to buy the baskets of information at a price based on the anticipated future revenues each basket represented.” 6 Note that due to considerable market power of the collecting companies songwriters actually receive much less than this (Kooistra 2015).
Laudon also states that organizations will be able to use the information that they have purchased for a limited amount of time. This is different the way in which information is ‘transacted’ now, because once information has been collected it can usually be stored indefinitely. Therefore, for every time a certain piece of information is sold the individual incurs a cost. One can interpret this as though the good is exclusive or as though the individual incurs the same cost of sharing the same piece of information every time it is sold. This is the complete description of the information market as proposed by Laudon. However, there are still some problems with this proposal. Firstly, how far are we to go with Laudon’s comparison of the stock exchange and the NIE? The initial value of a stock is normally calculated by an investment bank7. In the information market the initial price is set by the first offer from an OUPI that is supposed to base its offer on the expected revenues from advertising. Obviously the OUPI that is buying the basket will have a good reason to offer less than its valuation of that good. Since LIB’s are subject to the price that is negotiated on the NIE there is a strong incentive for OUPIs to collude and to keep the value of baskets low. I believe that there is reason to worry about this because there are a few OUPIs with considerable market power in the real world and a small number of buyers will be more likely to lead to an oligopsony.
Secondly, there is a problem with setting a defined period in which information is to be used. In the real world we know that it is extremely difficult and costly for a company to protect data against hackers, and that firms and governments have repeatedly been caught violating individuals’ privacy rights. Laudon proposes that an individual’s National Information Account be assigned “a unique identifier number and barcode symbol” (Laudon, 1996, p. 100). In this way, any and all information can be tracked. If information is used without consent of the individual then this is a crime and offenders will be prosecuted. This ex-post apprehension of an offender does not protect the individual beforehand. However, the infringement is no longer a case of an invasion of privacy, but one of foregone earnings, since the individual already consented to sharing her information. If it is possible to completely track information then this should prevent companies from illegitimately using an individual’s information. What follows is a description of the actors in the market and their interactions, which result in a two-stage dynamic game. The model I use is based on the Kreps-Scheinkman model, but there are several differences. This is why one cannot immediately assume that the market will have a unique Cournot equilibrium.
7 There are exceptions. For example, Google set their IPO value through a combination of a Dutch auction and a valuation (Levy 2014)..
2.3 The NIM Model
For the moment, I will assume that the market consists of four actors, namely two individuals, the LIB, and the OUPI. In the first period individuals determine how much information to sell to LIBs and to OUPIs. If an individual sells information to an LIB then she earns (1 - hiLIB)P(qiLIB + qiOUPI)qiLIB - xiqiLIB where
the amount of information that the individual sells to the bank is given by qiLIB. hiLIB represents the
handling costs of the LIB that it charges to individuals for processing and selling their information and 0 < hiLIB < 1. The cases where h = 0 or h = 1 are not included because then either the LIB’s revenue is zero or
the individual’s revenue is zero and then they will have no incentive to participate in this market. xi
represents the individual’s desire to keep that information private8, expressed in monetary terms. An individual that sells information to an OUPI earns (1 - α)P(qiLIB + qiOUPI)qiOUPI - xiqiOUPI. The OUPI keeps αP
and the individual therefore receives (1-α)P. α is the fraction of the revenue that the OUPI is entitled to by law. It can be any number between 0 and 1 and is determined according to some fairness principle. The LIB sells baskets of information on the NIE and gives the money that is earned on the exchange to the individual, minus the handling costs. For the sake of analysis I make the assumption that handling costs are set as a fraction of the price and not as a fixed fee. This makes comparison between the OUPI and the LIB easier because the profit functions are similar. The LIB’s profit function is hipiqLIB - Ci(qiLIB),
whereby the firm sets pi, assuming that hi has already been determined. The restrictions on hi are hi =< 1
- xi/p so that the individual does not make a loss and hi >= ci/p so that the bank also does not make a loss.
The quantity has already been determined by the individual in the first round.
The OUPI makes a profit of αpiqiOUPI - Ci(qOUPI). The organization sets the price so as to maximize this
profit function, whereby α is predetermined.
8Several attempts have been made to determine this value, but it has proven to be extremely difficult. One of the main arguments of those who oppose of the commercialization of privacy is that people are not able to view privacy in terms of currency in any coherent way. People are therefore likely to be exploited.
2.4 Application to the market
If α = hiLIB and Ci(qOUPI) = Ci(qLIB) then this model and the Kreps-Scheinkman model are identical except for
one aspect. In the Kreps-Scheinkman model each firm chooses the quantity and the price, and the resulting equilibrium outcome is the Cournot outcome. If both firms set quantity equal to the Cournot quantity then neither firm can do better by unilaterally deviating. However, in the National Information Market the decision in the first round is made by the individual instead of the firm. The individual that chooses quantity in the first round does not have to take into account the profits that the firm will stand to make given the quantity produced9.
Does this affect the equilibrium outcome? Given certain quantities in period 2 the OUPI and LIB will compete on prices. Kreps and Scheinkman (1983) show in Lemma 2, 3, and 4, that if the two firms are identical and the quantity produced is at most the best-response quantity then each firm will name price pi = P(q1 + q2). Knowing this individuals maximize their profits in the first round. The optimization
problem for the individual that sells to the LIB is
max(qLIB) π
LIB = (1 - hiLIB)P(qiLIB + qiOUPI)qiLIB - xiqiLIB
and given the assumptions of the simple model the optimization problem of the other individual is similar, namely
max(qOUPI) π
OUPI = (1 - hiLIB)P(qiLIB + qiOUPI)qiOUPI - xiqiOUPI.
The best response function of each individual is the solution in qi of
(1 - h)P’(qi + qj)qi + (1 - h)P(qi + qj) - xi = 0.
For any cost function, xi, the best-response of the individual is lower than the best-response function in
the standard Cournot setting, due to the handling costs h. This means that an equilibrium in this market will result in lower production than in a standard Cournot setting with a single firm making the decisions in each round. By separating the task of producing and selling goods and assigning each task to a separate entity the amount of production is reduced in the first round. In my model I assume that demand is linear because this greatly simplifies calculations and it is unlikely that the NIM is characterised by a convex demand function. With a simple demand function, where P(qi + qj) = a - b(qi + qj), the best response function for each type of individual is 9 Given that profits must not be negative.
qi* = (a - xi/(1 - h))/2b - qj/2. Since both types of individual face a similar best response function we can substitute for qj10 in the above function to get qi* = [a - 2xi/(1 - h) + xj/(1 - h)]/3b. The total market demand is Q* = qi* + qj* = [2a - (xi + xj)/(1 - h)]/3b and price is p* = [a + (xi + xj)/(1 - h)]/3.
Overall, the effect of an increase in privacy sensitivity (measured by the costs xi and xj) of either player
results in a decrease in total equilibrium output, given that the other player’s privacy sensitivity is unaffected. Privacy sensitivity has a positive effect on the equilibrium price. However, an increase in the other player’s privacy sensitivity results in an increase in the equilibrium output of the first player. If the game is symmetric then both types of players incur the same cost from selling information, that is, xi = xj = x. One can then subtract the cost functions from each other in the best-response function to get qi* = [a - x/(1 - h)]/3b, Q* = [2a - x/(1-h)]/3b, and p* = [a + 2x/(1-h)]/3. Thus far the rational-actor model. Now I will introduce behavioural deviations to see how people might really act in this market. 10 Where q j* = (a - xj/(1 - h))/2b - qi/2
3. Behavioural Economics and Privacy: Limits to rationality
The first problem is that the individual is faced with incomplete information, because there is no way in which they can take into account all of the mathematical complexities that are required to determine privacy costs, and to subsequently weigh these against the benefits. In comparison, a producer of a traditional good, such as shoes, does not care very much what happens to the product once it has been sold. Whether people share shoes or mistreat them is not taken into account when determining the price of the shoe. But an individual cares very much about what happens to her information once it has been sold. All eventualities need to be taken into account when setting the price and it has been found that “[t]he complexity of the privacy decision environment leads individuals to arrive at highly imprecise estimates of the likelihood and consequences of adverse events, and altogether ignore privacy threats and modes of protection” (Acquisti & Grossklags 2007, p. 4). In a survey the same authors found that people even forget that banks are also informed about their financial transactions during transactions with a credit card (Acquisti & Grossklags 2005, p. 31). So obviously the complexity and the scope of privacy issues will make it very difficult for individuals to determine what the value is of their information. However, this does not mean that people are incapable of successfully becoming information-suppliers. Even without an exact knowledge of this value, prices may be set by the “invisible hand” that guides market transactions and sets prices. If at any point individuals consider the risk far greater than what they had initially estimated then they can raise the price and find out through experimentation at which price they are happy to sell. There are several serious problems with this approach though. For it to work individuals need to respond to fluctuations in the perceived value of their information. According to Tversky and Kahneman (1973) “availability is correlated with ecological frequency”. For valuations this means that when there is a perceived higher risk of sharing information due to an incident that people can recall then their valuation of information will increase, but as the incident becomes more distant in time it will have less of an effect on people’s valuation of their information. This could mean, for example, that the Snowden leaks lead people to become fearful for what is done with their information for a while, but even as the circumstances do not change people will still attach less weight to these leaks as time passes. The question is whether these kinds of events actually affect peoples’ perceptions of risk at all.
Sören Preibusch (2015) found that the public’s interest in mass surveillance and the protection of information was very short-lived, and that the news about the NSA and the PRISM programme struggled
to compete for the readers’ attention with celebrity and sports news. This suggests that privacy concerns do not linger in the individuals’ minds for a long time and they do not take a prominent place. Therefore, instead of having a short effect on privacy-valuations, these kinds of shocks may not have any effect on valuations. This leads to the question whether this matters for a functioning NIM. If changes in the perception of risk that privacy-sensitive information falls into third-party hands does not affect individuals’ valuations, then this might be because the third party is a party that people trust. So for example, people might be fine with the NSA keeping track of their online behaviour because they do not perceive a threat in the NSA, but people would act differently if they felt at risk to criminals who wish to abuse that information. This, however, does not bode well for the individual who is to become a producer of information, because the risk of privacy breaches are an integral part of the cost that is to be calculated. Oftentimes the threats to privacy breaches are explained by referring to dystopian novels, but it seems unlikely that a supermarket collecting information about an individual’s groceries will lead to a dystopian future. However, with health insurance companies the threat might be more quantifiable. Health insurance companies stand to gain from charging people with unhealthy lifestyles more for the insurance. People will want to be compensated for the risk that their health insurance might increase when they release information.
Lastly, the discussion of a market for privacy-sensitive information has until now completely ignored whether individuals are capable or willing to consider privacy an economic good. Attaching a monetary value to privacy may seem as foreign to people as attaching a value to a social phenomenon such as trust. If somebody were to sell trustworthiness, then the fact that it is being sold is enough to obliterate any trust in the seller. The same could be true for privacy. If an organisation wishes to pay you for casually being able to observe your behaviour you may be inclined to question their motives. The fact that it has become a transactable good is a good reason to suspect that it might be sold on to other parties. In total there are three behavioural economics limits to rationality that need to be assessed in order to determine the viability of an information market, namely; - Incomplete information - Calculation of risk - Attitudes to privacy.
4. Survey
The aim of the survey was to determine how well informed people are about their privacy rights, whether they know what happens with their information, and to determine how people feel about sharing and selling their information. The target group of this survey was broad, allowing anyone to participate who is active on the web, older than 10, lives in the Netherlands, and participates in some form of loyalty or reward programme. The National Information Market does not delineate market participants according to a minimum age and in fact children are becoming ever more confronted with invasions of privacy online (Lwin, Stanaland, & Miyazaki 2008, p. 205). The internet’s inherent openness and inability to discriminate obfuscates the boundary between adulthood and adolescence. Therefore in this survey the age group below adults (11-20) is also included. 4.1 Method The survey was conducted using thesistools.com. This platform is appealing due to its simplicity and ease of use. In order to participate in the survey it was only necessary to click on a link, and respondents were able to participate using their computer, smart-phone or tablet. Even though the target demographic was people living in the Netherlands, the survey itself was in English. The reason for this was that I suspected that there would be participants who did not speak Dutch, but I assumed that all participants understand English to a good degree. Adding a Dutch survey may have resulted in extra discrepancy between the results, making it necessary to acquire twice as many results for comparison. The survey was distributed via Facebook, where it was shared several times, via Reddit, and among the more than 80 employees of a Dutch startup company. Of the 115 participants, 16 for some reason did not proceed beyond the basic demographic questions. There are too many possible reasons for this to have occurred to include them in the analysis. It could have been that people were distracted and forgot about the survey, or they abandoned their survey and started a new one at a later time, or maybe they were uncomfortable with sharing the information that was required. Whatever the cause, these answers were not usable and were therefore discarded, leaving a sample size of 99. 4.2 Limitations
As for the limitations of this survey, there are two major problems with the sample. Firstly, these methods of approaching people for the survey is far from a random allocation. While it is appealing to approach people through social media, this means that the answers will be from people who know the
surveyor, and whereby the greatest distance between the surveyor and the participant will be of the second degree. Approaching people through Reddit, an internet-forum, is appealing because it has a special section for surveys, in which a relatively random sample of people are approached. The problem with this is that there are few people who qualify for this survey and are active on that particular forum. Also, these types of fora are generally thought to have somewhat of a hive-mind. While the forum consists of people with differing views on a wide range of subjects, the average user tends to be tech-savvy and knowledgeable about issues relating to their online presence. The employees of the startup that participated in the survey are also all tech-savvy, and only separated from the surveyor by a single degree of separation. Therefore, both problems that occur with the previous groups are also found here. Does this mean that the survey results have no validity? Scientifically this method of sampling cannot be justified. As a result, the conclusions that this research arrives at may not be considered definite and cannot be alluded to in further research. The reason that the survey was not carried out in a manner adherent to the strict scientific method was due to the fact that this method is expensive and requires a team of researchers to gather the data. Neither money nor colleagues were available for this thesis. However, given the results of the survey as they were carried out it is possible to carry out a preliminary analysis and to determine what these results might indicate. Based on the 99 observations I have carried out a data analysis that is not scientifically valid but which can be used to determine possible necessary adjustments to the basic NIM model.
5. Analysis of results
5.1 Characteristics of respondents
Table 1 shows basic information about the respondents. Some data is missing because eight respondents did not fill in their age, while they did answer all other questions. Their results are included nonetheless.
Table 1: Characteristics of participants Age % 11-20 9.9 21-30 69.2 31-40 9.9 41-50 1.1 51-60 4.4 61-70 5.5 71-80 0.0 80+ 0.0 Highest level of education No schooling 0.0 Primary school 0.0 High school 6.1 College/bachelor 43.4 Master 48.5 Post-master 2.0 Number of appliances in possession capable of surfing the web 0 0.0 1 0.0 2 38.4 3 39.4 4 17.2 5 5.1 # of given11 loyalty programmes that you participate in 0 21.21 1 51.52 2 22.22 3 5.05 4 0.00 Do you share promotional content? Yes, if I can win a prize 5.05 Yes, if I think that it is important 10.10 Yes, other 3.03 No, never 81.82 Number of social media accounts 0 2.02 1-2 40.40 3-4 42.42 5-6 12.12 7 3.03 The data shows that the demographic of the sample group was relatively homogenous, consisting mostly of people between 21 and 30 years old who have enjoyed some form of higher education. Every
11 There were four options available to select from. These were AH Bonus Card, AirMiles, Coupons, and Frequent Flyer Points.
respondent has at least two appliances with access to the internet. Of all of the respondents that participate in some form of a loyalty programme, only 3 collect coupons, which are generally a non-privacy invasive type of loyalty programme. A large majority never shares promotional content on social media, while only one respondent did not have any kind of social media account.
5.2 Variables
Using the answers represented in tables 2 and 3 (appendix) I made four index variables to study people’s attitudes toward privacy issues and their stance on selling information. The four variables are pseindex, wtsindex, lvlindex, and cnfindex. pseindex is a measure of privacy sensitivity, wtsindex measures the willingness to sell information, lvlindex measures the level of trust that individuals have in organizations that collect information, and cnfindex measures the confidence that people have in their own knowledge of privacy issues.
The index variables are created by assigning values to respondents’ answers. For example, if somebody answers ‘Agree’ to the statement “I trust apps with my information” then they score one point and that point is added to the total of answers that makes up the index lvlindex. Any other answer to this statement does not score any points. The total number of points scored per individual is divided by the total number of points attainable to get a score out of 1. If somebody scores “1” on lvlindex then they are very trusting of organizations that use their information. If somebody scores “0” then they do not trust those organizations. Any score in between measures the degree of trust of that individual. Instead of directly measuring the knowledge that people have about about privacy issues I have chosen to measure the confidence that people place in their own knowledge. There are two reasons for this. Firstly, measuring the real knowledge that people have about privacy issues is difficult because many privacy policies vary and it is not possible to find out exactly what happens with information, even when privacy policies are given. For example, while apps are allowed to share information with third parties as long as they have the user’s consent, legislation is not clear on when an individual consents, and research has shown that many apps send information to third parties regardless of the stated policy. Therefore, people’s answers may often show up as ‘incorrect’ even though they are correct and vice versa. The confidence that people exhibit in their knowledge can instead be used to measure the extent to which people feel like they can make choices based on the available information in an information market.
5.3 Comparison to survey carried out by Turow et al
Researchers at the University of Pennsylvania (Turow et al. 2015) carried out a survey amongst 1,506 US adults, in which respondents were asked to state whether they agree or disagree with several statements. Theses statements can be used to determine whether participants of this survey have the same attitude as people in the US. I will compare the results of the four most important of their statements to the results in this survey.
In the US study, 91% of respondents disagreed with the statement “If companies give me a discount, it is a fair exchange for them to collect information about me without my knowing.” In order to test respondents’ stance on this issue I asked them to assess two similar statements, namely “I am fine with supermarkets tracking what I buy in return for a discount” and “Receiving a discount in return for information about me is a good deal”. The second statement tests whether people think that they are getting a good deal when trading information for a discount. Only 36% of respondents think that this is the case. The first statement is less stringent, in that it does not assess whether people think that they are getting a good deal, but instead whether they can live with this situation in which information is traded for a discount. In this case 63% agree that they will accept the situation as it is. This result suggests that a majority of respondents may have gotten used to exchanging their information in return for a discount, while most of them actually do not consider it to a be a good deal. Of the people who were fine with supermarkets tracking their groceries, just under half consider this a good deal. From this data one can conclude that respondents of this survey are not fiercely against the idea of exchanging information for discount, but a majority does not consider it a fair deal. In Turow et al’s survey 71% disagree with the statement that “It’s fair for an online or physical store to monitor what I’m doing online when I’m there, in exchange for letting me use the store’s wireless internet, or Wi-Fi, without charge.” This is in essence a question about whether people consider it a fair deal to trade information for a service. In my survey I tried to get respondents to think of another similar situation, which may be a bit more specific, namely the statement “I pay for Facebook and Google with my information.” The purpose of this question was to determine whether respondents feel that they are involved in an exchange of information for services. 62% agree with this statement, 23% disagree, and 15% are not sure. This shows that a majority of people feel that they are involved in an exchange of information for services. In order to determine whether people thought this was a fair deal they were asked to rate the statement “Any amount of information is a fair deal for their free services.” Only 19% agree, while 62% disagree. Of the people that considered their use of Facebook and Google an exchange of information in return for services only 18% considered it fair that the services were able to collect any and all information. These results suggest that the respondents of my survey are more opposed to
exchanging information for services than the US respondents. It also supports the idea that people are resigned to giving up their data.
Of Turow et al’s survey 84% agree that “I want to have control over what marketers can learn about me online.” I tried to assess this by asking which steps people undertake to have control of their information. 75% state that they “try to protect privacy online”, where 11% are not sure about whether they try to. 56% regularly delete cookies on their computer, and 12% have ever asked Facebook or Google to send them a list of all the information that has been collected on them. The last statement does not ask about steps undertaken to protect privacy, but is a test of whether people are curious about the information that is collected on them. These results suggest that people do want to have control over their information, but are not eager to find out what has already been collected. Also, 60% of respondents believe that “Companies are collecting too much personal information about me”, which suggests that a majority of respondents would like more control over their personal information. The last statement from the US survey that I wish to examine is the statement that supports the authors’ conclusion of their research, namely that the reason for the privacy paradox is that people are resigned to giving away information. As soon as people become active online or in supermarkets they may feel that they have no control over what happens with their information. The statement is “I’ve come to accept that I have little control over what marketers can learn about me online”, to which 65% of respondents agreed. In my survey 60% agree with the statement “Companies are collecting too much information about me.” Of the respondents who say that they try to protect their online privacy, 66% nevertheless feel that companies are collecting too much information about them. Of the respondents who say that they do not try to protect their privacy online or are not sure 56% feel that “It is not possible to stop companies from collecting personal information about me.”
The data suggests that a majority of people do not consider information for discounts or services a fair deal, and most people would like to have more control over the dispersion of their information. However, the survey results also show that people are willing to sell information. In order to determine what affects a person’s willingness to sell I carried regressions containing the following variables:
The variables represent respectively; age, level of education, number of online appliances in possession, number of loyalty programmes, number of social media accounts, privacy sensitivity, willingness to sell, level of trust in organizations that use personal information, and confidence in own knowledge of information dispersion. All variables were indexed so that the maximum attainable value for each is 1 and the minimum is 0. This does not affect the outcome of the regression output but it does make analysis easier. For example, it is clear from the summary above that the mean age of respondents is in the lower third of all available age categories from 11 to 80+, and the level of education of respondents on average is high.
The variables that have a significant influence on willingness to sell12 are the number of loyalty programmes that a person participates in, the privacy sensitivity of the individual, and the trust that a person places in organization that use personal information. Privacy sensitivity and willingness to sell are negatively related, which makes sense because people who are more concerned with what happens with their information should be less likely to want to sell that information. The number of loyalty programmes that a person participates in has a positive effect on a person’s willingness to sell according to the regression. However, one would think that the effect is the other way around, namely that people who are more willing to sell information are more likely to participate in loyalty programmes. Therefore, that variable cannot be considered an explanatory variable. Respondents who are ranked higher on the level of trust that they place in organizations that use information are more willing to sell information. This positive relationship makes sense, because when people believe that their information is safe with these organizations they consider the transaction to be less risky. However, it is pertinent to realize that 12 See appendix 3 for regression outputs
the overall trust in organizations is low, which indicates that most people do not place a lot of faith in those organizations’ security or integrity.
For the sake of further understanding why some people are more willing to sell their information than others it is necessary to understand why people vary in their degree of privacy sensitivity. The only variable that has a statistically significant effect on privacy sensitivity is the level of education. The higher the level of education, the greater that person’s privacy sensitivity. The variable pseindex measures the extent to which people wish to prevent personal information from spreading. One of the reasons for this relationship could be that people with a higher level of education are more concerned with (future) employers finding certain information about them. Many of the respondents selected ‘other’ and filled in ‘employers’ under the question “Which of the following groups will make you reconsider putting something online”. The reason that people with a higher education are likely to be more concerned about this is because applicants for jobs that require a high education tend to be screened more vigorously than applicants for jobs that require a low education. Also, it might be the case that people with a higher level of education are more aware of the risks that surround privacy issues.
Strangely enough there is not a statistically significant relationship between privacy sensitivity and the confidence that people have in their knowledge of privacy issues. It would have made sense to see at least some relationship between the two, for example because people who are highly sensitive to privacy issues spend more time delving into the matter. One obfuscating factor could be that people who spend more time researching privacy issues are more aware of the things that they do not know, while people who spend little time informing themselves are blissfully unaware of the issues that surround the matter13.
The absence of a relationship between the confidence in knowledge and willingness to sell is also peculiar. Turow et al (2015) found that “people who know more about ways marketers can use their personal information are more likely rather than less likely to accept discounts in exchange for data when presented with a real-life scenario.”. If this finding also holds for the Dutch population then that would mean that there is no relationship between the a person’s knowledge on privacy issues and the confidence in their knowledge. McDonald & Cranor (2008) found that it would take more than 25 8-hour workdays per year for the average internet-user to read the privacy policies of the sites that they visit. Perhaps the required time to fully delve into the subject matter causes individuals to resign. As a result of the arduous nature of privacy policies people have created a heuristic in order to avoid having to read all of the lengthy texts. The mere presence of a privacy policy or indication of regulation is 13 Situations in which ignorance lead to greater confidence have been observed previously and this is known as the Dunning-Kruger effect (Dunning & Kruger 1999).
often enough for individuals to place more trust in an online environment (Berendt et al, 2005), without any knowledge of what is done with their information.
6. Application of results to the model
6.1 The Bounded Rationality Model
The combination of the National Information Market model with the Kreps-Scheinkman model requires a few adaptations in order to properly represent the individual. Firstly, the individual’s privacy sensitivity is measured by the cost incurred. To recap, the individual’s payoff function is πi = (1 - h)P(qi + qj)qi - xiqi. In
the Model Formulation section no explanation was given as to what the cost function is made up of. Now, with a variable for privacy sensitivity, it is possible to quantify the costs of selling information. In order to determine the cost of selling information the individual needs to make an assessment of the risk that her information will be revealed to other parties, especially those parties that the individual explicitly does not want to share the information with.
In the survey privacy sensitivity is measured by an index variable. This means that the index variable needs to be represented by the cost in the payoff function for the individual. In the Limits to rationality section I explained how risk perceptions can be affected by global, political events and also how this effect wears off over time. Regardless of the nature, individuals have some perception of risk, in this case represented by r14. r is the monetary value of the pain of having privacy breached. The cost function is x
i
= rχi, where χi is individual i’s privacy sensitivity. To interpret this, an individual with low (high) sensitivity
is affected less (more) by a higher perception of risk.
Secondly, the individual in the NIM is not sure of the payoff, because prices are determined after quantities have been set. Prices are affected by the amount of information that is supplied. It is assumed that the demand for information is stable. In the basic model of this thesis three different price situations are possible, namely a low price when both types of individuals have a low privacy sensitivity, a medium price when the individuals differ in their price sensitivity, and a high price when both individuals have a low sensitivity. If the types of individual have the same levels of privacy sensitivity then the model is symmetric and the individuals’ best responses are qi* = [a - rχi/(1 - h)]/3b, with a resulting total quantity and price of 14 There is a serious problem with the idea that the risk associated with privacy breaches can be quantified in monetary terms. In the extreme it means that individuals need to be able to express the amount of money for which they would share their deepest secrets with people that they want to keep this information from. Perhaps the problem is less severe with basic information such as name, date of birth, etc.
Q* = 2/3b[a - rχi/(1 - h)] and
p* = [a + 2rχi/(1 - h)]/3.
The individual’s best response function shows that the individual will sell less information if their privacy sensitivity, their risk perception, or the handling costs increase. The same effect can be found in the total quantity, since that is simply the sum of the individual quantities. As economic theory dictates, the price should increase as the quantity decreases, therefore the price should be positively related to the individuals’ privacy sensitivity, risk perception, and the handling costs. This is the case in this model. The mean of the privacy sensitivity index from the survey is 0.6. Taking this as the value for a symmetric situation, the equilibrium price is
p* = [a + 1.2r/(1 - h)]/3.
What if the game is not symmetric? In that case the types of players can vary in their level of privacy sensitivity, and one type of individual may not know the level of privacy sensitivity of the other type. Individuals will therefore have to make an estimate of the privacy sensitivity of others. Given that the best response function in the non-symmetric case is
q* = [a - 2rχi/(1 - h) + rχj/(1 - h)]/3b,
individuals underproduce if others are more privacy sensitive than they assumed, and overproduce if others are less privacy sensitive than assumed. In the former situation the individual will earn less than the optimal amount. In the latter case the individual earns more, but also divulges more information than she would have liked.
From the survey data, let’s call the lower half of pseindex individuals with low privacy sensitivity (χi), and
the rest individuals with high privacy sensititvity (χj), and take their respective mean values as their index
representation, so χi = 0.46 and χj = 0.76. The equilibrium price and the total quantity sold remain the
same, but the best response functions are now qi* = [a - 0.92r/(1 - h) + 0.76r/(1 - h)]/3b = [a - 0.16r/(1 - h)]/3b qj* = [a - 1.52r/(1 - h) + 0.46r/(1 - h)]/3b = [a - 1.06r/(1 - h)]/3b. The individuals’ payoffs are
πi = qi*[(a(1 - h) + 1.2r)/3 - 0.46r] = qi*[a(1 - h)/3 - 0.06r] πj = qj*[(a(1 - h) + 1.2r)/3 - 0.76r] = qj*[a(1 - h)/3 - 0.36r] The low sensitivity individual sells more (qi > qj), which results in higher revenue, and that individual also has lower marginal costs. Therefore, the low-sensitivity individual is better off in a NIM than the high-sensitivity individual. 6.2 The Bounded Rationality Model with many individuals In a real-world market there could potentially be millions of individuals who sell their information. In that case a single individual will not affect prices. Therefore this section introduces a market model wherein prices do not vary with quantities, that is, prices are fixed. In this model the individual maximizes the following payoff function max(qi) πi = qi[(1 - h)p - rχi]. In this situation the individual either sells all of her information or none at all. Her best response function is max. if (1 - h)p > rχi qi = 0, max. if (1 - h)p = rχi 0 if (1 - h)p < rχi. An average individual would therefore sell her information if (1 - h)p >= 0.6r, and not sell otherwise. For the average low-sensitivity individual this is (1 - h)p >= 0.46r, and for the average high-sensitivity individual it is (1 - h)p >= 0.76r. What this means is that the choice of whether to sell is based on the ratio of marginal revenue and marginal cost, just as a firm, but that the marginal cost is affected by some subjective valuation of privacy.
7. Conclusion This thesis was a first attempt at getting closer to understanding the factors that underlie decision-making in an information market. The baseline model and the bounded rationality model are not that different. In fact, the only thing change in the bounded rationality model is the way in which costs are calculated. The baseline model assumes that individuals have knowledge of the costs associated with selling information, while in the bounded rationality model the individual requires two steps to calculate the costs; first she need to know how much she values her privacy and secondly, she needs to know the cost of the risk that they are taking. The first step, identifying the level of privacy sensitivity, was done by asking questions in a survey. The reason that I chose for this approach is because privacy sensitivity is a non-quantifiable feeling. Therefore, the best way to determine people’s privacy sensitivity is to ask how they feel about certain topics surrounding privacy issues. The questions were not comprehensive but the questions asked were encompassing enough to allow analysis of individuals’ privacy sensitivity. Another reason that the survey method was opted for is because they contained questions that individuals can ask themselves. The measure of privacy sensitivity needs to be a measure that individuals can measure themselves. The only variable that has an effect on the individual’s privacy sensitivity is the level of education. This suggests that privacy sensitivity does not fluctuate much as the level of education takes a while to change. Therefore, once an individual has determined her own privacy preferences she does not need to adjust this constantly when calculating the benefits of selling information. The second step, identifying the cost of the risk of selling information, cannot be identified through a survey such as the one used in this thesis. In order to calculate this it would be necessary to set up an experimental procedure, but even then there remains one important question; is a loss of privacy quantifiable? This question is far more philosophical than of an economic nature. However, it might be possible to set an experiment in a real-life situation to determine whether people are capable of making these kinds of calculations. The last addition to the model, a large number of individuals that causes the price to be unaffected by the single individual, results in a completely different outcome for the individual. Since prices are unaffected by the quantity set by the individual, the choice is between all or nothing. If the price of information is higher than the associated cost then the individual should sell all of her information, and she should sell none if this is not the case. If the individual knows what the costs are of selling information then she can simply compare this to the price and determine whether she wants to sell or
not. In this case, privacy sensitivity does not result in varying quantities of information sold, it only determines whether individuals want to sell all of their information or none of it. In the market model with two different individuals, one with a low sensitivity to privacy breaches and the other with a high sensitivity, the individual with a low sensitivity stands to gain more from participation in the NIM. This does not necessarily mean that that individual is also better off. In fact, an individual with a high privacy sensitivity might gain more utility from not sharing information and may therefore be better off or the same. This supports the idea that individuals are empowered in a NIM. Equipped with these conclusions it is possible to answer part of the questions that were asked at the beginning of this thesis, namely: - Is the transition from a universal basic human right to a market commodity a feasible option? - Are individuals capable of acting rationally enough to benefit from taking on the role of suppliers of their own information? The survey results show that there is a willingness to sell information. In fact, some people even claim to view the exchange of goods and services for information as a transaction that they are comfortable with, despite the very low levels of trust toward organizations using personal information. What this might suggest though is that people do consider risk at all when dealing in transactions of this nature. If this is the case then that might be the answer to the second question, namely that people cannot interpret the risk associated with transacting in information in any meaningful way for a calculation of payoffs. In other words, the decision whether to transact in private information is completely unaffected by the risks associated with it. Unfortunately I cannot say whether this is the case, but this is an interesting topic for further research. An answer to this question could determine whether the National Information Market is a feasible solution to the privacy problem.
8. Appendix
Appendix 1: Survey questions
Table 2: Attitudes to privacy I try to protect my privacy online % Agree 76.8 Disagree 13.1 Not sure 10.1 I trust apps with my information Agree 38.4 Disagree 26.3 Neither 35.4 I trust social media sites with my information Agree 29.3 Disagree 42.4 Not sure 28.3 I am fine with supermarkets tracking what I buy in return for a discount Agree 63.6 Disagree 24.2 Not sure 12.1 I expect a greater discount if information about my shopping behaviour is not anonymous Agree 57.6 Disagree 25.3 Not sure 17.2 Companies are collecting too much personal information about me Agree 59.6 Disagree 15.2 Not sure 25.3 I trust that shops will not sell information about my shopping behaviour to third parties Agree 20.2 Disagree 61.6 Not sure 18.2 Receiving a discount in return for information about me is a good deal Agree 35.4 Disagree 35.4 Not sure 29.3 Before I upload a photo or a video I first think about who will see this Agree 89.9 Disagree 7.1 Not sure 3.0 How many groups will make you reconsider putting something online (family, friends, acquaintances, advertisers, government, employers,...) 0 5.32 1-2 45.74
3-4 37.23 5-6 10.64 7+ 1.06 In the supermarket I sometimes do not use my loyalty card so that my groceries will not be tracked Agree 16.2 Disagree 77.8 Not sure 6.1 There are some products for which I will not use my loyalty card even if there is a large discount (Food, snacks and sweets, alcohol, cigarettes, hygienic products, magazines, birth control products…) Agree 9.1 Disagree 80.8 Not sure 10.1 If a company’s database has been hacked before I will not share my information with that company Agree 33.3 Disagree 39.4 Not sure 27.3 I pay for Facebook and Google with my information Agree 62.6 Disagree 23.2 Not sure 14.1 Any amount of information is a fair deal for their free services Agree 18.2 Disagree 62.6 Not sure 19.2 I would prefer to share less information than I am now sharing Agree 74.7 Disagree 14.1 Not sure 11.1 Companies that use or sell my information should pay me for it with money Agree 40.4 Disagree 41.4 Not sure 18.2 I want to have more time to think about which information I share online Agree 52.5 Disagree 35.4 Not sure 12.1 Instead of discounts that I cannot influence, supermarkets should give me money every month in return for the information that they collect on me Agree 27.3 Disagree 54.5 Not sure 18.2 Getting a discount in return for my information is a good deal Agree 43.2 Disagree 31.6
Not sure 25.3 Table 3: Relationship to organizations Supermarkets are allowed to share my information with third parties % Agree 15.2 Disagree 70.7 Not sure 14.1 It is possible that I will be charged a higher price as a result of information collection Agree 34.3 Disagree 32.3 Not sure 33.3 I use an Ad Blocker on my browser Agree 70.7 Disagree 26.3 Not sure 3.0 I regularly delete the cookies on my computer Agree 56.6 Disagree 41.4 Not sure 2.0 When I use my credit card online only the company I am buying something from knows about this transaction Agree 28.3 Disagree 30.3 Not sure 41.4 Social media sites are allowed to use my photos and videos for their own purposes Agree 20.2 Disagree 71.7 Not sure 8.1 Facebook, Google, and Twitter can follow what I am doing online even when I am not on their sites Agree 58.6 Disagree 16.2 Not sure 25.3 I know what happens to the information that supermarkets collect about me Agree 24.2 Disagree 39.4 Not sure 36.4 It is not possible to stop companies from collecting personal information about me Agree 50.5 Disagree 38.4 Not sure 11.1 I have asked Facebook and/or Google to send me a list of all of the information that they have collected on me Agree 12.1 Disagree 85.9 Not sure 2.0
Appendix 2: Construction of character attribute variables
Variable name Variable description Questions included to make index Answer score
pseindex Privacy sensitivity: Index rating of how dedicated a person is to protecting privacy and how reluctant that person is to give
away information I try to protect my privacy online Agree: 1 0: not at all sensitive Companies are collecting too much information about me Agree: 1 1: very sensitive Before I upload a photo or a video I first think about who will see this Agree: 1
How many groups will make you reconsider putting something online? 0-1: 0, 2-3: 1, 4+: 2 In the supermarket I sometimes do not use my loyalty card so that my groceries will not be tracked Agree: 1 I would prefer to share less information than I am now sharing Agree: 1 I want to have more time to think about which information I share online Agree: 1 I regularly delete the cookies on my computer Agree: 1 wtsindex Willingness to sell information: Index rating of likelihood that this person will accept money in return for information I am fine with supermarkets tracking what I buy in return for a discount Agree: 1 0: not willing to sell I expect a greater discount if information about my shopping behaviour is not anonymous Disagree: 1 1: very willing to sell Receiving a discount in return for information about me is a good deal Agree: 1 There are some products for which I will not use my loyalty card even if there is a large discount Disagree: 1 I pay for Facebook and Google with my information Agree: 1 Any amount of information is a fair deal for their free services Agree: 1 Companies that use or sell my information should pay me for it with money Agree: 1 Instead of discounts that I cannot Agree: 1
influence, supermarkets should give me money every month in return for the information that they collect about me lvlindex Trust in organizations: Index rating of how trusting a person is of the organizations that collect information and the belief that the information be
safely stored I trust apps with my information Agree: 1 0: not trusting I trust social media sites with my information Agree: 1
1: very trusting I trust that shops will not sell information about my shopping behaviour to third parties Agree: 1 If a company's database has been hacked before I will not share my information with that company Disagree: 1 Social media sites are allowed to use my photos and videos for their own purposes Agree: 1 cnfindex Confidence in knowledge: How certain is a person of their knowledge about what happens with their information I know what happens to the information that supermarkets
collect about me Agree, Disagree: 1 Not Sure: 0 It is not possible to stop companies
from collecting personal
information a... Agree, Disagree: 1 Not Sure: 0 Supermarkets are allowed to share
my information with third parties Agree, Disagree: 1 Not Sure: 0 It is possible that I will be charged a
higher price as a result of
informa... Agree, Disagree: 1 Not Sure: 0 I use an Ad Blocker on my browser Agree, Disagree: 1 Not Sure: 0 I regularly delete the cookies on
my computer Agree, Disagree: 1 Not Sure: 0 When I use my credit card online
only the company I am buying
something fro... Agree, Disagree: 1 Not Sure: 0 Social media sites are allowed to
use my photos and videos for their
own pu... Agree, Disagree: 1 Not Sure: 0 I have asked Facebook and/or
Google to send me a list of all of
Facebook, Google, and Twitter can follow what I am doing online even
when I... Agree, Disagree: 1 Not Sure: 0 I would prefer to share less
information than I am now sharing Agree, Disagree: 1 Not Sure: 0
9. Bibliography
Acquisti, A. (2004). Privacy in electronic commerce and the economics of immediate gratification. Proceedings of the 5th ACM Conference on Electronic Commerce - EC '04, 21-29. Acquisti, A. & Grossklags, J. (2005). Uncertainty, ambiguity and privacy. Workshop on Economics and Information Security (WEIS 2005).
Grossklags, J., & Acquisti, A. (2007). What Can Behavioral Economics Teach Us about Privacy? Digital Privacy Theory, Technologies, and Practices, 363-377. doi:10.1201/9781420052183.ch18 Almuhimedi, H., Schaub, F., Sadeh, N., Adjerid, I., Acquisti, A., Gluck, J., . . . Agarwal, Y. (2015). Your Location has been Shared 5,398 Times! Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems - CHI '15. doi:10.1145/2702123.2702210 Ayalew, R. (2011). Consumer behaviour in Apple's App Store. Master scriptie: Human Computer Interaction Programme, 1-103. Berendt, B., Günther, O. & Spiekermann, S. (2005). Privacy in e-commerce: stated preferences vs. actual behavior. Communications of the ACM, 48(4), 101-106.
Beresford, A. R., Kübler, D., & Preibusch, S. (2012). Unwillingness to pay for privacy: A field experiment. Economics Letters, 117(1), 25-27. doi:10.1016/j.econlet.2012.04.077
Borgesius, F. Z. (2015). Privacybescherming online kan beter. Nederlands Juristenblad, 14, 878-883.
Camerer, C., & Weber, M. (1992). Recent developments in modeling preferences: Uncertainty and ambiguity. Journal of Risk and Uncertainty, 5(4), 325-370. doi:10.1007/bf00122575
Kruger, J., & Dunning, D. (1999). Unskilled and unaware of it: How difficulties in recognizing one's own incompetence lead to inflated self-assessments. Journal of Personality and Social
Psychology, 77(6), 1121-1134. doi:10.1037/0022-3514.77.6.1121
Earp, J., Anton, A., Aiman-Smith, L., & Stufflebeam, W. (2005). Examining Internet Privacy Policies Within the Context of User Privacy Values. IEEE Transactions on Engineering Management IEEE Trans. Eng. Manage., 52(2), 227-237. doi:10.1109/tem.2005.844927
Elsen, M., Elshout, S., Kieruj, N., & Benning, T. (2014). Onderzoek naar privacyafwegingen van internetgebruikers. Tilburg: CentERdata.
Fisher, R. J. (1993). Social Desirability Bias and the Validity of Indirect Questioning. Journal of Consumer Research, 20(2), 303. doi:10.1086/209351
