Examining influential aspects of piracy-related behaviour concerning (e-)books : a vignette experiment

(1)

Examining influential aspects of piracy-related

behaviour concerning (e-)books: a vignette experiment

Luuc Overvoorde

Faculty of Economics and Business

University of Amsterdam 10197508

Supervisor: dr. Joep Sonnemans 10 november 2015

This research aims to assess the effect of different aspects of norms regarding the appropriation of copyright goods, in this case (e-)books. These norms precede the act of appropriation, e.g., downloading e-books without payment. Students participated in a vi-gnette experiment, judging the acceptability of twenty appropriation scenarios. Whereas most aspects show an expected effect on the acceptability of the scenarios, such as that fur-ther sharing of the good renders the behaviour less acceptable, two aspects show unexpected results. Firstly, the physicality of the book (being either an e-book or a physical book) shows a reverse effect, suggesting stealing a real book is morally more acceptable than stealing an e-book, all else equal. However, this might be due to the interactions with other aspects. Secondly, the acceptability of the act does not change when the producer incurs a loss. A

number of explanations are discussed, in addition to the implications for further research. 1

1_{I would like to thank Joep Sonnemans for giving invaluable advice on how to design the experiment,}

and in general for supervising this thesis. In addition, I would like to thank Anouar El Haji for sparking my interest in this topic. I would like to thank my friends and family for their support. Lastly, I am grateful for all participants in the experiment.

(2)

Contents

1 Introduction 3

2 Literature review 5

2.1 Vignette experiment and students . . . 5

2.2 Individual and social norms . . . 6

2.3 Loss producer . . . 6

2.4 Moral cognition . . . 7

3 Hypotheses 8 4 Methodology 10 4.1 Respondents . . . 10

4.2 Scenarios and variables . . . 11

4.3 Treatments and incentives . . . 15

5 Results 16 5.1 Assumptions . . . 16

5.2 General regression . . . 16

5.3 Heavy users . . . 18

5.4 Treatment effects . . . 18

5.5 Interaction terms and complete model . . . 18

6 Discussion and conclusion 22 7 References 25 8 Appendix 27 8.1 List of scenarios . . . 27

8.2 Density graph of respondents’ age . . . 29

(3)

ASPECTS OF PIRACY-RELATED BEHAVIOUR 3

1 Introduction

Digital copyright infringement, also known as piracy (Picard, 2004), is regarded by many as a significant problem for producers and retailers of copyright goods. Although pre-cise estimations are difficult to make, which does not prevent a vast body of estimations to be published, the foregone gains from digital sales by the illegal appropriation of digital in-tellectual property are considerable (Siwek, 2007). In the past years, many on-line platforms arose through which copyright goods, such as music, films, tv series, and books, are shared with anyone, which essentially allows free-riding on those who purchased the good initially (Cubitt, Drouvelis, Gächter, & Kabalin, 2011). Compared to this, the occurrence of illegal appropriation of physical goods containing intellectual property is negligible. In parallel, the norms regarding piracy are considerably more loose than theft. Although scarcely re-searched, this difference in moral perception of two very similar acts of illegal appropriation of copyrighted material is peculiar. Indeed, the norms one has regarding certain behaviour are taken into account when considering behaviour (Krupka & Weber, 2013). On the other hand, as copyright laws differ across countries, norms regarding the behaviour do not neces-sarily follow (Svensson & Larsson, 2012). What aspects regarding piracy and piracy-related behaviours activate which norms?

The drivers of this difference may lie in the aspects involved in the individual and social norms. For example, copying a digital good seems to be socially appropriate, whereas stealing a physical good is not. Some different aspects may change which social norm is activated, such as rivalry of the good: stealing a book is different from copying one. A number of models to describe this behaviour have been proposed (Peitz & Waelbroeck, 2006). However, these do not adequately investigate the norms at the basis of such behaviour. Why is digital theft not considered immoral, while physical theft is? And do individual norms differ from social norms in this context? In order to answer this question, books and e-books are used as copyright goods in the following experiment.

Respondents participated in a vignette experiment based on Hardy and colleagues’ study (2013), consisting of 20 scenarios regarding the appropriation of copyright goods, in this case (e-)books. Respondents rated the scenarios on their moral acceptability. The scenarios closely resemble one another, except for five key aspects. These varying aspects, such as the physicality of the good (either a book or an e-book), result in the 20 different scenarios. In addition, two treatments determined whether the moral acceptability rating was the individual’s own norm, or what the individual regarded as the social norm. Using these treatments, the respondents were incentivised in the second, social norm treatment to predict the ’right’ social norm answers. This was done by averaging the responses in the first, individual norm treatment. Comparing these averages to the responses in the social norm,

(4)

respondents were ranked on performance, and received the reward accordingly. Afterwards, the influence of the varying aspects is analysed on the acceptability of the behaviour is analysed, in order to get a better view on what determines the individual and social norms on this type of behaviour.

(5)

2 Literature review

In this section, the literature regarding piracy and its normative aspects will be dis-cussed, in addition to the literature on the method used for the experiment. An important distinction made throughout this research is that differences in downloading behaviour and the respective norms stem from different circumstances and beliefs, instead of inherently different individuals with certain predetermined tendencies.

First, an overview of the literature on vignette experiments and the implications of using a student sample will be discussed. Second, the norms regarding piracy will be dis-cussed. Third, the research on the loss incurred by the producer is explained, in addition to a cognitive explanation of the difference in norms regarding downloading behaviour.

2.1 Vignette experiment and students

A common and effective way to study moral judgement is through a vignette experiment (Presser et al., 2004; Martin, 2004; Alexander & Becker, 1978). Vignette experiments offer a set of scenarios, each scenario different in certain variables. A scenario can describe any hypothetical event, hence the variables can determine any aspect of the scenario. The dependent variable(s) should be the same for every scenario. This way, the direction and size of each variable’s influence on the dependent variable can be determined through statistical analysis. A large advantage of vignette experiments is the possibility to research more sensitive topics, including student cheating (Rettinger, Jordan, & Peschiera, 2004), abortion refusal (Rasinski, Yoon, Kalad, & Curlin, 2011), and mental illness stigmatisation (Link, Yang, Phelan, & Collins, 2004). Despite this, few vignette experiments consider the moral discrepancies regarding copyright goods.

One study, conducted by Hardy, Krawczyk and Tyrowicz, upon which this research is based, does take moral considerations into account (Hardy et al., 2013). They use a triple treatment vignette experiment to elicit norms regarding piracy-related behaviour concerning films. Asking subjects to rate a number of scenarios in which the piracy-related behaviour is described on the moral acceptability, they attempt to distinguish the impact of the dif-ferent aspects in a larger model. In addition, the treatments (individual judgement, social judgement, and incentivised social judgement) enabled an in-experiment evaluation of the validity of the responses.

Aside from this, results obtained using students as participants paint a realistic and generalisable picture regarding piracy. In contrast to the view that students are different from the general population for their more lax norms, financially constraints, and lower ethical standards (Al-Rafee & Cronan, 2006; Gunter, 2009), Krawczyk and colleagues (2015) show there is very little difference to be found between students and members of a digital rights

(6)

defenders group in a resembling vignette experiment regarding piracy norms. Similarly, pro-fessionals do not behave more closely to theoretical predictions than students in economic experiments (Fréchette, 2011). While the difference of students’ perception of piracy com-pared to the general population is to be agreed upon, the issue of how these perceptions or norms are constructed remains unanswered.

2.2 Individual and social norms

Distinguishing individual norms from social norms likely improves the validity of a study concerning a morally sensitive topic, such as piracy. Individual norms are the norms one has itself regarding a certain behaviour, whereas social norms are the norms shared by the direct environment. Aiming to explain the normative difference between physical and digital theft of copyright goods, the literature proposes several factors influencing individual and social norms regarding theft. These factors include availability of alternatives, physicality of the good, distributing the good after obtaining it, the loss to the direct source (e.g., a store or a friend), and the loss to the producer. Considering these factors, amongst others, but with the exception of the loss to the producer, Hardy and colleagues find using a similar vignette experiment most show expected, but varying strengths of the isolated factors (Hardy et al., 2013). Of these factors, physicality of the good showed the strongest negative influence on the norm. As the physicality variable entails either a physical or a digital good, this can be explained by the difference in rivalry: stealing a physical good from a store implies a decreased inventory for the store, whereas stealing a digital good merely implies copying the good. However, the physicality of the good alone might not change the norm, but another variable correlating well with it: the loss of the producer.

2.3 Loss producer

The costs incurred by the producer has been scarcely researched as a determinant in the decision process of theft, even though this is suggested to be equally detrimental to producers’ incomes as to the re-sellers (Poort & Leenheer, 2012). An important distinction herein is that loss is only incurred by the producer when piracy leads to foregone gains. In other words, piracy affects to producers’ earnings only if it prevents otherwise consuming persons to buy the good (Picard, 2004). Building on this distinction, a study proposes the fixed and variable costs incurred by the producer are regarded ethically different by consumers, leading to a more theft when variable costs are relatively low (Nunes, Hsee, & Weber, 2004). However, this difference between fixed costs and variable costs often coincides with the physicality of the good: an e-book has significantly lower variable costs than a

(7)

physical book. What remains interesting is the way consumers incorporate cost structures in their norms and decisions regarding information goods.

2.4 Moral cognition

Aside from this, a more fundamental hypothesis might cause the difficulty in perceiving the morality of digital behaviour. A still unpublished study tests the hypothesis that moral cognition cannot fully grasp digital theft, due to the absent rivalry of the good (i.e., when stealing a digital good, one often merely copies it) (El Haji & Leenders, n.d.). This could extend to all digital behaviour, as all content is virtual, requiring a medium of display. Even though this is a compelling argument, I do not intend to assess the influence of these proposed factors at this point.

(8)

3 Hypotheses

The following hypotheses are tested in the subsequent analysis. The first hypotheses are based on the single aspects used in the scenarios, after which the other hypotheses are dis-cussed. In the hypotheses, the source entails the last owner of the good, from which the good is taken, and the producer entails the individual or group that produced the (intellectual) good.

The most visible difference in piracy versus physical theft is the difference in the physi-cality of the good. As in a similar research the physiphysi-cality strongly decreased the acceptability of the behaviour (i.e., if the only difference between two scenarios is the good being physical, instead of digital) concerning films (Hardy et al., 2013; Krawczyk et al., 2015), the same influence can be assumed in this experiment, as stated in the first hypothesis:

1. Piracy-related behaviour is found less acceptable if physical books are used, as com-pared to e-books.

When a copyright good is obtained without payment, the producer of the good faces a foregone gain, which can be considered a loss. This loss is used in the scenario by stating the individual "would [not] have bought the good otherwise", hence, the willingness to pay (WTP) is greater [smaller] than the purchasing price (P). If the individual would have bought the good, W T P > P , the behaviour is expected to be found less acceptable, as the producer is not reimbursed for producing the good, as stated in the second hypothesis:

2. Piracy-related behaviour is found less acceptable when W T P < P than W T P > P The following hypothesis concerns the direct source from which the good is taken or copied. In accordance with previously found results (Hardy et al., 2013), if the source is a friend, respondents are expected to find the behaviour less acceptable, compared to the source being an (on-line or physical) store:

3. Piracy-related behaviour is found less acceptable when the direct source is a friend The next hypothesis concerns an additional act. When the scenario continues after the act of appropriation, and describes the further sharing of the good on-line, this should decrease the acceptability of the scenario (Hardy et al., 2013):

4. Piracy-related behaviour is found less acceptable when it entails further sharing the good

Similarly, if the behaviour entails a source incurring a loss, it is found less acceptable. As stealing is typically perceived as immoral, this addition should lower the moral acceptability, e.g., stealing an e-book from a source that owns the e-book versus copying an e-book.

(9)

ASPECTS OF PIRACY-RELATED BEHAVIOUR 9 5. Piracy-related behaviour is found less acceptable when the source incurs a loss

Downloading behaviour being different per person, more experienced users should find piracy-related behaviour more acceptable (Hardy et al., 2013), as norms should directly influence behaviour:

6. Experienced downloaders find piracy-related behaviour more acceptable

Lastly, individual norms on piracy-related behaviour should be weaker than the social norms (Hardy et al., 2013), as this is a controversial issue.

7. The individual norms are weaker than the social norms regarding piracy-related be-haviour

(10)

4 Methodology

The methodology section consists of three subsections. The research entailed a vignette experiment, based on the design in Hardy et al. (2013). Firstly, the actual experimental circumstances, such as the subject sample, will be discussed. Secondly, the scenarios (vi-gnettes) and the variables will be discussed. Thirdly, the treatments and incentives in the experiment will be explained.

4.1 Respondents

The experiment ran on-line, a link to the survey either given on-line through digital media, or off-line by recruiting participants in university libraries. Several libraries from different faculties from the University of Amsterdam were used. As all participants were either enrolled or had been enrolled in higher education, the sample is not quite represen-tative for the population. However, ensuring a certain type of respondents did enable the research to have a well-mixed sample inside this subgroup. Moreover, this sample is relatively experienced with respect to piracy-related activities.

Some general statistics are shown in Table 1. In total, 90 respondents submitted

responses. Excluding the incomplete or irrelevant responses, 71 responses remained.2 _Firstly,

the majority of respondents was Dutch, with various Western nationalities in the remaining respondents. Secondly, a slight majority of respondents was female. Thirdly, the average age was almost 23. The age distribution is also shown in Figure 1 in the Appendix on page 29.

Table 1

Respondents’ descriptives

Dutch male age

share or average 72% 43% 22.89

Aside from the two treatments containing the 20 scenarios, explained in section 4.2, subjects responded to six questions regarding their downloading behaviour. All six ques-tions were designed to elicit information as to whether the subject ever shown a particular behaviour, e.g., whether the subject had once downloaded a film. The questions differed with respect to the good or the behaviour, including fiction books, nonfiction books, films, music, and sharing copyright goods on-line or with friends. As the answers were either yes

2_{One respondent aged 55 was omitted, in order to be more certain only students’ responses are used in}

(11)

or no, the averaged answers offer a general picture of their downloading behaviour, as shown in Table 1.

Table 2

Share of respondents having downloaded the copyright goods

Fiction book Nonfiction book Film Music Friends On-line

share 38% 58% 94% 90% 62% 17%

4.2 Scenarios and variables

The experiment largely consisted of a set of 20 scenarios, in which a non-specified ’person A’ obtains an (e-)book in some way. While the actions are morally more or less wrong, normative terms were explicitly avoided in the scenarios, in order to ensure objectivity. Each scenario was to be rated by the subjects for its acceptability, i.e., reports of individual and social norms. The Likert scale used ranged from completely unacceptable (1) to completely acceptable (7).

The scenarios differed through five binary variables, which led to a possible set of

25 _{= 32 scenarios. By differing each variable, the isolated effect should appear in statistical}

analysis. However, a number combinations of values did not produce plausible scenarios, hence the list was shortened to 20 scenarios. The omitted scenarios will be discussed. Firstly, the variable physicality determined whether the act entailed stealing a physical or an e-book (Hardy et al., 2013). For instance, as the variable changes, ceteris paribus, the scenario changes from physical theft from a store to a digital theft from a store.

Secondly, the variable loss producer was included to measure the moral considerations of not paying the producer, or essentially all other entities in the value chain, except for the reseller. The share of the purchasing price that is not taken by the reseller differs for copyright goods, even between goods of one type. Hence, this variable ought to be sufficiently general to incorporate it in any scenario. The addition of whether the person would have bought the good otherwise should do the trick: the willingness to pay (W T P ) for the good either being larger or smaller than the price (P ) ensures in any scenario whether the producer has incurred a loss.

Thirdly, the loss incurred by the direct source determined whether or not the source lost a copy of the (e-)book (Hardy et al., 2013). For example, the variable changes stealing an e-book from a friend to copying an e-book from a friend. This variable refines the rivalry associated with the physicality variable in the following way. When the act entails stealing a copy of an e-book (physicality = 0 and losssource = 0) is compared to stealing a copy of a book physicality = 1 and losssource = 1, one might easily attribute the difference in norm

(12)

to the rivalry of the good (due to the good being ’real’, the act is morally worse). However, whether or not the source incurs a loss is an important factor to take into account when judging the act: stealing an e-book (without copying it), physicality = 0 and losssource = 1 is also possible. The opposite, where physicality = 1 and losssource = 0, was not a plausible scenario, as it implies the laborious act of scanning a complete book from a store and then returning it without consequences.

The other variables were peer and sharing (Hardy et al., 2013). The peer variable determined whether the direct source was a friend or not, e.g., copying an e-book from a friend, or downloading it from an unknown source. The sharing variable entailed whether or not the person shared the newly obtained good on-line with others. This is regarded by some as a "noble act" (Sandler, 2013, p. 305), essentially a deliberate extension of regular downloading (which often is accompanied with temporarily uploading the good as well).

(13)

ASPECTS OF PIRA C Y-RELA TED BEHA VIOUR 13 Table 3 List of scenarios

No. Scenario Physical Loss producer Loss source Peer Sharing

1 Download e-book, no payment, private use, would not have bought. 0 0 0 0 0

2 Download e-book, no payment, shares it, would not have bought. 0 0 0 0 1

3 Copies e-book friend, no permission,_{private use, would not have bought.} 0 0 0 1 0

4 Copies e-book friend, no permission, shares it, would not have bought. 0 0 0 1 1

5 Download e-book via friend’s code, no permission,_{website, private use, would not have bought.} 0 0 1 1 0

6 Download e-book via friend’s code, no permission,_{website, shares it, would not have bought.} 0 0 1 1 1

7 Obtain store, no payment, private use, would not have bought. 1 0 1 0 0

8 Obtain store, no payment, private use,_{shares it, would not have bought.} 1 0 1 0 1

9 Obtain friend, no permission, private use, would not have bought. 1 0 1 1 0

10 Obtain friend, no permission, shares it, would not have bought. 1 0 1 1 1

11 Download e-book, no payment, private use, would have bought. 0 1 0 0 0

12 Download e-book, no payment, shares it, would have bought. 0 1 0 0 1

13 Copies e-book friend, no permission, private use, would have bought. 0 1 0 1 0

14 Copies e-book friend, no permission, shares it, would have bought. 0 1 0 1 1

15 Download e-book via friend’s code, no permission,_{website, private use, would have bought.} 0 1 1 1 0

16 Download e-book via friend’s code, no permission,_{website, shares it, would have bought.} 0 1 1 1 1

17 Obtain store, no payment, private use, would have bought. 1 1 1 0 0

18 Obtain store, no payment, private use, shares it, would have bought. 1 1 1 0 1

19 Obtain friend, no permission, private use, would have bought. 1 1 1 1 0

20 Obtain friend, no permission, shares it, would have bought. 1 1 1 1 1

(14)

Table 4

List of averaged ratings across treatments, without sharing

No. Scenario Physical LossProd Loss source Peer Sharing Rating

1 Download e-book, no payment, private use, would not have bought. 0 0 0 0 0 4.30 3 Copies e-book friend, no permission, private use, would not have bought. 0 0 0 1 0 2.99 5 Download e-book via friend’s code, no permission,_{website, private use, would not have bought.} 0 0 1 1 0 1.87 11 Download e-book, no payment, private use, would have bought. 0 1 0 0 0 3.95 13 Copies e-book friend, no permission, private use, would have bought. 0 1 0 1 0 3.03 15 Download e-book via friend’s code, no permission,_{website, private use, would have bought.} 0 1 1 1 0 2.04 7 Obtain store, no payment, private use, would not have bought. 1 0 1 0 0 2.73 9 Obtain friend, no permission, private use, would not have bought. 1 0 1 1 0 2.78 17 Obtain store, no payment, private use, would have bought. 1 1 1 0 0 2.77 19 Obtain friend, no permission, private use, would have bought. 1 1 1 1 0 2.56 Note: The five aspects are shown as binary variables: if the variable shows "1", it is present, if it shows "0",

it is not.

Tables 4 and 5 show the averaged ratings for the scenarios (both treatments averaged). In Table 4 the scenarios did not include sharing the good, whereas Table 5 did. Although this merely gives a general insight into the impact of the different variables, some larger differences appear in Table 4, such as lower ratings when obtaining the good from a peer and when the good would have been purchased.

Table 5 shows similar differences, as ratings decrease when a peer is involved. In addition, the difference in the physicality of the good seems to affect ratings negatively as well.

After the two sets of 10 scenarios each, the subjects were asked to report basic infor-mation, such as age, gender, and nationality. In addition, they were asked to report on their downloading behaviour, with questions regarding the downloading and sharing of various copyright goods.

Table 5

List of averaged ratings across treatments, with sharing

No. Scenario Physical LossProd Loss source Peer Sharing Rating

2 Download e-book, no payment, shares it, would not have bought. 0 0 0 0 1 3.22 4 Copies e-book friend, no permission, shares it, would not have bought. 0 0 0 1 1 2.17 6 Download e-book via friend’s code, no permission,_{website, shares it, would not have bought.} 0 0 1 1 1 1.97 12 Download e-book, no payment, shares it, would have bought. 0 1 0 0 1 3.37 14 Copies e-book friend, no permission, shares it, would have bought. 0 1 0 1 1 2.26 16 Download e-book via friend’s code, no permission,_{website, shares it, would have bought.} 0 1 1 1 1 1.96 8 Obtain store, no payment, private use, shares it, would not have bought. 1 0 1 0 1 1.93 10 Obtain friend, no permission, shares it, would not have bought. 1 0 1 1 1 2.05 18 Obtain store, no payment, private use, shares it, would have bought. 1 1 1 0 1 1.97 20 Obtain friend, no permission, shares it, would have bought. 1 1 1 1 1 2.16 Note: The five aspects are shown as binary variables: if the variable shows "1", it is present, if it shows "0",

(15)

4.3 Treatments and incentives

As seen in Tables 4 and 5, each scenario is ethically judged by the participants. How-ever, two treatments form the within-subject experiment to enable an incentive mechanism. A key element of experimental economics, incentives ensure participants treat the question or task seriously and give the answer they expect to yield the highest reward. In this exper-iment, the incentives worked in the following way. First, individual judgement is required for the first set of 10 scenarios, as the question for each scenario reads "please rate how

appropriate you find this scenario". Following this, the social norm is asked for the latter

set of 10 scenarios, with the question "please rate how appropriate you think others in your

environment would find this scenario". The incentive mechanism consists of a comparison

of the averaged individual judgement ratings for every scenario to the social norm ratings given by every participant. The closer the participant’s answers, or predictions, for the 10 scenarios in the social norm treatment were to the averaged individual norm ratings of all participants for the same 10 scenarios, the better the participant’s score. The aggregated

prediction errors determined the ranking amongst participants.3 _{The best scoring half of}

participants received a €5 voucher (either a physical or a digital voucher) for purchasing (e-)books in the Netherlands, as was stated at the start of the experiment.

3_{For example, when the averaged individual norms for scenario 1 is 2, and the subject predicts (in the}

social norm treatment) 3, the error is 1. This calculation occurs 10 times per subject, for every subject reports 10 scenarios in the social norm treatment

(16)

5 Results

In this section I will discuss the various models used in the analysis, in addition to the underlying assumptions. First, the assumptions will be discussed. Second, the general model is discussed, and an adjustment of the general model with respect to the time it took participants to complete the experiment. Third, a model with different effects for the self-reported downloading behaviour is discussed. Fourth, the difference of estimation results between the two treatments will be examined. Fifth, an extended model with interaction terms will be discussed.

Ordered logit regressions are used, as the dependent variable, the ratings of the scenar-ios in different treatments, is ordinal. This way, the effects of the different variables can be compared as to what strength they exert on the ratings. In addition, the model uses standard errors clustered on respondents, as individuals might display biased responses throughout the experiment.

5.1 Assumptions

The assumptions underlying ordered logit regressions are the same as for the generalised linear model: homoskedasticity of the error terms, and parallel lines or proportional odds. First, homoskedastic errors are generally not an issue in survey experiments with an ordered response variable, such as in this experiment. Second, the parallel lines assumption should be evaluated. All models in the subsequent analyses are subjected to Wald tests, in order to assess whether the parallel lines assumption holds for all variables (Williams, 2006). All models, except for the first model in Table 8 on page 19, showed an insignificant Wald test statistic. In other words, for all models except one the parallel lines assumption holds.

5.2 General regression

In Table 6 the results are shown for the first regression: the first model includes all respondents without controlling for age, gender, and nationality, including respondents having only submitted the first part of the experiment. This is why this model yielded the highest number of observations: the participants that only submitted the first part of the experiment (either the individual or the social treatment scenarios) are included. Most factors show a significant influence (physical, loss source, peer, and sharing), except for loss

producer. Three of the four significant factors show a negative effect on ratings, with loss source being the largest, which implies respondents decreased their ratings if the factor was

included in the scenario. In other words, when either one or a combination of loss source,

(17)

ASPECTS OF PIRACY-RELATED BEHAVIOUR 17 Table 6

Ratings for (1) all observations, (2) controlled for characteristics, (3) controlled for time

Model (1) (2) (3)

coefficient SE coefficient SE coefficient SE

Physical 0.586∗∗∗ _0.150 _0.625∗∗∗ _0.167 _0.795∗∗∗ _0.212 Loss producer -0.0298 0.102 -0.0811 0.113 0.0114 0.141 Loss source -1.639∗∗∗ _0.151 _-1.743∗∗∗ _0.168 _-2.111∗∗∗ _0.218 Peer -0.546∗∗∗ _0.0986 _-0.476∗∗∗ _0.109 _-0.504∗∗∗ _0.139 Sharing -0.851∗∗∗ _0.0958 _-0.887∗∗∗ _0.107 _-1.078∗∗∗ _0.139 Treatment -0.107 0.0973 -0.0302 0.109 0.156 0.138 Age -0.0122 0.0258 0.0267 0.0362 Gender -0.179 0.107 -0.0104 0.138 Dutch 0.0125 0.119 0.417∗∗ _0.160 N 1590 1320 820

Note: Dependent variable in all models is the rating given by respondents, ranging from completely unacceptable (1) to completely acceptable (7).

Significance levels of 0.05 * 0.01 ** 0.001 *** are used. Standard errors clustered on respondents.

the other hand, physical shows a positive effect, which suggests respondents increased their ratings if the scenario concerned a physical book; all else equal, the book being physical led respondents to find the appropriation behaviour more acceptable, compared to e-books.

The second (base) model in Table 6 controls for age, gender, and nationality.4 _Even

though the effects of the other variables do not change compared to the first model, any differences between these characteristics should be omitted. Aside from the generally un-changed coefficients, the number of observations decreases, as now only full responses are included.

The third model in Table 6 excludes respondents having completed the experiment in

less than 400 seconds.5 _{This exclusion might be relevant, as the available reward might}

in-centivise respondents to quickly finish the experiment without properly reading the questions and scenarios. Most of the factors slightly increase in strength. Moreover, nationality now shows a significant positive effect on ratings; among those that completed in more than 400 seconds, Dutch respondents find the behaviour described in the scenarios more acceptable than other nationalities. What underlies this difference might be cultural, but is unclear.

4_{The nationality variable is created as follows: Dutch = 1, other nationalities = 0.}

5_{This number of seconds entails the respondent using at least 10 seconds per scenario, in addition to 200}

(18)

5.3 Heavy users

The second regression examines heavy downloaders, as seen in Table 7. Heavy down-loaders might have distinctly different views on piracy-related behaviour through their more frequent piracy-related behaviour. Hence, controlling for this should yield more accurate results for the regression. The first model is the base, or second model from the previous regression in Table 6. Here, in the second and third model, a variable for heavy downloaders is used. The variable is constructed in the following way. First, the difference between the individual answers and the question averages, as shown in Table 2 on page 11, is calculated. Second, the differences are summed: having scored less than the average results in a negative difference, which means the individual showed more downloading behaviour, and vice versa. Third, the scores are made positive by adding the absolute value of the measure to all values. This way, the lower the value for the variable heavy, the more piracy-related behaviour is shown.

The regression model including the new variable heavy shows no significant effects. As the expected direction of heavy was negative, i.e., the more piracy-related behaviour one shows, the more acceptable one finds the scenarios, the positive direction in Table 7 is surprising, albeit insignificant. Similarly, the interaction between heavy and treatment does not show any significant effects. Here, too, the expected direction was negative, i.e., in the individual treatment, heavy downloaders would find the scenarios less acceptable.

5.4 Treatment effects

In Table 8, the base model (3) is divided into the two different treatments: individual (1) and social (2) norms. No large differences appear, except for the variable peer showing a significant negative effect in the individual norms treatment, whereas no effect remains in the social norms treatment. This suggests that if a friend was the source of the appropri-ated (e-)book, respondents found the behaviour less acceptable, whereas in the social norm treatment, no such effect was visible. However, the first model in Table 8 violated the par-allel lines assumption underlying the ordered logit regression, as the Wald test presented a significant value. Hence, the results from this particular model can only provide suggestions for further analysis.

5.5 Interaction terms and complete model

Lastly, in Table 9, the general model (1), a model including all interaction effects (2), and the second model omitting quick responses (3) are shown. In the second and third model, the following interaction terms are included. Firstly, the interaction term of physical

(19)

Ratings in (1) the base model, (2) controlling for heavy users, and (3) including an interaction term for heavy users

Model (1) (2) (3)

Physical 0.625∗∗∗ _0.167 _0.624∗∗∗ _0.167 _0.624∗∗∗ _0.167 Loss producer -0.0811 0.113 -0.0810 0.113 -0.0809 0.113 Loss source -1.743∗∗∗ _0.168 _-1.743∗∗∗ _0.168 _-1.743∗∗∗ _0.168 Peer -0.476∗∗∗ _0.109 _-0.476∗∗∗ _0.109 _-0.476∗∗∗ _0.109 Sharing -0.887∗∗∗ _0.107 _-0.888∗∗∗ _0.107 _-0.888∗∗∗ _0.107 Treatment -0.0302 0.109 -0.0303 0.109 -0.0314 0.112 Age -0.0122 0.0258 -0.0127 0.0259 -0.0127 0.0259 Gender -0.179 0.107 -0.174 0.109 -0.174 0.109 Dutch 0.0125 0.119 0.0139 0.119 0.0138 0.119 Heavy 0.0640 0.189 0.0532 0.259 Heavy * treatment 0.0217 0.365 N 1320 1320 1320

Significance levels of 0.05 * 0.01 ** 0.001 *** are used.

Table 8

Ratings in (1) only for the individual norm treatment, (2) only for the social norm treatment, and (3) the base model

Model (1) (2) (3)

Physical 0.961∗∗ _0.306 _0.598∗∗ _0.211 _0.625∗∗∗ _0.167 Loss producer 0.278 0.176 -0.261 0.149 -0.0811 0.113 Loss source -2.133∗∗∗ _0.345 _-1.530∗∗∗ _0.225 _-1.743∗∗∗ _0.168 Peer -0.620∗∗∗ _0.153 _-0.237 _0.177 _-0.476∗∗∗ _0.109 Sharing -1.040∗∗∗ _0.183 _-1.027∗∗∗ _0.177 _-0.887∗∗∗ _0.107 Age -0.0848∗ _0.0379 _0.0622 _0.0356 _-0.0122 _0.0258 Gender -0.165 0.149 -0.241 0.155 -0.179 0.107 Dutch -0.356∗ _0.170 _0.396∗ _0.169 _0.0125 _0.119 Treatment -0.0302 0.109 N 660 660 1320

(20)

and peer is used, as obtaining a book from a friend might be different from obtaining an e-book from a friend, or a book from a store. Second, the interaction term of physical and

sharing is used, as sharing a book might be different from sharing an e-book. In addition,

the variable for heavy users, from the regression models in Table 8 is included. The third model omits the responses submitted in under 400 seconds, as used in Table 6.

The interaction terms generally show significant effects. First, the interaction term of physical and peer shows a significant, positive effect on respondents’ ratings, suggest-ing the scenarios in which a physical book is obtained from a friend are viewed as more acceptable. Second, the interaction term of physical and sharing shows a weakly signifi-cant, negative effect, suggesting sharing a physical book is considered worse than sharing an e-book. Third, the variable for heavy users now shows a positive significant coefficient, which implies respondents that show less downloading behaviour find the scenarios more acceptable, contradicting the expected negative value. However, this is only significant in the second model. Fourth, omitting the quick responses does not alter the results in a large way, except for three coefficients. Firstly, nationality now shows a significant, positive effect, suggesting Dutch respondents generally find the scenarios more acceptable, similar to the effect found in the first regression models in Table 6 on page 17. Secondly, the interaction term of physical and sharing now becomes insignificant, suggesting the effect found in the second model does not exist. As not many scenarios included this combination, possibly the effect disappears through the now decreased number of observations satisfying the interaction term. Thirdly, as mentioned, the coefficient heavy becomes insignificant.

(21)

Ratings in (1) the base model, (2) including all interaction terms, and (3) omitting quick responses

Model (1) (2) (3)

Physical 0.625∗∗∗ _0.167 _-0.426 _0.276 _-0.372 _0.364 Loss producer -0.0811 0.113 -0.0805 0.115 -0.00481 0.143 Loss source -1.743∗∗∗ _0.168 _-1.448∗∗∗ _0.172 _-1.808∗∗∗ _0.225 Peer -0.476∗∗∗ _0.109 _-1.104∗∗∗ _0.202 _-1.216∗∗∗ _0.260 Sharing -0.887∗∗∗ _0.107 _-0.739∗∗∗ _0.145 _-0.942∗∗∗ _0.182 Treatment -0.0302 0.109 0.122 0.189 0.276 0.239 Age -0.0122 0.0258 -0.0107 0.0259 0.0308 0.0368 Gender -0.179 0.107 -0.246∗ _0.110 _-0.0577 _0.141 Dutch 0.0125 0.119 0.00454 0.118 0.413∗∗ _0.160 Physical * peer 1.718∗∗∗ _0.227 _1.873∗∗∗ _0.297 Physical * sharing -0.337 0.222 -0.323 0.283 Heavy 0.105∗ _0.0473 _0.0575 _0.0657 Peer * treatment -0.461∗ _0.225 _-0.437 _0.282 N 1320 1320 820

(22)

6 Discussion and conclusion

This research shows mixed results: on the one side, a number of hypotheses regarding the influence of the varying aspects on the norms reported by the respondents were confirmed, yet on the other side, some surprising results appeared. A number of findings contradicted with the hypotheses concerning the effect on the acceptability rating of the scenarios. An overview of the status of every hypothesis is shown in Table 10.

First, the influence of peer, i.e., whether the scenario entailed the (e-)book being taken from a friend, instead of store, was significantly negative in any model, except when the social norm treatment was considered alone in Table 8 on page 19. In the models where

peer showed this significant effect, most often the influence was weaker than any of the other

significant effects. These results suggest that the act of piracy, or related behaviour, is found significantly less acceptable when the (e-)book is taken from a friend, as compared to a store. This effect is generalises to films and to other samples (Hardy et al., 2013; Krawczyk et al., 2015), further strengthening the importance of the source from where the copyright was taken.

Second, the further sharing of the good, as examined through the variable sharing, showed a similarly consistent negative influence on the scenarios’ acceptability, in all models. In other words, respondents decreased the reported norms in scenarios where the (e-)book was shared with others digitally. In most models, the effect was twice as strong as the effect of the peer variable, suggesting sharing is seen as considerably worse than obtaining the good from a friend. Similar to the peer variable, this result extends to films and other samples (Hardy et al., 2013; Krawczyk et al., 2015).

Third, the variable loss source, i.e., whether the source the (e-)book is taken from incurs a loss, shows the strongest effect on the reported norms. As this effect persists in all regression models, this might be the most important aspect when evaluating a behaviour in terms of acceptability. As this, too, can be extended to films and other samples, inflicting a loss appears to decrease the acceptability of piracy-related behaviour (Hardy et al., 2013; Krawczyk et al., 2015).

On the other hand, a number of unexpected results appeared. First, the difference between physical and e-books in the scenarios showed a contradictory result: in most mod-els, the influence of a physical book in the scenario significantly made the behaviour more acceptable for the respondents. This is directly opposite to earlier studies (Hardy et al., 2013; Krawczyk et al., 2015). The following might have caused this. Firstly, the effect might be caused by an interaction with another variable, as piracy-related behaviour with physical books often coincides with other variables, such as the source incurring a loss. Indeed, when an interaction term between physical and peer is included, i.e., the specific combination

(23)

where the book was physical and was taken from a friend, the original effect of physical disappears. Instead, the interaction term shows an even stronger effect, which implies the particular combination made the scenarios more acceptable. The reason for this is unclear. Possibly, the scenario was perceived as borrowing, instead of stealing. However, when this interaction was included, the remaining difference due to the physicality of the book does not alter the other scenarios, such as stealing a book from a store, compared to stealing an e-book from a store. Secondly, as the other findings in the literature concern films (Hardy et al., 2013; Krawczyk et al., 2015), the opposite effect might be due to the difference in copyright good.

Second, whether or not the producer incurred a loss does not seem to affect the accept-ability of the behaviour. The variable loss producer does not show any significant effects in any regression. Moreover, no interaction terms with loss producer showed any effects on the norms related to the behaviour. Two possible interpretations arise. Firstly, the producer in-curring a loss might not affect the acceptability of the behaviour at all. However, this might contradict the results on the direct source incurring a loss, which shows a significant negative effect on the ethical judgement; it is difficult to rhyme the indifference to the producer in-curring a loss with lower acceptability ratings when the direct source incurs a loss. Secondly, the way the influence was tested might not be adequate. Perhaps, the consequences for the producer of not purchasing the good were unclear.

Third, the effects of treatments and the downloading behaviour did not become clear. The absent difference between individual and social norms is surprising, as this was a solid finding in the related literature (Hardy et al., 2013; Krawczyk et al., 2015). Again, this might be due to the difference in copyright goods, or the experiment did not make the difference in treatments as clear as possible. Only the scenarios where the source being a friend and with the individual norm treatment result in lower acceptability ratings. In addition, heavy downloaders do not seem to heavy very different norms with regard to piracy-related behaviour.

In summary, this research offers more insight in how norms regarding downloading behaviour are constructed. Some aspects of piracy-related behaviour concerning (e-)books show similar effects as in other studies, such as the source incurring a loss, and the source being a friend, whereas other aspects show opposite effects, or none whatsoever. The most surprising finding is that the physicality of the book seems to affect individual and social norms differently, as compared to other copyright goods; here, when the behaviour concerns a physical book, the act is seen as more acceptable, which contradicts previous studies (Hardy et al., 2013; Krawczyk et al., 2015). The best explanation for this is that the behaviour might have been perceived as borrowing when a friend was involved, yet this does not explain the

(24)

Table 10

List of hypotheses and rejection status

Hypothesis Status

1: Physicality rejected

2: Loss producer rejected

3: Source friend accepted

4: Sharing accepted

5: Loss source accepted

6: Experienced rejected

7: Individual or social norms rejected

absence of the general effect. Further research could examine these opposed results, for example by using a large range of different copyright goods in a vignette experiment. In addition, the producer incurring a loss does not seem to affect the ethical judgement of the behaviour. Either the implementation of this aspect was not sufficient, or it does not affect the acceptability in this context. In the latter case, in more specific situations it might indeed play a role, for example when the producer’s loss is emphasised more (e.g., comparing financially struggling producers versus producers earning millions). As this information might affect ethical judgement, further research could explicitly differentiatie between these scenarios. Aside from this, other aspects regarding the direct source, and sharing of the good, show similar effects, compared to related copyright good studies, the strongest being the direct source incurring a loss.

The different aspects affecting the acceptability of downloading behaviour could be useful in two ways. First, the weak or absent effects of certain aspects should be emphasised more to elicit a (stronger) negative influence on norms. This particularly holds for the absent effect of the producer incurring a loss; this research suggests this is rarely taken into account when showing piracy-related behaviour. Second, the strong, or established, effects of the other aspects should be emphasised, as they are sure to affect norms negatively. As a large number of parties claim to face significant losses, it is in their interest to lower the acceptability of downloading behaviour.

(25)

ASPECTS OF PIRACY-RELATED BEHAVIOUR 25 7 References

Alexander, C. S., & Becker, H. J. (1978). The use of vignettes in survey research. Public

opinion quarterly, 42 (1), 93–104.

Al-Rafee, S., & Cronan, T. P. (2006). Digital piracy: Factors that influence attitude toward behavior. Journal of Business Ethics, 63 (3), 237–259.

Cubitt, R. P., Drouvelis, M., Gächter, S., & Kabalin, R. (2011). Moral judgments in social dilemmas: How bad is free riding? Journal of Public Economics, 95 (3), 253–264. El Haji, A., & Leenders, M. A. A. M. (n.d.). A moral dissociation between theft and piracy:

An experimental study among u.s. and indian consumers.

Fréchette, G. R. (2011). Laboratory experiments: Professionals versus students. Available

at SSRN 1939219.

Gunter, W. D. (2009). Internet scallywags: A comparative analysis of multiple forms and measurements of digital piracy. W. Criminology Rev., 10 , 15.

Hardy, W., Krawczyk, M., & Tyrowicz, J. (2013). Why is online piracy ethically different

from theft? a vignette experiment. (Tech. Rep.).

Krawczyk, M., Tyrowicz, J., Kukla-Gryz, A., & Hardy, W. (2015). “piracy is not theft!” is it just students who think so? Journal of Behavioral and Experimental Economics,

54, 32–39.

Krupka, E. L., & Weber, R. A. (2013). Identifying social norms using coordination games: Why does dictator game sharing vary? Journal of the European Economic Association,

11(3), 495–524.

Link, B. G., Yang, L. H., Phelan, J. C., & Collins, P. Y. (2004). Measuring mental illness stigma. Schizophrenia bulletin, 30 (3), 511–541.

Martin, E. (2004). Vignettes and respondent debriefing for questionnaire design and evalu-ation. Methods for testing and evaluating survey questionnaires, 149–171.

Nunes, J. C., Hsee, C. K., & Weber, E. U. (2004). Why are people so prone to steal software? the effect of cost structure on consumer purchase and payment intentions. Journal of

Public Policy and Marketing, 23 (1), 43–53.

Peitz, M., & Waelbroeck, P. (2006). Piracy of digital products: A critical review of the theoretical literature. Information Economics and Policy, 18 (4), 449–476.

Picard, R. G. (2004). A note on economic losses due to theft, infringement, and piracy of protected works. Journal of Media Economics, 17 (3), 207–217.

Presser, S., Couper, M. P., Lessler, J. T., Martin, E., Martin, J., Rothgeb,

J. M., & Singer, E. (2004). Methods for testing and evaluating

(26)

from http://poq.oxfordjournals.org/content/68/1/109.short doi: 10.1093/poq/nfh008

Rasinski, K. A., Yoon, J. D., Kalad, Y. G., & Curlin, F. A. (2011). Obstetrician-gynaecologists’ opinions about conscientious refusal of a request for abortion: results from a national vignette experiment. Journal of medical ethics, 37 (12), 711–714. Rettinger, D. A., Jordan, A. E., & Peschiera, F. (2004). Evaluating the motivation of

other students to cheat: A vignette experiment. Research in Higher Education, 45 (8), 873–890.

Sandler, R. L. (2013). Ethics and emerging technologies. Palgrave Macmillan.

Siwek, S. E. (2007). The true cost of copyright industry piracy to the us economy. IPI Center for Technology Freedom.

Svensson, M., & Larsson, S. (2012). Intellectual property law compliance in europe: Illegal file sharing and the role of social norms. New media & society, 14 (7), 1147–1163. Williams, R. (2006). Generalized ordered logit/partial proportional odds models for ordinal

(27)

8 Appendix 8.1 List of scenarios

1. Person A downloaded an e-book on the internet without any payment, for private use. Person A would not have bought the e-book otherwise.

2. Person A downloaded an e-book on the internet without any payment, and shares the e-book on-line. Person A would not have bought the e-book otherwise.

3. Person A copies an e-book from a friend for private use, without the friend?s knowledge. Person A would not have bought the e-book otherwise.

4. Person A copies an e-book from a friend, without the friend?s knowledge, and shares it on-line. Person A would not have bought the e-book otherwise.

5. Person A uses a friend’s one-use only access code, without the friend’s knowledge, for a payable website with e-books to download an e-book for private use. Person A would not have bought the e-book otherwise.

6. Person A obtains a friend’s one-use only access code, without the friend’s knowledge, for a payable website with e-books to download an e-book, and shares the e-book on-line.

7. Person A obtains a book from a store without payment, and scans the book for private use. Person A would not have bought the book otherwise.

8. Person A obtains a book from a store without payment, scans the book, and shares the book on-line. Person A would not have bought the book otherwise.

9. Person A obtains a book from a friend, without the friend?s knowledge, and scans the book for private use. Person A would not have bought the book otherwise.

10. Person A obtains a book from a friend, without the friend?s knowledge, scans the book, and shares the book on-line. Person A would not have bought the book otherwise. 11. Person A downloaded an e-book on the internet without any payment, for private use.

Person A would have bought the e-book otherwise.

12. Person A downloaded an e-book on the internet without any payment, and shares the e-book on-line. Person A would have bought the e-book otherwise.

(28)

13. Person A copies an e-book from a friend, without the friend?s knowledge, for private use. Person A would have bought the e-book otherwise.

14. Person A copies an e-book from a friend, without the friend?s knowledge, and shares the e-book on-line. Person A would have bought the e-book otherwise.

15. Person A uses a friend’s one-use only access code, without the friend’s knowledge, for a payable website with e-books to download an e-book for private use. Person A would have bought the e-book otherwise.

16. Person A uses a friend’s one-use only access code, without the friend’s knowledge, for a payable website with e-books to download an e-book, and shares the e-book on-line. Person A would have bought the e-book otherwise.

17. Person A obtains a book from a store without payment, and scans the book for private use. Person A would have bought the book otherwise.

18. Person A obtains a book from a store without payment, scans the book, and shares the book on-line. Person A would have bought the book otherwise.

19. Person A obtains a book from a friend, without the friend?s knowledge, and scans the book for private use. Person A would have bought the book otherwise.

20. Person A obtains a book from a friend, without the friend?s knowledge, scans the book, and shares the book on-line. Person A would have bought the book otherwise.

(29)

8.2 Density graph of respondents’ age

Figure 1. Distribution of respondents’ age

20 25 30 0.00 0.05 0.10 0.15 N = 66 Bandwidth = 0.7756 Density 8.3 Experimental instructions

(30)

Main questions

Welcome to this short experiment.

Please note that your data will only be used for this research. No other party will see the raw data. Anonymised data can be viewed by other researchers.

On the next two pages you will find several scenarios. You will be asked to rate each scenario. The score ranges from 1 to 7: from completely unacceptable (1), to completely acceptable (7).

There are two sets of questions. The questions in set 1 require your individual judgement. The questions in set 2 will determine whether you will receive the reward (a €5 coupon for purchasing a book in any book store, a Boekenbon). Half of all participants will receive the reward.

Please rate how acceptable you find person A’s behaviour in the following scenarios:

unacceptable unacceptablecompletely unacceptableslightly neutral acceptableslightly acceptable completelyacceptable Person A copies an e-book

from a friend without the friend's permission, for private use. Person A would not have bought the e-book otherwise.

Person A obtains a book from a store without payment, scans the book, and shares the book on-line. Person A would have bought the book otherwise.

Person A uses a friend's one-use only access code, without the friend's permission, for a payable website with e-books to download an e-book, and shares the e-book on-line. Person A would have bought the e-book otherwise.

Person A obtains a book from a friend without the friend's permission, and scans the book for private use. Person A would have bought the book otherwise.

Person A uses a friend's one-use only access code, without the friend's permission, for a payable website with e-books to download an e-book for private use. Person A would have bought the e-book otherwise.

Person A obtains a book from a friend without the friend's permission, scans the book,

(31)

and shares the book on-line. Person A would have bought the book otherwise. Person A copies an e-book from a friend without the friend's permission, for private use. Person A would have bought the e-book otherwise.

Person A obtains a book from a store without payment, scans the book, and shares the book on-line. Person A would not have bought the book otherwise.

Person A obtains a friend's one-use only access code, without the friend's permission, for a payable website with e-books to download an e-book, and shares the e-book on-line.

Person A downloaded an e-book on the internet without any payment, for private use. Person A would have bought the e-book otherwise.

The second set consists of questions regarding your view on the social norm. This is the norm that is shared by your friends, family, colleagues, et cetera. You will have higher chances of winning the prize, the closer your answers in the second set of questions are to the actual social norm. This is calculated in the following way:

Just like you, all other participants answer the previous questions (individual judgement). The average of the answers given by all participants for these questions will be calculated. The closer your answers for the following questions are to these values, the more likely you will receive the reward. In other words, the closer your answers are to the social norm, the more likely you will get the €5 coupon. If the calculated social norm for a question is 3.3, and you answer 3 (slightly unacceptable), your error is 0.3. The sum of your errors will be calculated and compared to the other participants. If you score better than at least half of the participants, you will receive the reward.

Please rate how acceptable you think others in your environment would find person A’s behaviour in the following scenarios: completely unacceptable unacceptable slightly unacceptable neutral slightly acceptable acceptable completely acceptable Person A downloaded an

e-book on the internet without any payment, for private use. Person A would not have bought the e-book otherwise.

Person A uses a friend's one-use only access code, without the friend's permission, for a payable website with e-books to download an e-book for private use. Person A would not have bought the e-book otherwise.

Person A downloaded an e-book on the internet without any payment, and shares the e-book on-line. Person A would not have bought the e-book otherwise.

Person A obtains a book from a friend without the friend's

(32)

Male Female Other

permission, and scans the book for private use. Person A would not have bought the book otherwise.

Person A obtains a book from a store without payment, and scans the book for private use. Person A would not have bought the book otherwise.

Person A obtains a book from a friend without the friend's permission, scans the book, and shares the book on-line. Person A would not have bought the book otherwise.

Person A copies an e-book from a friend without the friend's permission, and shares it on-line. Person A would not have bought the e-book otherwise.

Person A copies an e-book from a friend without the friend's permission, and shares the e-book on-line. Person A would have bought the e-book otherwise.

Person A downloaded an e-book on the internet without any payment, and shares the e-book on-line. Person A would have bought the e-book otherwise.

Person A obtains a book from a store without payment, and scans the book for private use. Person A would have bought the book otherwise.

These page timer metrics will not be displayed to the recipient.

First Click: 0 seconds Last Click: 0 seconds

#QuestionText, TimingPageSubmit#: 0 seconds #QuestionText, TimingClickCount#: 0 clicks

Other questions

Lastly, please find below a small number of questions regarding your downloading behaviour. Here, too, your answers will be anonymised.

Please state your age

(33)

Please state your nationality

Please state the last university you attended

Yes No

Have you once downloaded a fiction e-book without paying? Have you once downloaded a nonfiction e-book without paying? Have you once downloaded a film without paying? Have you once downloaded music without paying? Did you share these kinds of files with friends? Did you share these kinds of files with others on-line?

Reward info

In order to transfer the prize, please write below your e-mail and home address (this should be in the Netherlands). When the data collection has ended, the latest on the 24th of July, I will send you the prize. This will be checked by my supervisor from the faculty of Economics and Business at the University of Amsterdam. If you do not wish to give your details this is fine, however, you will not receive a reward.

Thank you for participating!