BISE – RESEARCH PAPER

Music as a Service as an Alternative to Piracy? An Empirical Investigation of the Intention to Use Music Streaming Services

Although there are no final results of music piracy’s influence on music industry revenues, music pirates are an important target group for the music industry. Music as a Service (MaaS) represents a new form of consuming music. The streaming service is offered on a monthly subscription or an ad-financed basis. Initial user surveys indicated that many music pirates also use such streaming offers. For this empirical study, we questioned 132 music pirates. We conclude that music pirates have a positive attitude towards MaaS. Although the free consumption of music received higher approval, MaaS – owing to social sharing functions and a new pricing model – is a viable alternative to illegal music consumption.

DOI 10.1007/s12599-013-0294-0

music piracy (Liebowitz 2006; Zentner The Authors This article is also available in Ger- 2006), studies have focused on aspects man in print and via http://www. of illegal music downloads. In the litera- wirtschaftsinformatik.de:DörrJ,Wag- ture, two main types of music pirates are Dr. Jonathan Dörr ner T, Benlian A, Hess T (2013) Mu- Dipl.-Volksw. Thomas Wagner () identified. The first group – the so-called sic as a Service als Alternative für savers – experience the current prices of Prof. Dr. Thomas Hess Musikpiraten? Eine empirische Unter- legal music as unfair and too expensive Munich School of Management suchung zur Nutzungsintention von (Al-Rafee and Cronan 2006;Levinetal. Institute for Information Systems Streaming-Services für Musik. WIRT- 2004). The second group – the so-called and New Media SCHAFTSINFORMATIK. doi: 10.1007/ Ludwig Maximilians University s11576-013-0387-x. samplers – use illegal downloads mainly Munich (LMU) to preview a song in order to buy it later Ludwigstrasse 28 © Springer Fachmedien Wiesbaden if they like it (Bhattacharjee et al. 2003b; 80539 Munich 2013 Peitz and Waelbroeck 2004). Germany There have been numerous discussions [email protected] about illegal copying. Digital rights man- url: http://www.wim.bwl.lmu.de agement systems (DRMS) seek to pro- 1Introduction tect digitally saved music from unau- Prof. Dr. Alexander Benlian thorized sharing via the internet. DRMS Department of Law and Economics In 2011, record labels generated 32 % consist of encryptions, authorization def- Chair of Information Systems oftheirglobalturnoverviadigitalsales inition languages, and watermarks. These & Electronic Services channels, a year-on-year growth of 8 %. security mechanisms were introduced to Darmstadt University of Technology Despite this positive development, global regulate the access and usage of the (TU Darmstadt) music industry revenues decreased by digital content for a specific group as Hochschulstrasse 1 31 % between 2004 and 2010 (IFPI well as to prosecute violations of laws 64289 Darmstadt (Hess and Ünlü 2004). However, the cus- Germany 2012, p. 6). Meanwhile, 29.8 million users shared illegal and free music in Europe’s tomer now experiences restrictions while top five markets in 2009 (IFPI 2010, using the music file (Buxmann et al. Received: 2011-03-21 2005). While pirates are skilled at cir- Accepted: 2013-05-17 p. 11). At the same time, only 35 % of cumventing DRMS, legal users are scared Accepted after five revisions by members of illegal networks also paid for off by the technical restrictions result- Prof. Dr. Heinzl. music (IFPI 2012,p.16).Withinthepast few years, the initially very popular peer- ing from the file protection, such as to-peerexchangessuchasBitTorrentand the limited choices of mobile devices or eDonkey (Grasmugg et al. 2003)havebe- the limited distribution options (Sinha gun to be replaced by so-called share- et al. 2010). Therefore, the application hosters such as Rapidshare for distribut- of DRMS may result in a drop in sales ing digital music illegally (IFPI 2010, (Jaisingh 2007). p. 19). By means of warnings and lawsuits, Besides the primary question of the record labels early attempted to stop correlation between the drop in sales and illegal downloading. Already in 2000,

Business & Information Systems Engineering BISE – RESEARCH PAPER

Recording Association of America sued able to empirically show that legal mu- The remainder of this study is struc- the illegal music provider (then sic download platforms such as Apple’s tured along seven sections. First, we state just two years old). At the time, Napster iTunes store are perceived as a substitute our motivation. Section 2 provides an had already managed to acquire a user to music piracy owing to their adequate overview of online music offers in gen- base of 44.6 million (Freedman 2003; price-quality ratio. eral, and especially MaaS. In Sect. 3,we Liebowitz 2006). Music pirates have con- To date, piracy’s de facto impact on present the Theory of Planned Behavior sidered themselves safe because the la- the revenues of the content industry can- as well as our research hypotheses and, bels’ legal departments do not have the not be clearly identified. On the one in Sect. 4, the methodological approach. capacity to properly prosecute each case hand, several studies postulate a nega- Section 5 contains our empirical results, (Al-Rafee and Cronan 2006; Plowman tive correlation (e.g., Sinha and Mandel Sect. 6 our results and an assessment and Goode 2009). 2008; Upshaw and Babin 2010;Wool- of the limitations, while Sect. 7 sum- Besides the protection of music, legal ley 2010), while some studies do not see marizes the results and future research download offers have been steadily im- any effect or a positive effect on revenues opportunities. proved. For instance, consumers can il- (e.g., Oberholzer-Gee and Strumpf 2007; legally download songs from exchange Peukert and Claussen 2012). platforms or from other websites and can A new way to consume digital music 2 Background Information: then decide whether to purchase them af- is now available in the form of MaaS. Current Music Offers on the terwards. The literature describes this as- In contrast to known music offers, so- Internet pect as sampling (Peitz and Waelbroeck called à-la-carte downloads such as the iTunes Store or Amazonmp3, MaaS is 2006). Bounie et al. (2007) and Gopal 2.1 Online Music Offers characterized by two features: instead of et al. (2006) discovered that better sam- downloading a song, music is streamed pling may reduce illegal downloads. It while it is consumed, and users pay a sub- Currently there are three legal distribu- has also been shown that individuals tend scription fee instead of paying per down- tion channels for digital music on the in- to buy music recommended by contacts load. Hence, purchasing and download- ternet which allow users to choose di- in their personal networks (Bounie et al. ing are substituted by a monthly sub- rectlyfromarangeoftitles,albums, 2007;HinzandEckert2010). With the scription fee or an ad-financed stream- and interpretations. In this context, ser- integration of the social network Ping in ing service. First results show that mu- vices similar to web radio outlets such 2010, Apple tried to serve these customer sic pirates also use MaaS, and some have as last.fm and Pandora are not consid- needs. subsequently given up illegal download- ered, since here users are unable to al- Price-setting has also been investigated ing (IFPI 2010,p.9).Ourstudyexam- ter prearranged playlists. Figure 1 pro- several times, since the vast amount of ines whether MaaS offers are accepted by vides an overview of the three legal ser- music pirates seek to legitimize their ille- music pirates as a purchase option; there- vices – differentiated along consumption, gal downloading by pointing to the music fore, it contributes to new technology payment method, and recommendation industry’s unfair price policy (Al-Rafee acceptance research. To date, acceptance system. The offers are arranged from left and Cronan 2006). With a price of €0.99 studies have examined a new technol- to right, following their occurrence. per download, Buxmann et al. (2007) ogy’s adaptability. In this context, it was After a strong drop in revenues at propose a reduction of the price in or- disregarded that comparable technolo- the turn of the millennium, the music der to increase profits. Deviating from gies were already in use, which would industry began to license digital music the initial uniform price-setting strat- make technology substitution more suit- for downloading via the internet. Such egy, the literature contains recommen- able (Polites and Karahanna 2012,p.22). consumption is called download-to-own dations for differentiating the price for Such product usage will be considered in and it is mainly charged for by pay-per- different customer groups (Sandulli and our study, since we consider the relative download. Initially secured by DRMS, Martin-Barbero 2007; Sinha et al. 2010) advantage of a new product – MaaS – since 2009 songs can be downloaded on or the songs offered to them (Buxmann in comparison to an existing product – any device without authentication pro- et al. 2005). Danaher et al. (2010)were illegal downloading. cedures; they can also be shared without

Fig. 1 Current online music consumption services

Business & Information Systems Engineering BISE – RESEARCH PAPER

restriction. However, songs can still con- MaaS provider’s core competency is con- cantly influenced especially by the mu- tain a watermark to detect their origin tent distribution through which it can sic quality offered and the contract du- on exchange platforms (Dörr et al. 2009). now be considered a media company in ration. After analyzing data on the web DRMS is still included in consumption the broad sense (Schumann and Hess radio outlet last.fm, Oestreicher-Singer via download-to-rent. In this case, music 2009,p.12).Thedistributioncoincides and Zalmanson (2009)identifiedacor- files are also located on the user’s device, with the consumption, since the content relation between willingness to pay and but may only be played after an online is streamed and only available during the community activities. Whether a service check of the legal situation. These offers consumption period. Finally, the user of is fully accepted by users and why, has usually apply a subscription model with theMaaSserviceisincludedasexternal not yet been investigated. Our study seeks recurring payment. During the subscrip- factor in the distribution process by him to close this research gap by attempting tion period, the user is able to download or her actively streaming the content. to explain MaaS acceptance by music pi- and play the songs without limitations. Most MaaS providers use freemium as rates, an important target group for the MaaS streams a music file to the user’s their revenue model (Anderson 2009), music industry. device while s/he is consuming the mu- according to which MaaS service cus- sic. Songs are not permanently saved tomers can choose between a free, ad- on the user’s device. MaaS services are financed and a fee-based alternative. The 3 Model Development usually free of charge. In most cases, fee-based premium offer, which costs up € there is also a fee-based premium MaaS to 10 per month, for instance, in- 3.1 Theory Selection offer. The prevailing payment method cludes additional features such as im- is subscription. Both download-to-own proved sound quality, applications for Our study asks whether MaaS is an at- (e.g., iTunes) and download-to-rent (e.g., mobile devices, offline accessibility, and tractive music distribution model for Napster) often offer users streaming op- no advertisements. With offline accessi- music pirates and why. The essential tions. In this context, MaaS services only bility, the user can listen to the music in question is more about the acceptance provide music streaming in combination any country and without any data plan. of a technology than of the acceptance with freemium as revenue model. Music Currently, out of its 10 million registered of a service. Service acceptance stud- streaming offers such as Google or Ama- users, has a paying customer base ies include the integration of the user zon are not considered MaaS, since cus- of 2.5 million. The company has a 15 % in co-production as an essential fac- tomers purchase download-to-own mu- conversion rate (IFPI 2012,p.12).On tor (Meuter et al. 2005). MaaS requires sic and then transfer it into the provider’s average the freemium service’s conver- the active streaming of content. In con- cloud storage. sion rate is around 5 % (Anderson 2009, trast to other so-called self-service tech- p. 28). MaaS offers extended recommendation nologies (e.g., self check-in at airports), Besides web radio stations and some systems in comparison to older, techni- MaaS does not require any deeper user national providers, Spotify, which made cally limited recommendation systems. involvement. In information systems, its debut in 2008 and is based in Sweden, the Theory of Planned Behavior (TPB) Usersarenowabletorecommendmusic is the first successful international on- (Ajzen 1991), the Technology Acceptance as well as receive recommendations. Gen- demand streaming service. Spotify has a Model (TAM) (Davis 1989), and the erated playlists and direct recommenda- user base of 13 million with a presence in Unified Theory of Acceptance and Use tions can also be shared via social net- Germany, the U.S., Belgium, the Nether- of Technology (UTAUT and UTAUT2) works like Facebook or Twitter, and via lands, Great Britain, Sweden, France, (Venkatesh et al. 2003, 2012)–asacom- e-mail. Spain, Denmark, Finland, Switzerland, bination of established theories – are usu- and many more (Spotify 2011). In 2009, ally applied to analyze a new technology’s 2.2 Music as a Service as a Distribution Steereo made its debut in Germany, and it acceptance. We chose TPB as the concep- Concept was taken over by Simfy, the most popu- tual framework, since we consider a com- lar German provider, by the end of 2010. ponent of social influence of third parties Immateriality, simultaneous production Simfy was restructured in 2010 and has a and seek to ascertain attitudes towards and consumption (the uno-actu princi- user base of 1.7 million and a library of 17 MaaS. ple), and integration of the user as ex- million music songs (Simfy 2011). Cur- Previous studies were able to demon- ternal factor are considered substantial rently, there are approximately a dozen strate that music pirates’ attitudes are in- characteristics of a service (Buhl and MaaS providers. Due to country-specific fluenced by social norms (e.g., d’Astous Weinhardt 2009, p. 469). legal regulation for music, most of the et al. 2005; Kwong and Park 2008;Plow- MaaS establishes immateriality services are not transnational. man and Goode 2009;Wangetal.2009). through of music as a Despite MaaS’s relative success, there is While identifying social influence as rel- whole and, as a result thereof, its intan- a limited amount of research and studies evant construct, we consider previous gibility. In contrast to traditional services dealing with the question why users pay studies’ results. At the same time, the (e.g., booking a holiday), production for services, while the service’s basic func- theoretical framework will provide the and consumption do not coincide. Mu- tionality is offered for free (Oestreicher- opportunity for a context-specific exten- sic provided by a MaaS vendor is first Singer and Zalmanson 2009,p.3).Dörr sion. Since UTAUT has emerged from es- produced in a studio. Here, a first copy et al. (2010) examined 132 MaaS users tablished theories, an extension may not is created that will then be distributed. in a survey concerning the influence of be appropriate. Furthermore, constructs However, the MaaS provider is not the service functionalities on users’ willing- such as performance expectancy are dif- content producer but a broker. Consid- ness to pay. From their results, they con- ficult to apply in the context of an enter- ering the music industry value chain, a clude that willingness to pay is signifi- tainment medium. However, TAM and

Business & Information Systems Engineering BISE – RESEARCH PAPER

TPB are often extended or even com- bined. Dibbern et al. (2007)combine TPB and TAM to obtain a different per- spective on users. Thus, user classifica- tions in three perspectives – consumer, technology user, and network member – were achieved. To answer our re- search question, we focus on attitudes to- wards MaaS. We will therefore only ap- ply TPB. Yet, this theory includes the technical dimension via an observed be- havioral check, the social dimension in form of a construct of subjective norm, and the consumption-oriented compo- nent in form of a construct of atti- tude. The study thus focuses on the consumption-orientated component. TPB is a general theory from the so- cial sciences that seeks to predict and ex- Fig. 2 Basic research model to explain MaaS acceptance plain individuals’ behavior (Ajzen 1991). The dependent variable is behavioral in- tention, which is explained by three in- the influence factor plays an important H3: Perceived behavioral control is pos- dependent variables: attitude, subjective role. However, existing studies on digi- itively related to intention to use norm, and perceived behavioral control. tal music show mixed results. While in a MaaS. A meta-analysis of 83 studies shows an student sample, LaRose and Kim (2007) Figure 2 summarizes the basic research average correlation of 0.53 between in- do not discover a significant influence of model, which is based on TPB. tention and actual behavior (Sheppard subjective norm on intention, this corre- et al. 1988). TPB is a further develop- lation is shown in other studies (d’Astous 3.3 Extension for Specifics of MaaS ment of the Theory of Reasoned Ac- et al. 2005; Kwong and Park 2008;Levin tion (TRA) (Ajzen and Fishbein 1980), et al. 2007; Plowman and Goode 2009; A pirate must decide whether s/he wants while adding perceived behavioral con- Wang et al. 2009). In the past few years, to use MaaS or to continue to ob- trol as additional predictor of user in- music labels initiated campaigns against tain music illegally. Both sources can be tention (Ajzen 1991). TRA is the most copyright violations in order to raise free, while MaaS is a legal offer. Dur- influential theory in terms of explaining public awareness of music piracy. Al- and predicting human behavior. It has ing the evaluation of the new offer, s/he though music pirates seek to trivialize been successfully applied in various con- will compare the offers’ characteristics in their crimes, they are aware that par- texts (Sheppard et al. 1988), such as in- both absolute and relative terms (Rogers ents, teachers, and some opinion-formers formation systems usage research (Ben- 1995, p. 212). lian et al. 2009; Hildenbrand et al. 2007; consider illegal music downloading to be Interviews in a previous study demon- Pavlou and Fygenson 2006). morally reprehensible and that they re- strate that MaaS recommendation fea- gard legal alternatives as preferable. Thus, tures are perceived as particularly appeal- 3.2 The Theory of Planned Behavior we derive our second hypothesis: ing. Students who already used the ser- H2: Subjective norm is positively related vice were asked which features they con- TPB assumes that attitudes towards the to intention to use MaaS. sider innovative and useful. The feature examined object influence behavior (in The last predictor of TPB is perceived of sharing entire playlists with friends this case, MaaS usage) based on the as- behavioral control, which describes the on social networks such as Facebook was sumption that attitude already includes subjectively perceived difficulty of an ac- considered particularly valuable. Online the essential conviction and assessment tion.Thisincludesbothexternalfactors recommendations can be understood as of the behavior’s expected results. In our such as time, situational opportunities, online word of mouth (Hennig-Thurau case, music pirates affectively evaluate and the cooperation of others involved, et al. 2004;Sunetal.2006) and therefore MaaS services and their various charac- as well as internal factors such as per- as a special form of conventional word of teristics and thus adopt an approach to- sonal skills to meet the object’s demands mouth. The key aspects of this model are wards them. Therefore, we hypothesize: (Ajzen 2002). In the case of MaaS, per- the constructs online opinion leadership H1: Attitude towards MaaS is positively ceived behavioral control represents both and online opinion search. Online opin- related to intention to use MaaS. the individual’s mental ability to handle ion leadership describes the characteris- The subjective norm describes the most MaaS and his or her technical equipment tic of an individual distributing his or her important reference persons’ expecta- for using the service. If an individual opinion over the internet. Online opin- tions of individual behavior. These ex- posses the necessary knowledge and/or ion search refers to the search for recom- pectations constitute a form of social the necessary equipment (e.g., a smart- mendations. This approach has already pressure and are considered relevant in- phone for mobile usage), this will pos- been applied in the context of digital mu- tention formation factors (Ajzen 1991). itively affect the intention to use MaaS. sic (Sun et al. 2006). In the process, music Especially in the context of illegal actions Thus, we hypothesize: recommendations are given or received

Business & Information Systems Engineering BISE – RESEARCH PAPER

in social networks or via e-mail. As de- that MaaS was compared with the previ- costs compared to illegal offers. Based on scribed in Sect. 2, MaaS offers have com- ously used sourcing channel in several as- this idea, Danaher et al. (2010) demon- prehensive recommendation systems that pects. Also a music pirate compares the strate, in the case of TV shows, that indi- facilitate social exchange, especially on new opportunity of receiving music with viduals are willing to pay small amounts social networks such as Facebook. Sev- his or her existing ones. Since there is rather than accept lengthy searches in eral providers have already introduced a no construct in the literature to describe illegal networks. mandatory Facebook link to fully inte- the comparison of legal and illegal mu- A measure against an unlawful act is grate social recommendations into the sic sourcing, we have developed a new usually a criminal proceeding, which is service. Individuals who like to give or construct in this context, referred to as considered to have a preventive effect on receive music recommendations will be relative advantage of MaaS. However, the the intention to commit a criminal of- positively influenced regarding their atti- construct’s aspects are based on the char- fense (Straub 1990). Depending on the tudes towards MaaS. We therefore derive acteristics identified as important influ- assessment of the law-abiding actions, the following hypotheses: ences on music piracy in the academic lit- music pirates consider their actions risky. H4: Submission of music recommenda- erature: sound quality (e.g., Bhattachar- It has been shown empirically that a tions is positively related to attitude jee et al. 2003a, 2003b,Fetscherinand higher perceived risk of criminal prose- towards MaaS. Zaugg 2004; Gopal et al. 2006; Plowman cution has a negative influence on ille- H5: Search for music recommendations and Goode 2009), search costs (e.g. Jais- gal downloading behavior (Chiang and is positively related to attitude to- ingh 2007;PeitzandWaelbroeck2006), Assane 2008; Chiou et al. 2005;Nanded- wards MaaS. law-abiding actions (e.g., Al-Rafee and kar and Midha 2009). If an individual be- While streaming the music, no music files Cronan 2006; Chiang and Assane 2008; lieves that a legal norm contradicts ille- are stored on the user’s device and there- Chiou et al. 2005;KwongandLee2002; gal music acquisition, s/he considers his fore no right of use is granted. Past stud- Nandedkar and Midha 2009; Plowman or her action to be risky and will prefer ies have investigated collectors and their and Goode 2009), and moral scruples legal music offers. desire to own media such as video cas- (e.g., Chen et al. 2008; Chiou et al. 2005; Based on equity theory (Adams 1963; settes, DVDs, or music CDs (Henke and Coyle et al. 2009;FraedrichandFerrell Kabanoff 1991), the perceived fairness Donohue 1989;Mann2010;Sullivanand 1992;KwongandLee2002; Plowman and results from an individual’s assessment Hibbert 2006). Despite advancing digi- Goode 2009). of own accomplishments, compared to tization, some collectors still prefer pos- The compression of digital music in others’ achievements (Glass and Wood session of digital content such as music MP3 format reduces sound quality. This 1996). In the context of piracy, (Burkart 2008). Since a MaaS user who is of great importance for users that con- Douglas et al. (2007) identify recipro- likes to own music only has the right sume a lot of music (Gopal and Sanders cal influencing variables such as sense of to access the music library during sub- 2003; Plowman and Goode 2009). De- guilt as the most influential determinants scription, s/he will tend to have a nega- pending on the provider and the pricing of perceived fairness. Music pirates per- tive attitude towards MaaS. We therefore model, MaaS services offer bit rates be- ceive the price-setting as unfair and assess hypothesize: tween 128 kB/s and 320 kB/s. Illegal mu- thevalueofmusictobelower.Theyare H6: The desire to own music is neg- sic platforms offer various bit rates. It was atively related to attitude towards shown that music rates were above 128 of the opinion that illegal downloading MaaS. kB/s and that the sound quality increased is neither unethical nor criminal (Chen To test the payment model’s influence on over time (Bhattacharjee et al. 2003a; et al. 2008;Coyleetal.2009). As a re- the attitude towards MaaS, we introduce Fetscherin and Zaugg 2004). However, sult, the positive attitude towards illegal the construct flat rate preference. This the sound quality (caused by bit rate music sources rises with a lacking sense construct is found in the taxi meter effect fluctuation) and failure-free function- of fairness (Kwong and Lee 2002;Plow- and can be traced back to the approach ing (caused by strategic distribution of man and Goode 2009). If a music pirate of mental accounts (Heath and Soll 1996; flawed files of record labels) of illegally considers his or her behavior as unfair, Thaler 2008). According to this approach, downloaded music files remain a prob- this takes the form of moral concerns. Be- a consumer possesses mental accounts lem of illegal platforms. Therefore, MaaS cause MaaS is a legal music offer, con- or budgets, where costs and the utility is more advantageous here than the illegal sumers need no longer worry about un- of a good are registered. Hence, a cus- channels. fair behavior. MaaS therefore offers a fur- tomer enjoys his or her products more Search costs refer to the time invested ther advantage in this aspect compared to when paying a usage-independent (flat) by an individual to meet his or her search illegal downloading. rate compared to a usage-dependent rate. objective. In literature, search costs are From an objective perspective, MaaS The costs occur only during selection considered the most decisive costs of il- offers advantages in the aforementioned and not during consumption (Prelec and legal music consumption and are inte- dimensions compared to music piracy. Loewenstein 1998). In contrast to usage- grated as a negative influence into several Overall, the dimensions have a positive dependent pay-per-download, MaaS of- theoretical models (Jaisingh 2007;Peitz influence on attitude towards MaaS. We fers such a usage-independent plan. We and Waelbroeck 2004, 2006). Besides the have combined these aspects in one con- therefore derive the following hypothe- time invested in the search, a user must structwhichwehaveaddedasafurther sis: sometimes repeat a search owing to in- determinant of attitude. Therefore, we H7: Flat rate preference is positively re- correct or faulty music files. In legal of- derive the following, last hypothesis: lated to attitude towards MaaS. fers, searching is quicker and mostly lim- H8: Relative advantage of MaaS is pos- A qualitative prestudy questioning stu- ited to one provider. Thus, MaaS users itively related to attitude towards dents on MaaS acceptance demonstrated can expect considerably lower search MaaS.

Business & Information Systems Engineering BISE – RESEARCH PAPER

Fig. 3 Extended research model to explain MaaS acceptance

Figure 3 summarizes our extended re- considered only those 132 datasets where mately 5 % of the answers.1 Furthermore, search model. It consists of 2 × 8 = 16 students stated that they had downloaded we tested answers for a nonresponse bias hypotheses. music mainly via illegal channels (ille- by comparing the answers of the last 25 % gal exchange platforms or websites) dur- of participants with the answers of the ing the past month. For this purpose, at remaining participants (Armstrong and 4 Research Methodology the start of the survey, participants were Overton 1977; Lambert and Harrington askedabouttheirusageofvarious(le- 1990). The results showed no significant 4.1 Data Collection and Sample gal and illegal) providers. 28 % of the differences and we could therefore rule Selection sample stated that they downloaded mu- out a nonresponse bias. sic mainly via illegal exchange services, To test our hypotheses, we developed while 72 % used illegal websites. Besides 4.2 Measures an online questionnaire. At the out- the illegal downloading, 37.1 % bought set of the survey, we showed a short music via legal download platforms such To operationalize our constructs, we only video explaining the functionality and as iTunes. 57.6 % bought CDs in retail applied questions that were used in previ- main features of MaaS to ensure that stores and 74.2 % used legal streaming ous studies and adopted them to the con- all participants had the same knowledge portalssuchasYouTube.Atthispoint,it text of MaaS. Since the items were used base. The questions followed. We used already becomes evident that the stream- in previous studies without any valid- the survey software Unipark by Global- ing service, generally free of charge, is ity problems, we thus ensured that con- park to create our online survey, which a favorite distribution channel of music tent validity was not a concern. All ques- was active for two weeks during Au- pirates. tions were rated on a Likert scale ranging gust 2010. We sent an invitation link Theaverageagewas24,while59% from1to5(where1referstothelowest via e-mail to 8,000 students of a Ger- of respondents were men and 41 % were score on the item scale and 5 the high- man university. Our survey followed the women. Participants took approximately est score). Attitude towards MaaS was usual approach of asking students about 15 minutes to complete the survey. Miss- measured with bipolar pairs of opposites. their habits regarding illegal download- ing values were replaced with the lin- In the literature, the relevant aspects of ing from the internet (e.g., Chen et al. ear trend for that point. This proce- MaaS’s relative advantage can be derived 2008; Gopal et al. 2006; Plowman and dure is used for the calculation of val- as identifiable, separate, and independent Goode 2009; Sinha et al. 2010). We col- ues by means of the assigned forecast val- constructs. These aspects (sound qual- lected 926 primary datasets. We then ues. We subsequently estimated approxi- ity, search costs, actions low in compli-

1The results also remain robust when using other replacement approaches and a case-based exclusion of data sets.

Business & Information Systems Engineering BISE – RESEARCH PAPER

ance with laws, and moral scruples) were 4.4 Validation of the Construct Relative The factor values of the factor analy- measured effectively in the past. In to- Advantage of MaaS sis were then included in the final dataset tal, they explain the higher-order con- as formative indicators for the structural struct relative advantage of MaaS and are The construct relative advantage of MaaS equation model. To validate the forma- therefore regarded as formative indica- describes a higher-order construct with tive construct, we consider the signifi- tors, being independent of each other. the structure reflective first and forma- cances of the path weights and the vari- According to Polites et al. (2012), thus tive second. There are two suitable ap- ance inflation factor (VIF) to test for multicollinearity. The indicators sound the prerequisites of a higher-order model proaches for validating such constructs quality and law-abiding actions show no are given. These models represent mul- (Becker et al. 2012): 1. In the two- stage approach (Becker et al. 2012;Ringle significant path weights, while the indica- tilevel constructs, which are represented et al. 2012), factor scores are generated tors search costs and moral scruples do. by constructs of their own (here: the as- for each of the first-order dimensions, The VIF value, which should not exceed pects of sound quality, search costs, law- which are then used as formative mea- 5, shows that multicollinearity between abiding actions, and moral scruples) in- sures of the second-order aggregate con- the remaining indicators is not a concern stead of by observable indicators (Jarvis structs. 2. In the indicator re-use tech- (Henseler et al. 2009). et al. 2003). Our study uses a multidi- nique (Becker et al. 2012;Ringleetal. Table 4 summarizes the described re- 2 mensional construct reflective first-order, 2012; Wetzels et al. 2009), a higher-order sults. formative second-order and can there- construct can be designed by specify- fore be classified as a Type II conception ing a latent construct that includes all 4.5 Validation of the Structural Equation (Becker et al. 2012). the manifest indicators of the underly- Model Table 1 shows the items and the sources ing lower-order latent constructs. Becker they were drawn from. et al. (2012) recognize the superiority of Since our study includes formative as well the two-step approach in cases where the as reflective constructs and due to a rel- researcher is especially interested in the ative small sample size, we decided to 4.3 Validation of the Reflective higher-level estimates. Since we are espe- use the partial least squares (PLS) ap- proach for the empirical validation of Measurement Model cially interested in the influence of rela- our model. PLS is a modeling technique tive advantage of MaaS on attitude, we well suited to assessing complex predic- To establish content validity, we adopted decided to use a two-step approach, fol- tive models. Compared to other proce- only constructs used in previous stud- lowing the procedures of Benlian and dures, PLS offers the advantage of mod- ies. Factor loadings should be above 0.70 Hess (2010), Choudhury and Karahanna eling latent constructs under conditions to establish an indicator reliability of at (2008), Lin et al. (2005), as well as Polites of non-normality and small sample sizes. least 50 % (Hair et al. 1998). A viola- and Karahanna (2012). The PLS algorithm minimizes residual tion of this threshold for subjective norm To validate the higher-order construct, variances to enhance optimal predictive and perceived behavioral control can be we first conducted a confirmatory factor power (Chin 1998;FornellandBook- accepted, since factor loadings are above analysis of the four constructs of the first stein 1982). To reach this goal, PLS esti- 0.40 while satisfying all other follow- dimension (sound quality, search costs, mation is performed by iterations of re- ing thresholds (Götz and Liehr-Gobbers law-abiding actions, and moral scruples) gression, which is why no further sample 2004, p. 723). All constructs showed a (Diamantopoulos and Winklhofer 2001). distribution assumptions needed to be composite reliability significantly above We then used these calculated factors as made (Lohmöller 1989). We used Smart- the critical value of 0.70. Furthermore, formative indicators for the higher-order PLS version 2.0 M3 for the path analy- the average variance extracted (AVE) val- construct (Jöreskog et al. 2006). sis and as a bootstrap re-sampling tool ues all exceeded the 0.50 thresholds (Chin For the confirmatory factor analysis, to determine the significance of the paths 1998). Discriminant validity was assessed we used LISREL 8.8. The theoretically within the model (Ringle et al. 2005). derived constructs could mostly be sup- by investigating the latent construct cor- To test the validity of the structural ported by the empirical data. The model 2 2 relations and the square root of their spe- equation model, we used R ,Ball’sQ shows a Chi-squared value 48.25 and a p- cific AVE. The square root of the AVE for as an indicator for predictive relevance, value of 0.12 at 38 degrees of freedom. each construct was much larger than the significance levels of the path weights, The higher-order model therefore rep- and effect size f 2.AQ2 > 0 indicates the specific construct’s correlation with any resents the data structure well (Jöreskog model’s predictive relevance. The Stone- of the other constructs in the model; the 1993, p. 298). Furthermore, fit indexes Geisser test shows that all reflective mea- condition for discriminant validity can should be considered in the evaluation of sured constructs have a Q2 > 0andare thus be regarded as fulfilled (Fornell and the model. The absolute fit index (root therefore relevant for the model’s predic- Larcker 1981). To summarize, all con- mean square error of approximation / tive power. We computed Cohen’s f 2 to structs satisfied the abovementioned re- RMSEA) (0.043) as well as the compar- check each construct’s effect size. A value liability and validity criteria. The sum- ative fit index / CFI (0.99) fulfill the re- of 0.02 indicates a small, a value of marized results can be found in Tables 2 quirements and provide support for a 0.15 a medium, and a value of 0.35 a and 3. goodmodelfit(HuandBentler1999). largeeffectsize(Cohen1988). All our

2In order to assure that the chosen methodical approach is not subject to systematic errors we also applied the item re-use technique (Mode B). Here, the results also show that moral scruples and search costs are the most important influence factors within the construct. The remaining path coefficients of the structural equation model did not result in significant changes confirming the robustness of the results.

Business & Information Systems Engineering BISE – RESEARCH PAPER

Table 1 Items and their origins

Construct Items Source

Intention to use MaaS I intend to use MaaS in its premium (free) version in the next three months. Kwong and Park (2008), (paid and free) I predict that I will use MaaS in its premium (free) version in the next three Venkatesh et al. (2003) months. I plan to use MaaS in its premium (free) version in the next three months. Attitude The idea of using MaaS Ajzen and Fishbein (. . . ) I like – I dislike∗ (1980), Graf (2007) (...)isboring–isexciting (...)isvaluable–isworthless∗ (...)isfavorable–isunfavorable∗ (...)isgood–isbad∗ Subjective norm People who are important to me think I should use MaaS. Ajzen and Fishbein (1980), People who influence my behavior think I should use MaaS. Mathieson (1991) People whose opinions I value recommend that I use MaaS. Perceived behavioral I have the necessary resources to use MaaS. Venkatesh (2000) control I have the necessary knowledge to use MaaS. With the necessary devices, the possibility of access, and the necessary knowledge provided, it would be easy to use MaaS. Submission of My friends think that I am a good source of information for new music in Sun et al. (2006) recommendations the internet. Compared to my friends I am asked more frequently about recommendations for music via the internet. I tend to influence the opinions of others regarding music via the internet. Search for I tend to search in the internet for others’ opinions to find new music. Sun et al. (2006) recommendations I tend to search for the latest information about music on the internet before purchasing music. When I am interested in new music, I search for recommendations via e-mail, chat rooms, or ratings on the internet. Desire to own I would be not sad if I lost my digital music library due to technical issues.∗ Belk (1985) Compared to others, I do not care as much about saving my digital music library.∗ Flat rate preference I like the flat rate, because I do not have to think about my music Lambrecht and Skiera consumption costs. (2006) I feel more comfortable when listening to music when I pay a flat rate. Relative Sound quality My decision to download music legally or illegally is strongly linked with Plowman and Goode advantage the music’s sound quality. (2009) of MaaS The sound quality of downloaded music is better when downloading it from legal providers than from illegal ones. Search costs Interesting and rare titles are easier to find via legal music providers. Trepte et al. (2004) The music I am looking for is easier to find via legal providers than illegal ones. Especially new and the latest titles are easier to find via legal providers than illegal ones. Law-abiding actions Existing laws prohibit effectively illegal music sharing. Kwong and Lee (2002) Existing laws effectively deter illegal downloading. The current prosecution of illegal music sharing is effective. Moralscruples Downloadingsongsillegallyis(...)ethicallyincorrect. Coyleetal.(2009) (...)thesameastheft. (...)areasontofeelguilty.

∗Reverse item

Business & Information Systems Engineering BISE – RESEARCH PAPER

Table 2 Factor loadings, means, standard deviations, and construct reliabilities

Construct Indicators Standardized Mean Standard deviation Construct factor loadings reliability

Intention to use MaaS (I) I_1 0.974 (0.963) 1.710 (3.507) 1.65 (3.57) 1.251 (1.483) 1.18 (1.41) 0.988 (0.978) I_2 0.987 (0.973) 1.641 (3.643) 1.166 (1.435) I_3 0.986 (0.971) 1.611 (3.564) 1.189 (1.471) Attitude (Att) Att_1 0.773 (0.793) 4.377 3.95 0.813 0.79 0.907 (0.908) Att_2 0.780 (0.812) 4.153 1.115 Att_3 0.818 (0.799) 3.569 1.026 Att_4 0.833 (0.833) 3.902 0.929 Att_5 0.861 (0.841) 3.754 0.963 Subjective norm (SN) SN_1 0.625 (0.745) 3.789 2.99 1.025 0.99 0.864 (0.866) SN_2 0.895 (0.870) 2.710 1.316 SN_3 0.931 (0.861) 2.500 1.234 Perceived behavioral PBC_1 0.777 (0.630) 4.669 4.63 0.647 0.57 0.854 (0.833) control (PBC) PBC_2 0.837 (0.870) 4.555 0.782 PBC_3 0.825 (0.858) 4.680 0.665 Submission of music OOL_1 0.893 (0.893) 3.200 2.82 1.377 1.23 0.921 (0.921) recommendations (OOL) OOL_2 0.873 (0.870) 2.663 1.376 OOL_3 0.912 (0.914) 2.624 1.398 Search for music OOS_1 0.879 (0.875) 2.916 2.79 1.493 1.27 0.887 (0.887) recommendations (OOS) OOS_2 0.929 (0.827) 3.137 1.517 OOS_3 0.844 (0.849) 2.322 1.479 Desire to own (DO) DO_1 0.867 (0.872) 1.660 1.89 1.102 1.10 0.875 (0.875) DO_2 0.896 (0.892) 2.130 1.401 Flat rate preference (FP) FP_1 0.958 (0.957) 4.189 4.13 1.089 1.06 0.949 (0.949) FP_2 0.943 (0.943) 4.090 1.143 Sound quality (SQ) SQ_1 0.897 (0.896) 2.910 2.89 1.359 1.20 0.906 (0.906) SQ_2 0.923 (0.923) 2.886 1.289 Search costs (SC) SC_1 0.901 (0.901) 3.035 2.85 1.306 1.11 0.901 (0.901) SC_2 0.907 (0.909) 2.795 1.340 SC_3 0.789 (0.786) 2.957 1.373 Law-abiding actions (LA) LA_1 0.873 (0.701) 2.180 2.14 1.316 1.02 0.861 (0.845) LA_2 0.876 (0.934) 2.297 1.268 LA_3 0.706 (0.854) 1.971 1.088 Moral scruples (MS) MS_1 0.895 (0.895) 2.653 2.49 1.372 1.26 0.935 (0.935) MS_2 0.928 (0.928) 2.390 1.366 MS_3 0.903 (0.903) 2.456 1.446

Notes: Values (in brackets) refer to the measurement model including the intention to use paid (free) MaaS. Since intention to use MaaS was mea- sured for paid MaaS as well as free MaaS, the means and standard deviation values are provided for both constructs. Values of SQ, SC, LA, and MS were calculated by using the indicator re-use technique (Mode B) significant variables showed at least a a clearly positive approach to MaaS. The shows that music pirates are able to use small effect size. Table 5 summarizes the mean of attitude was 3.95 on a five- MaaS from a cognitive perspective and results. point scale. Questions regarding the in- a technical perspective, while the latter tention to use MaaS were asked for the shows a broad acceptance of a flat rate free version and for the premium paid pricing model for music. 5 Results version – with significant differences in the answers. While most pirates would 5.2 Results from the Structural Equation 5.1 Descriptive Results use the free version (mean = 3.57), few Model would pay for MaaS (mean = 1.65). Concerning the descriptive results, we are Strikingly high are the means of per- Our main model regarding usage inten- especially interested in the results for at- ceived behavioral control (4.63) and flat tion comprises 2 × 3 = 6 hypotheses. titude and intentions. Music pirates show rate preference (4.13). The first result Overall, we found support of five of these

Business & Information Systems Engineering BISE – RESEARCH PAPER

Table 3 Correlations and AVEs

1234 5 6 789101112

I 0.965 (0.938) Att 0.387∗∗ 0.662 (0.681∗∗) (0.665) SN 0.536∗∗ 0.481∗∗ 0.686 (0.492∗∗) (0.505∗∗) (0.684) PBC 0.023 0.189∗ 0.181∗ 0.661 (0.285∗∗) (0.198∗) (0.176∗) (0.630) OOL 0.359∗∗ 0.302∗∗ 0.469∗∗ 0.280∗∗ 0.796 (0.312∗∗) (0.286∗∗) (0.451∗∗) (0.275∗∗) (0.796) OOS 0.324∗∗ 0.369∗∗ 0.462∗∗ 0.160 0.595∗∗ 0.724 (0284∗∗) (0.350∗∗) (0.447∗∗) (0.151) (0.596∗∗) (0.723) DO 0.189∗ −0.073 −0.083 −0.361∗∗ −0.273∗∗ −0.210∗ 0.777 (0.041) (−0.072) (−0.110) (−0.359∗∗) (−0.273∗∗) (−0.208∗) (0.777) FP 0.126 0.332∗∗ 0.321∗∗ 0.168 0.231∗∗ 0.275∗∗ −0.021 0.903 (0.246∗∗) (0.336∗∗) (0.325∗∗) (0.155) (0.232∗∗) (0.277∗∗) (0.020) (0.903) SQ 0.380∗∗ 0.181∗∗ 0.445∗∗ −0.041 0.190∗ 0.191∗ 0.152 0.195∗ 0.827 (0.114) (0.184∗) (0.439∗∗) (−0.050) (0.189∗) (0.192∗) (0.154) (0.195∗) (0.827) SC 0.357∗∗ 0.238∗∗ 0.435∗∗ 0.001 0.054 0.213∗ 0.088 0.130 0.434∗∗ 0.752 (0.147) (0.236∗∗) (0.429∗∗) (0.000) (0.056) (0.215∗) (0.088) (0.130) (0.435∗∗) (0.752) LA 0.244∗∗ −0.014 −0.042 −0.132 −0.126 −0.053 0.345∗∗ 0.163 0.182∗ 0.145 0.676 (0.018) (0.013) (−0.013) (−0.089) (−0.112) (−0.100) (0.338∗∗) (0.110) (0.151) (0.159) (0.652) MS 0.410∗∗ 0.324∗∗ 0.375∗∗ 0.070 0.266∗∗ 0.313∗∗ 0.097 0.231∗∗ 0.330∗∗ 0.375∗∗ 0.281∗∗ 0.826 (0.267∗∗) (0.314∗∗) (0.348∗∗) (0.064) (0.266∗∗) (0.314∗∗) (0.098) (0.231∗∗) (0.329∗∗) (0.376∗∗) (0.277∗∗) (0.826) Notes: Diagonal elements are AVEs and off-diagonal elements are correlations between constructs. Values (in brackets) refer to the measurement model, including the intention to use paid (free) MaaS. Values of SQ, SC, LA, and MS were calculated by using the indicator re-use technique (Mode B) ∗p < 0.05; ∗∗p < 0.01; all other correlations are insignificant

Table 4 Validation of the formative measurement model we had 2 × 5 = 10 hypotheses. There are Sound quality Search costs Law-abiding actions Moral scruples only small differences between the path coefficients, since attitude was only asked Standardized factor ns 0.403∗ ns 0.745∗∗∗ for once – for MaaS in general. Out of loadings (0.416∗) (0.734∗∗∗) these 10 hypotheses, we found support Variance – 1.28 – 1.28 for six. We could explain 24 % respec- inflationfactor tively 23 % of the variance of attitude to- (VIF) wards MaaS. We found that the search for music recommendations and the flat Notes: Values (in brackets) refer to the measurement model, including the intention to use paid (free) MaaS rate preference influence the attitude to- ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ns = not significant wards MaaS positively and significantly. The new recommendation functions and the pricing model are therefore functions six hypotheses. Only the influence of per- able to explain 51 % of the variance of users rated highly and which can help ceived behavioral control on intention in intention to use free MaaS. We can ob- build a strongly positive attitude towards the case of paid MaaS was found to be serve a strong difference between the in- MaaS. not significant. This might be explained fluence of attitude on intention to use The most important influence, how- by the low variance in the variable. free MaaS and the intention to use paid ever, can be found in the relative advan- In our model for paid MaaS, subjective MaaS: building a strong attitude there- tage of MaaS compared to illegal chan- norm and attitude were found to posi- fore does not lead directly to high will- nels. Reduced search costs and preven- tively influence usage intention. We could ingness to pay; instead, it is the inten- tion of moral scruples are seen as con- thus explain approximately 32 % of the tion to test free MaaS that is influenced siderable advantages of MaaS. As shown, variance of intention paid. Regarding the by attitudes. sound quality and law-abiding actions model for free MaaS, attitude was found Our extended model focused on the at- are not regarded as advantages by the to have the strongest influence, followed titude towards MaaS. Since we ran two music pirates. Pirates seem satisfied with by subjective norm and perceived behav- models (one for free MaaS usage inten- the sound quality of tracks provided via ioral control. These three constructs were tion, one for paid MaaS usage intention) illegal networks and seem to feel safe

Business & Information Systems Engineering BISE – RESEARCH PAPER

Table 5 Structural equation model results from prosecution. Figure 4 summarizes our results. Predictor Paid MaaS Free MaaS Intention Intention R2 = 0.32 R2 = 0.51 Pathcoefficient Effectsize Pathcoefficient Effectsize 6 Implications and Limitations

H1 Attitude 0.181∗∗∗ 0.04 0.561∗∗∗ 0.48 Our study demonstrated the attractive- H2 Subjective norm 0.466∗∗∗ 0.24 0.184∗∗∗ 0.07 ness of MaaS offers to music pirates. Nev- H3 Perceived behavioral −0.095ns –0.141∗∗ 0.05 ertheless, most music pirates prefer free control MaaS. In this context, providers can gen- erate advertising-based revenues. The in- Predictor Attitude towards MaaS Attitude towards MaaS tegration of fee-based services seems par- R2 = 0.24 R2 = 0.23 ticularly profitable for MaaS providers. Pathcoefficient Effectsize Pathcoefficient Effectsize While configuring the service, the pricing model and additional features offered are H4 Submission of music 0.078ns – 0.068ns – of great importance. recommendations We showed that flat rates are regarded ∗ ∗ H5 Search for music 0.179 0.03 0.162 0.02 as an attractive pricing model by mu- recommendations sic pirates and that this constitutes a H6 Desire to own −0.034ns – −0.040ns – suitable alternative to pay-per-download, H7 Flat rate preference 0.213∗ 0.04 0.224∗ 0.05 which is often considered too expensive H8 Relative advantage of 0.219∗∗∗ 0.05 0.217∗∗∗ 0.05 (Al-Rafee and Cronan 2006). Although a MaaS flatrateisnottheonlyreasonformu- sic pirates to use MaaS offers, it influ- ∗ ∗∗ ∗∗∗ p < 0.05; p < 0.01; p < 0.005; ences the general attitude towards the nsNot significant: n = 132. Effect size: f 2 > 0.02 = small; f 2 > 0.15 = medium; f 2 > 0.35 = large service. Users who do not consider mu- sic piracy owing to moral scruples and higher search costs also show a positive

Fig. 4 Structural equation model results

Business & Information Systems Engineering BISE – RESEARCH PAPER

attitude towards MaaS. This positive ef- is not possible due to a sample bias. Im- Abstract fect may be strengthened by the social en- plications for the general acceptance of Jonathan Dörr, Thomas Wagner, vironment, such as close friends and rel- MaaS cannot be derived from a survey Alexander Benlian, Thomas Hess atives who disapprove of music piracy. among music pirates. Thus, future stud- A reason for the increased willingness to ies should try to develop and empiri- Music as a Service as an pay may also relate to hedonistic social cally test a general, valid research model. Alternative to Music Piracy? benefits, established by integrating social The limitation to German-speaking ar- features into recommendation systems. eas further prevents a generalization of An Empirical Investigation of the Childers et al. (2001) concluded, among the results. Also, mobile network cov- Intention to Use Music Streaming others, that users experience far more Services erage plays a crucial role in the mo- enjoyment in a pleasant and entertain- bile use of MaaS. Furthermore, MaaS ing online shopping environment. Cus- Despite increasing acceptance of dig- providers have to deal with different na- tomers also developed a positive attitude ital channels, total sales in the music tional music rights protection authori- towards a service that has strong design business decreased by 31 % from 2004 ties (e.g., GEMA in Germany). Although features and is easy to use. Technically to 2010. Music piracy is still consid- we provided an information video about and in terms of design, MaaS offers dif- ered one of the main causes for this. MaaS, it must be considered that the fer from existing services; they match the However, several studies found no ef- desired requirements. MaaS providers surveyed students did not have equal fects or even positive effects of illegal should therefore focus on comprehen- knowledge about MaaS. Future studies downloading on record sales. In the should therefore directly focus on MaaS past, piracy has been counteracted es- sive, user-friendly recommendation sys- tems that support social exchanges be- users in order to ensure sufficient knowl- pecially by prosecution and legal offers. edge about MaaS. The survey partici- Music as a Service (MaaS) represents a tween MaaS users. Our study results clearly demonstrate that a platform’s fea- pants were asked about their usage in- new, differing distribution approach in tention rather than their de facto MaaS digital music. In contrast to the well- tures positively influence the attitude to- use. Therefore, there may be a parallel use known music platforms for so-called à- wards MaaS. Besides direct recommen- of MaaS and illegal platforms. Especially la-carte downloads, such as the iTunes dations from friends, users can receive Store, MaaS possesses two important recommendations based on tagged music German MaaS providers do not always characteristics: transmission (stream- channels or collaborative filtering. Fur- have the latest music tracks available on ing instead of downloading) and pric- ther measures based in this relative ad- their portal. This fact may encourage to ing model (flat rate instead of pay-per- vantage, such as adequate search func- continue using illegal downloads in the download). Therefore, the consump- tions, positively influence attitudes to- first place. tion of music by means of purchas- wards MaaS. Our study also derives that ing and downloading is replaced by a measures which make music pirates feel monthly payment service (paid MaaS) that they are behaving unfairly shall be and an ad-supported (free MaaS) ser- continued. However, this is more within 7 Summary and Outlook vice. First user surveys suggest that the power of the industry than of a single many music pirates are making use of online music provider. MaaS is a business model for the distri- these offers. To find out if MaaS is an at- Fromaresearchperspective,ourstudy bution of digital music. The presented tractive distribution channel for music contributes to the existing academic lit- study demonstrates that new offers of pirates, we developed a model to ex- erature by analyzing the acceptance of music consumption can also be an at- plaintheintentiontouseMaaSbased MaaS as a new technology. TPB’s ex- tractive alternative for music pirates. Al- on the Theory of Planned Behavior. planatory power was confirmed. Our pa- though there is no indication of the re- To empirically test this model, we sur- per also uses MaaS as an example to ex- ductionofillegaldownloadsingeneral, veyed 132 music pirates. Among oth- plain attitudes towards new technology music pirates consider the free ad-based more precisely by comparing alternatives ers, the outcome shows that the inten- version of MaaS an alternative. Music pi- tion to use free MaaS is mainly affected (see Fig. 3). To date, acceptance stud- rates who have rejected legal music con- by the attitude towards MaaS, while us- ies have solely focused on the intention sumption due to high prices in the past ing paid MaaS is predominantly a re- to use a new service or product, while may well switch to legal consumption. sult of the influence of users’ closest ignoring other services or products be- peers. The attitude towards MaaS is ing used by customers (Polites and Kara- Not only music is offered via stream- positively influenced by the desire to hanna 2012). Our study includes prior ing services. Streaming portals for videos, receive music recommendations, the use of other products by adding the new games, and e-books are being developed payment type (in the form of a flat rate product’s relative advantage. The study along the same principle as MaaS. In model), and the relative advantage of therefore contributes to both the practi- this context, this represents a significant MaaS compared to illegal choices. cal relevance of and the general research change in property rights. Rather than Keywords: Music as a Service, MaaS, on acceptance. paying for possession, the user pays for Digital goods, Music streaming, Mu- Some limitations must be considered unlimited access to content. After the sic piracy, Business models, Theory of for the results’ interpretation. A student transmission from physical carrier media planned behavior sample cannot be considered representa- to the internet, these offers may be the tive of the population of music pirates. next large change in the consumption of Therefore, a generalization of the results digital goods.

Business & Information Systems Engineering BISE – RESEARCH PAPER

8Remarks Buxmann P, Strube J, Pohl G (2007) Cooper- piracy behavior. Information & Manage- ative pricing in digital value chains – the ment 44(5):503–512 case of online music. Journal of Electronic Fetscherin M, Zaugg S (2004) Music piracy on A previous version of the paper has Commerce Research 8(1):32–40 peer-to-peer networks. In: Proceedings of been published in German within the Chen Y-C, Shang R-A, Lin A-K (2008) The in- the 2004 IEEE international conference on cumulative thesis by Dörr (2012). tention to download music files in a P2P e-technology, e-commerce and e-service, environment: consumption value, fashion, Taipei, p 9 and ethical decision perspectives. Elec- Fornell C, Bookstein F (1982) Two structural tronic Commerce Research & Applications equation models: LISREL and PLS applied References 7(4):411–422 to consumer exit-voice theory. Journal of Chiang EP, Assane D (2008) Music piracy Marketing Research 19(4):440–452 among students on the university campus: Fornell C, Larcker DF (1981) Evaluating struc- Adams JS (1963) Toward an understanding of do males and females react differently? tural equation models with unobservable inequity. Journal of Abnormal and Social Journal of Socio-Economics 37(4):1371– variables and measurement error. Journal Psychology 67(5):422–436 1380 of Marketing Research 18(1):39–50 Ajzen I (1991) The theory of planned behav- Childers TL, Carr CL, Peckc J, Carson S (2001) Fraedrich JP, Ferrell OC (1992) The impact of ior. Organizational Behavior and Human Hedonic and utilitarian motivations for on- perceived risk and moral philosophy type Decision Processes 50(2):179–211 line retail shopping behavior. Journal of on ethical decision making in business or- Ajzen I (2002) Perceived behavioral control, Retailing 77(4):511–535 ganizations. Journal of Business Research self-efficacy, locus of control, and the the- Chin WW (1998) The partial least squares ap- 24(4):283–295 ory of planned behavior. Journal of Applied proach for structural equation modeling. Freedman D (2003) Managing pirate cul- Social Psychology 32(4):665–683 In: Marcoulides GA (ed) Modern methods ture: corporate responses to peer-to-peer Ajzen I, Fishbein M (1980) Understanding for business research. Lawrence Erlbaum, networking. The International Journal on attitudes and predicting social behavior. Hillsdale, pp 295–336 Media Management 5(3):173–179 Prentice Hall, New York Chiou J-S, Huang C-Y, Lee H-H (2005) The Glass RS, Wood WA (1996) Situational deter- Al-Rafee S, Cronan T (2006) Digital piracy: fac- antecedents of music piracy attitudes minants of software piracy: an equity the- tors that influence attitude toward behav- and intentions. Journal of Business Ethics ory perspective. Journal of Business Ethics ior. Journal of Business Ethics 63(3):237– 57(2):161–174 15(11):1189–1198 259 Choudhury V, Karahanna E (2008) The relative Gopal R, Bhattacharjee S, Sanders L (2006) Do Anderson C (2009) Free – the future of a advantage of electronic channels: a multi- artists benefit from online music sharing? radical price. Random House, London dimensional view. MIS Quarterly 32(1):179– Journal of Business 79(3):1503–1533 Armstrong JS, Overton TS (1977) Estimating 200 Gopal RD, Sanders GL (2003) Digital music nonresponse bias in mail surveys. Journal Cohen J (1988) Statistical power and analysis and online sharing: software piracy 2.0? of Marketing Research 14(3):396–402 for behavioral sciences. Lawrence Erlbaum, Communications of the ACM 46(7):107– Arndt J (1967) Role of product-related conver- Hillsdale 111 sations in the diffusion of a new product. Coyle JR, Gould SJ, Gupta P, Gupta R (2009) Götz O, Liehr-Gobbers K (2004) Analyse Journal of Marketing Research 4(3):291– To buy or to pirate: the matrix of music von Strukturgleichungsmodellen mit Hil- 295 consumers’ acquisition-mode decision- fe der Partial-Least-Squares(PLS)-Methode. Becker J-M, Klein K, Wetzels M (2012) Hierar- making. Journal of Business Research Betriebswirtschaft 64(6):714–738 chical latent variable models in PLS-SEM: 62(10):1031–1037 Graf D (2007) Die Theorie des geplanten Ver- guidelines for using reflective-formative d’Astous A, Colbert FO, Montpetit D (2005) haltens.In:KrügerD,VogtG(eds)Theori- type models. Long Range Planning 45(5– Music piracy on the web – how effective en in der biologiedidaktischen Forschung. 6):359–394 are anti-piracy arguments? Evidence from Springer, Heidelberg, pp 33–43 Belk RW (1985) Materialism: trait aspects of the theory of planned behaviour. Journal of Grasmugg S, Schmitt C, Veit D (2003) Internet- living in the material world. Journal of Consumer Policy 28(3):289–310 Quellen zu Peer-to-Peer (P2P)-Systemen. Consumer Research 12(3):265–280 Danaher B, Dhanasobhon S, Smith MD, Telang WIRTSCHAFTSINFORMATIK 45(3):335–344 Benlian A, Hess T (2010) IT standard imple- R (2010) Converting pirates without canni- Hair J, Anderson R, Tatham R, Black W (1998) mentation and business process outcomes balizing purchasers: the impact of digital Multivariate data analyses with readings. – an empirical analysis of XML in the pub- distribution on physical sales and Internet Prentice Hall, Englewood Cliffs lishing industry. In: Proceedings of the 31st piracy. Marketing Science 29(6):1138–1151 Heath C, Soll JB (1996) Mental budgeting and international conference on information Davis FD (1989) Perceived usefulness, per- consumer decisions. Journal of Consumer systems, St. Louis, p 21 ceived ease of use, and user acceptance Research 23(1):40–52 Benlian A, Hess T, Buxmann P (2009) Drivers of of information technology. MIS Quarterly Henke LL, Donohue TR (1989) Functional dis- saas-adoption: an empirical study of differ- 13(3):319–340 placement of traditional TV viewing by VCR ent application types. Business & Informa- Diamantopoulos A, Winklhofer HM (2001) In- owners. Journal of Advertising Research tion Systems Engineering 1(5):357–369 dex construction with formative indicators: 29(2):18–23 Bhattacharjee S, Gopal RD, Lertwachara K, an alternative to scale development. Jour- Hennig-Thurau T, Gwinner KP, Walsh G, Grem- Marsden JR (2003a) No more shadow box- nal of Marketing Research 38(2):269–277 ler DD (2004) Electronic word-of-mouth via ing with online music piracy: strategic busi- Dibbern J, Heinzl A, Schaub N (2007) Determi- consumer-opinion platforms: what moti- ness models to enhance revenues. In: Pro- nanten der Akzeptanz von mobilen Bank- vates consumers to articulate themselves ceedings of the 36th annual Hawaii inter- diensten: Test eines Drei-Perspektiven- on the Internet. Journal of Interactive Mar- national conference on system sciences, Modells. In: Bayón T, Herrmann A, Huber keting 18(1):38–52 Hawaii, p 11 F (eds) Vielfalt und Einheit in der Marke- Henseler J, Ringle CM, Sinkovics RR (2009) The Bhattacharjee S, Gopal RD, Sanders GL tingwissenschaft. Gabler, Wiesbaden, pp use of partial least squares path model- (2003b) Digital music and online sharing: 449–478 ing in international marketing. Advances in software piracy 2.0? Communications of Dörr J (2012) Music as a Service – Ein neues International Marketing 20(4):277–319 the ACM 46(7):107–111 Geschäftsmodell für digitale Musik. epubli, Hess T, Ünlü V (2004) Systeme für das Ma- Bounie D, Bourrreau M, Waelbroeck P (2007) Berlin nagement digitaler Rechte. WIRTSCHAFTS- Pirates or explorers? Analysis of music con- Dörr J, Benlian A, Grau C, Wilde T (2009) INFORMATIK 46(4):273–280 sumption in French graduate schools. Brus- Musikdistribution ohne Digital Rights Ma- Hildenbrand T, Korchminskaya A, Oswald S, sels Economic Review 50(2):167–192 nagement – Eine empirische Analyse der Bieber E, Berchez J-P, Maché N (2007) Buhl HU, Weinhardt C (2009) BISE’s responsi- Lock-in- und Netzeffekte im Ecosystem Konzeption einer Kollaborationsplattform bility in service research. Business & Infor- iTunes. In: Proceedings of the 9th inter- für die zwischenbetriebliche Softwareer- mation Systems Engineering 1(6):405–407 national conference Wirtschaftsinformatik, stellung. WIRTSCHAFTSINFORMATIK 49(4): Burkart P (2008) Trends in digital music Vienna, p 9 247–256 archiving. Information Society 24(4):246– Dörr J, Benlian A, Vetter J, Hess T (2010) Pric- Hinz O, Eckert J (2010) The impact of search 250 ing of content services – an empirical inves- and recommendation systems on sales in BuxmannP,PohlG,JohnscherP,StrubeJ tigation of music as a service. In: Proceed- electronic commerce. Business & Informa- (2005) Strategien für den digitalen Musik- ings of the 16th Americas conference on tion Systems Engineering 2(2):67–77 markt – Preissetzung und Effektivität von information systems, Lima, p 9 Hu L, Bentler PM (1999) Cutoff criteria for Maßnahmen gegen Raubkopien. WIRT- Douglas DE, Cronan TP, Behel JD (2007) Eq- fit indexes in covariance structure analysis. SCHAFTSINFORMATIK 47(2):118–125 uity perceptions as a deterrent to software Structural Equation Modeling 6(1):1–55

Business & Information Systems Engineering BISE – RESEARCH PAPER

IFPI (2010) IFPI digital music report 2010 behavior. Information Systems Research Sheppard BH, Hartwick J, Warshaw P (1988) – music how, when, where you want it. 2(3):173–191 The theory of reasoned action: a meta- London Meuter ML, Bitner MJ, Ostrom AL, Brown analysis of past research with recom- IFPI (2012) IFPI digital music report 2012 – SW (2005) Choosing among alternative ser- mendations for modifications and future expanding choice. Going Global, London vice delivery modes: an investigation of research. Journal of Consumer Research Jaisingh J (2007) Piracy on file-sharing net- customer trial of self-service technologies. 15(3):325–343 works: strategies for recording companies. Journal of Marketing 69(2):61–83 Simfy (2011) Simfy – a music networx com- Journal of Organizational Computing & Nandedkar A, Midha V (2009) Optimism in pany. http://corporate.simfy.de/. Accessed Electronic Commerce 17(4):329–348 music piracy: a pilot study. In: Proceed- 2011-11-25 Jarvis CB, Mackenzie SB, Podsakoff PM (2003) ings of the 30th international conference Sinha RK, Mandel N (2008) Preventing dig- Acriticalreviewofconstructindicatorsand on information systems, Phoenix, p 10 ital music piracy: the carrot or the stick? measurement model misspecification in Oberholzer-Gee F, Strumpf K (2007) The effect Journal of Marketing 72(1):1–15 marketing and consumer research. Journal of file sharing on record sales: an empir- Sinha RK, Machado FS, Sellman C (2010) Don’t of Consumer Research 30(2):199–218 ical analysis. Journal of Political Economy think twice, it’s all right: music piracy and Jöreskog K (1993) Testing structural equation 115(1):1–42 pricing in a DRM-free environment. Journal models. In: Bollen KA, Long JS (eds) Testing Oestreicher-Singer G, Zalmanson L (2009) of Marketing 74(2):40–54 structural equation models. Sage, Newbury “Paying for content or paying for commu- Straub DW Jr. (1990) Effective IS security: Park, pp 294–316 nity?” The effect of social involvement on an empirical study. Information Systems Jöreskog KG, Sörbom D, Wallentin FY (2006) subscribing to media web sites. In: Pro- Research 1(3):255–276 Latent variable scores and observational ceedings of the 30th international confer- Sullivan M, Hibbert S (2006) The music col- residuals. http://www.ssicentral.com/lisrel/ ence on information systems, Phoenix, p lector. In: Advances in consumer research – techdocs/obsres.pdf. Accessed 2012-12-02 17 7th European conference proceedings, pp Kabanoff B (1991) Equity, equality, power and Pavlou PA, Fygenson M (2006) Understanding 285–291 conflict. Academy of Management Review and prediction electronic commerce adop- Sun T, Youn S, Wu G, Kuntaraporn M (2006) 16(2):416–441 tion: an extension of the theory of planned Online word-of-mouth (or mouse): an ex- Kwong SW, Park J (2008) Digital music ser- behavior. MIS Quarterly 30(1):115–143 ploration of its antecedents and conse- vices: consumer intention and adoption. Peitz M, Waelbroeck P (2004) The effect of In- quences. Journal of Computer-Mediated Service Industries Journal 28(10):1463– ternet piracy on CD sales – cross section ev- Communication 11(4):1104–1127 1481 idence. Review of the Economic Research Thaler RH (2008) Mental accounting and con- Kwong TCH, Lee MKO (2002) Behavioral inten- on Copyright Issues 1(2):71–79 sumer choice. Marketing Science 27(1):15– tion model for the exchange mode Internet Peitz M, Waelbroeck P (2006) Why the music 25 music piracy. In: Proceedings of the 35th industry may gain from free downloading – Trepte S, Reinecke L, Richter-Matthies A, Adel- Hawaii international conference on system the role of sampling. International Journal berger C, Fittkau J-T (2004) Von Jägern und sciences, Hawaii, p 10 of Industrial Organization 24(5):907–913 Sammlern. Motive des MP3-Sharings in Ab- Lambert DM, Harrington TC (1990) Measur- Peukert C, Claussen J (2012) Piracy and movie grenzung zum CD-Kauf. In: Hasebrink U, ing nonresponse bias in customer service revenues: evidence from . Mikos L, Pommer E (eds) Mediennutzung mail surveys. Journal of Business Logistics http://papers.ssrn.com/sol3/papers.cfm? in konvergierenden Medienumgebungen. 11(2):5–25 abstract_id=2176246. Accessed 2012-12 Reinhardt Fischer, München, pp 199–219 Lambrecht A, Skiera B (2006) Ursachen eines -01 Upshaw D, Babin LA (2010) - Flatrate-Bias – Systematisierung und Mes- Plowman S, Goode S (2009) Factors affect- ing: competing against online piracy. In- sung der Einflussfaktoren. Schmalenbachs ing the intention to download music: qual- ternational Journal of Business & Public Zeitschrift für betriebswirtschaftliche For- ity perceptions and downloading intensity. Administration 7(1):14–26 schung 58(5):588–617 Journal of Computer Information Systems LaRose R, Kim J (2007) Share, steal, or buy? Venkatesh V (2000) Determinants of per- 49(4):84–97 ceived ease of use: integrating control, in- A social cognitive perspective of music Polites GL, Karahanna E (2012) Shackled to downloading. CyberPsychology & Behavior trinsic motivation, and emotion into the the status quo: the inhibiting effects of technology acceptance model. Informa- 10(2):267–277 incumbent system habit, switching costs, Levin AM, Dato-On MC, Manolis C (2007) De- tion Systems Research 11(4):342–365 and inertia on new system acceptance. MIS Venkatesh V, Thong JYL, Xu X (2012) Con- terring illegal downloading: the effects of Quarterly 36(1):21–42 threat appeals, past behavior, subjective sumer acceptance and use of information norms, and attributions of harm. Journal of Polites GL, Roberts N, Thatcher J (2012) Con- technology: extending the unified theory Consumer Behavior 6(2–3):111–122 ceptualizing models using multidimen- of acceptance and use of technology. MIS Levin AM, Dato-On MC, Rhee K (2004) Money sional constructs: a review and guidelines Quarterly 36(1):157–178 for nothing and hits for free: the ethics for their use. European Journal of Informa- Venkatesh V, Morris MG, Davis GB, Davis of downloading music from peer-to-peer tion Systems 21(1):22–48 FD (2003) User acceptance of information web sites. Journal of Marketing Theory and Prelec D, Loewenstein G (1998) The red and technology: toward a unified view. MIS Practice 12(1):48–60 the black: mental accounting of savings Quarterly 27(3):425–478 Liebowitz SJ (2006) File sharing: creative de- and debt. Marketing Science 17(1):4–28 Wang C-C, Chen C-T, Yang S-C, Farn C-K struction or just plain destruction? Journal Ringle CM, Wende S, Will S (2005). Smart- (2009) Pirate or buy? The moderating ef- of Law & Economics 49(1):1–28 PLS 2.0 (M3) beta. Hamburg. http://www. fect of idolatry. Journal of Business Ethics Lin C-H, Sher PJ, Shih H-Y (2005) Past progress smartpls.de 90(1):81–93 and future directions in conceptualizing Ringle CM, Sarstedt M, Straub DW (2012) Wetzels M, Odekerken-Schröder G, Van Op- customer perceived value. International A critical look at the use of PLS-SEM in MIS pen C (2009) Using PLS path modeling Journal of Service Industry Management Quarterly. MIS Quarterly 36(1):iiv-8 for assessing hierarchical construct models: 16(4):318–336 Rogers EM (1995) Diffusion of innovations. guidelines and empirical illustration. MIS Lohmöller J-B (1989) Latent variable path Free Press, New York Quarterly 33(1):177–195 modeling with partial least squares. Sandulli FD, Martin-Barbero S (2007) 68 cents Woolley DJ (2010) The cynical pirate: how Springer, Heidelberg per song: a socio-economic survey on cynicsm effects music piracy. Academy of Mann F (2010) Filmdistribution über internet- the Internet. The International Journal of Information & Management Sciences Jour- basierte Abrufdienste. epubli, München Research into New Media Technologies nal 13(1):31–43 Mathieson K (1991) Predicting user inten- 13(1):63–78 Zentner A (2006) Measuring the effect of file tions: comparing the technology accep- Schumann M, Hess T (2009) Grundfragen der sharing on music purchases. Journal of Law tance model with the theory of planned Medienwirtschaft. Springer, Heidelberg and Economics 49(4):63–90

Business & Information Systems Engineering