Platform governance and engagement behaviour in the sharing economy

(1)

Platform governance and engagement behaviour in the sharing economy.

How can platforms diminish consumer misbehaviour and foster value co-creation in platform sharing?

Philip Kirch - 11440767 University of Amsterdam

Bachelor Thesis

BSc Business Administration - Management in the Digital Age track Nicole Stofberg

(2)

Statement of Originality

This document is written by Philip Kirch who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

2

Abstract

The growth of sharing platforms depends on the continued participation of its members and is influenced by the value co-creation and prosocial behaviour of these members, which is mainly outside of the direct control of the platform. To explain when consumers act pro-social and misbehave less we draw on organizational trust and power. We propose that consumers engagement behaviours are influenced by different types of governance. More specifically, we hypothesize that if a platform implements platform sociality, the peer trust among users will have a positive influence on consumer engagement behaviours. Whereas, we hypothesize that if a platform implements platform power, the enforced compliance will have a negative influence on consumer engagement behaviours. We test these hypotheses with a sample of 1175 respondents in a vignette study. The hypothesis of platform trust was supported, suggesting that platform trust is more likely to contribute to positive consumer engagement behaviours, when users express a strong sense of peer trust. Whereas, the hypothesis of platform power was not supported, yet suggesting that platform power is more likely to contribute to positive consumer engagement behaviours, when users express a strong sense of enforced compliance.

Keywords: sharing economy, carsharing, organizational principles, organizational power, sociality, peer trust, enforced compliance

(4)

Acknowledgements

First of all, I would like to express my gratitude to my family, friends and significant other for the encouragement, patience and support they have provided me with throughout this whole process.

Second, my gratitude towards my fellow peers with whom the whole design and data collection was conducted.

Finally, and most important: I am very thankful for the help, guidance and spirit Nicole Stofberg gave me during her supervision of this thesis. Without her continuous guidance, upon the last second, this would all have not been possible.

(5)

4

A new type of business in which consumers rent out or borrow their personal belongings has arisen. The “sharing economy” offers the opportunity for consumers, who connect via the internet, to rent or lend out their belongings, such as cars, homes and tools to complete strangers (Frenken & Schor, 2017). The consumer is willing to pay a price premium, in order to acquire consumption time, for the use of that object (Bardhi & Eckhardt, 2012). In the sharing economy, business and their new technologies enable strangers to connect in networks through online platforms and to share services, goods, space, money and transportation solutions (Möhlmann 2015). Sharing between strangers embodies the principles of the sharing economy (Frenken & Schor, 2017; Murillo, Buckland, & Val, 2017; Muñoz & Cohen), since it allows to create an opportunity of utilizing resources and tackle overproduction by the exploitation of underutilized assets that otherwise would have sit idle (Benjaafar, Kong, Li, & Courcoubetis, 2019). The growth of both P2P and B2C platforms is dependent on the current platform members’ continued participation and to behave towards other participants, who are usually strangers, in a prosocial manner (Stofberg et al., 2019). Sharing and interacting with strangers creates room for opportunistic behaviours, which might hinder pro-social behaviour and therefore value co-creation (Bardhi & Eckhardt, 2012). The key in all sharing context is the adaptation of this type of behaviour, and is an essential precondition to proper functioning over time in particular of P2P business models. A sharing citizenship (pro-social) behaviour is essential in both P2P and B2C business models, to ensure that other users have access to a good that is not broken, dirty or damaged. The number of undesired consumers must therefore be limited by platforms (Harris & Reynolds, 2004), to safeguard the trust of consumers in the platform and their participating peers (Bardhi & Eckhardt, 2012). An essential purpose for sharing platforms is to create governance mechanisms that deter misbehaviour and encourage cooperative (citizenship) behaviours.

This research explores these issues in the context of what is known as carsharing. Most cars, which are privately owned, sit idle 95% of the time (Belk, 2010), this makes cars one of the candidates for sharing at times when there is no need from the owners, the number of privately shared cars is small in reality. Within carsharing, consumers can choose between different forms of sharing, for instance, when sharing a car, they can share a car with a peer (peer-to-peer, P2P; Cohen & Kietzmann, 2014), or use a company’s service (business-to-consumer, B2C; Cohen & Kietzmann, 2014). What distinguishes these sharing business models from traditional ones, is that the main share of value is created by the consumer community (both providers and users), which implies that the value creation lies outside of the direct control of the business (Parker, Van Alstyne, & Jiang, 2016). This paper will focus on both P2P and B2C carsharing

(7)

6

platforms to further understand whether and how the platforms’ organizational structure influences consumer engagement behaviour and can ensure a safe experience.

(8)

2. Literature

2.1 The difference Peer-to-Peer and Business-to-Consumer platforms.

Peer-to-peer platforms have been conceptualized as a kind of ‘platform markets’ (Rietveld & Eggers, 2018) or as ‘two-sided markets’ (Rochet & Tirole, 2006), because these platforms do not take ownership of the transferred goods and are identified as intermediaries enabling interactions with a minimum of two sets of actors (users and providers). Examples of P2P platforms in the car sharing economy are Turo, Helbiz, GoMore and SocialCar. Business-to-consumer platforms, on the other hand, are for example used by established producing firms to distribute their products (Matzler, Veider, & Kathan, 2015). Producers of the automotive industry, especially, have started to share cars via B2C sharing platforms such as Zipcar or Car2Go (Provin, D., Angerer, P., & Zimmermann, S., 2016). Not only, is this platform type used by established car producers, but are also implemented by companies such as GreenWheels, MyWheels and WattCar. The dependency of both P2P and B2C sharing platforms on their users’ behaviour is essential to co-create value. The growth of both P2P and B2C platforms is dependent on the current platform members’ continued participation and to behave towards other participants, who are usually strangers, in a prosocial manner (Stofberg et al., 2019). When lacking trust in the behaviour of their peers, users ‘call for governance and regulation’, Bardhi and Eckhardt (2012) have maintained that governance is preferred by users of a B2C platform, and also necessary for users of a P2P platform to ensure a safe experience.

2.2 Consumer engagement behaviour

How behaviours of users on the platform can influence both the members they interact with, in a negative (taking advantage of others) or positive sense (aiding others) as well as their own experience, is captured by consumer engagement behaviours (CEB) (Van Doorn et al., 2010; Brodie et al., 2011). CEB are defined as: “consumers making voluntary resource contributions that have a brand or firm focus but go beyond what is fundamental to transactions, occur in interactions between the focal object and/ or other actors and result from motivational drivers” (Jaakkola & Alexander, 2014: 247). Behaviours can be both positive (create value) as well as negative (destroy value) and these benevolent behaviours are directed at the other sharing partners and not only the platform (Stofberg et al, 2019). The CEB conceptualization is more all-embracing than related concepts such as organizational citizenship behaviours or voluntary performance (Jaakkola & Alexander, 2014).

(9)

8

Negative engagement behaviours

When users encounter negative reciprocity (misbehaviour), chances are high they reproduce this misbehaviour themselves, particularly if relation bonds are lacking (Schaefers, Wittkowski, Benoit, & Ferraro, 2016). Misbehaviour, i.e. “the act of deliberately disobeying commonly accepted codes of conduct in a consumption situation by inappropriate handling, damage or overuse of the product” (Schaefers et al., 2016, p. 3), negatively impacts the access to it by other users, causing ‘general non-availability’ or ‘malfunctions’ and is therefore particularly harmful within the sharing economy (Schaefers et al., 2016). This misbehaviour can lead to negative emotions, experiences and can be seen as value destroying and is therefore particularly harmful within the sharing economy (Schaefers et al., 2016). Example of these acts of misbehaviour include; owners providing a car missing a floorboard, seeing straight to the asphalt (Burke, 2020), “people say they filled up the car with gas when they didn’t” (Quora, 2017) and “When returning my car, a renter parked in front of a fire hydrant.” (Quora, 2017). The number of such undesired consumers must therefore be limited by platforms (Harris & Reynolds, 2004) to safeguard the trust of consumers in the platform and their participating peers (Bardhi & Eckhardt, 2012).

Positive engagement behaviours

The opposite of misbehaviour is a ‘sharing citizenship behaviour’, in a sharing economy context. In particular, sharing citizenship behaviour can be defined as a behaviour that is ‘affiliative, cooperative and directed at other participants’ (Mossholder, Richardson, & Settoon, 2011, p. 13) in the sharing economy. The key in all sharing context is the adaptation of this type of behaviour, and is an essential precondition to proper functioning over time in particular P2P business models. A sharing citizenship behaviour is essential in both P2P and B2C business models, to ensure that other users have access to a good that is not broken, dirty or damaged. Leaving a car parked in an inaccessible location or leaving a car in bad conditions, are examples of behavioural actions that hampers another users’ access to it. Particularly in P2P contexts a sharing citizenship behaviour is critical: after one’s intimate possessions have been broken or damaged, the willingness to share these again is significantly lower. According to Belk (2010) users feel a ‘de facto sense of joint possession’ when sharing platforms are characterised by social bonds, in which they feel responsible to not damage or overuse the product, take advantage of it and leave it in a similar state as they found it.

Similar, Stofberg et al. (2019) demonstrate users will only engage in citizenship behaviours when they frame their relationship with other participants as belonging to the same community (generalized reciprocity) or view them as being equal partners (balanced reciprocity). A direct connection between citizenship behaviour and the relationship users develop with the sharing platform has been established by (Lee, Yang, & Koo, 2019). Sharing citizenship behaviour will aid value co-creation since it will provide

(10)

additional acts beyond expectation and ensures that users stick to the rules and not cause trouble for other members. Which will result in positive emotions such as; gratitude, happiness and joy (Stofberg et al., 2019; Lee et al., 2019). There are adequate examples of sharing citizenship behaviour in the P2P carsharing contexts. For example, it is frequent among Turo car sharers, to go out of their way to make their users comfortable through bringing the car to the renter, provide safety instructions specific for the car or disinfect the car before usage. When users of Turo pick up a car, the owner will walk around the vehicle with the user to check for any previous scratches and other damages, level of gas and check the users’ driver licence. Users are expected to treat the owners’ car as if it were their friends’ car (Burke, 2020).

2.3 Organizational Principles

An essential purpose for sharing platforms is to create governance mechanisms that deter misbehaviour and encourage cooperative (citizenship) behaviours. Research in sociology has investigated how organizational principles (a measure of solving the problem of uncertainty and interdependence) (McEvily, Perrone, & Zaheer, 2003, p. 102) help foster these behaviours. Uncertainty and interdependence are factors which diminishes trust between users and between platform and users (Ouchi, 1980), which in turn lowers the willingness to participate (Hartl, Hofmann, & Kirchler, 2016) and makes goal obtainment for platforms more difficult, without creating the necessity for organizational solutions (Ouchi, 1980). The expectation of other participants refraining from misbehaving, increases participants’ willingness to continue participating on a sharing platform (Stofberg et al., 2019). In other words, without the trust in other participants’ behaviour, users are less willing to continue. Without being able to fully monitor or having complete control over other’s behaviours, sharing partners must exchange information and rely on others to accomplish these organizational goals.

Sharing partners cannot assume that their motives, competencies, goals and interests are perfectly aligned, which further complicates coordinating actions (Ouchi 1980). Examples of organization principles include hierarchy, market and clan (Ouchi 1980). These organizational principles have been referred by others as authority (power), price, and norms (trust) (Adler 2001, Bradach and Eccles 1989; Powell 1990). Below we will discuss how particularly power and trust are important in sharing.

2.3.1 Organizational Power

The example of an organizational principle which solves the problem of coordinating action regarding uncertainty and interdependence by reallocating decision-making rights is authority (Simon 1957; Coleman 1990). A platform could force their users to act according to the platform’s will, utilizing its superior attributes, i.e. the platform could exert power (Bachmann, 2001). To understand how relationships of a sharing platform are coordinated, power is a fundamental principle of coordinating (Hurni

(11)

10

& Huber, 2014). Examples of power in the case of carsharing include fines, regulations and reporting mechanisms, Hartl et al. (2016) found that consumers support power structures because they regard humans as egoistic and therefore in need of regulation. Without the enforcement mechanisms, consumers might think that the system will break down and therefore see governance as a necessary device (Bardhi & Eckhardt, 2012). When cooperation would be highly unlikely without sanctions, benefits of a power structure may emerge under these circumstances (Jiang, Perc, & Szolnoki, 2013). To date few studies have explicitly investigated how wielding power can help regulate misbehaviour in sharing, by setting up command and control type sanctions, with the exception of Hofmann, Hartl, & Penz (2017). However, this paper asserts, that sharing cannot be under full control of the platform in the carsharing context. Even in a B2C context, i.e. by being responsible for the cleanliness of the vehicle, users take over service quality roles, previously executed by employees. These actions are not under the direct supervision of the platform and power may not be the most effective mechanism of governance to ensure consumer engagement (Eckhardt et al., 2019). Next, we discuss trust as another organizational principle, which may be more suited in situations of uncertainty.

2.3.2 Organizational Trust

McEvily et al. (2003) propose another organizational principle, trust. Conceptualizing trust as an organizational principle and integrating the research on trust and generalizable implications provides a powerful mean for researching how it affects organizing. Trust can be viewed closest related to the clan organizing principle which relies on trust, by definition of Ouchi (1980), and the organizational principles of norms, by definition of (Powell 1990). Organizational clan is defined by Ouchi (1980, p. 136) as “the solidarity which contemplates the union of objectives between individuals which stems from their necessary dependence upon one another.” Not only in the organizational contexts of clans, but also outside clans, organizational trust can and does occur. When there exists performance ambiguity or behaviours that cannot be controlled or observed, organizational trust is important (Dirks, 2000).

Trust is based on the expectancy of another party’s integrity, competency and benevolence (Mayer, Davis, & Schoorman, 1995) and captures the amount of confidence a user has in their sharing partner to not take advantage of their vulnerability. It allows trustors to “take a leap of faith” that the transacting party will do right by them, since trust is a positive behavioural expectation (Hurni & Huber, 2014). Otherwise put by McEvily et al. (2003): trust as an organizational principle implies an expectation on the side of the trustor that the party’s intentions and motivations are at least benign or benevolent, which diminishes their natural impulse to protect themselves under conditions of uncertainty from opportunistic behaviours.

The degree of sociality on the platform is critical for the ability of platforms to coordinate sharing transactions through trust. Perren and Kozinets (2018, p. 23) defined platform sociality as “the physical

(12)

and/ or virtual copresence of social actors in a network, which provides opportunity for social interaction between them.” The possibilities of negotiation and communication for the combination of virtual and physical copresence, allow users to work out a successful exchange and establish trust (Perren & Kozinets, 2018). Building on the insights Perren and Kozinets (2018), Stofberg and Bridoux (2019) empirically demonstrate in a discrete choice experiment that sharing becomes a private experience that has nothing to do with sociality when sharing is facilitated by remote technologies and in-person interactions with other consumers. In turn, this sends out cues that participation is driven by self-interest and norms of negative reciprocity. Contrary, cooperation in the form of tit for tat and generalized reciprocity is an expected behavioural outcome when sociality is an integral part of the sharing transaction. The work by Stofberg et al. (2019) empirically substantiates these findings amongst users and demonstrates these perceptions to drive citizenship behaviours.

2.4 How platform sociality influences altruism and expected negative normative behaviour

In other words, sociality refers to tendencies to build cooperative and associative relationships between users (Wittel, 2001). Platforms with low sociality is embedded in the anonymity of its users, while platform sociality allows for free-flowing interaction between their users, resulting in social benefits and more trust (Perren and Kozinets, 2018). The organizational principle of clan (Ouchi, 1980), which entails solidarity and union of objectives between individuals, implies the trust and social benefits between users on a platform. These social benefits and trust positively influence citizenship behaviour, since reciprocal altruism rests on the expectation of future material benefits from cooperation and is self-interested and future-oriented. When such feelings of sociality are facilitated, Möhlmann (2015) states that the probability of repeat sharing increases. As described before, the CEB of altruism aids value co-creation, since it provides additional acts beyond expectation and ensures that users do not cause trouble for other members and stick to the rules, which will result in positive emotions such as; joy, gratitude and happiness (Stofberg et al., 2019; Lee et al., 2019). When a platform trusts its users, this organizational trust represents a positive assumption about the intentions and motives of the other parties (Mayer et al., 1995; Rousseau, Sitkin, Burt, & Camerer, 1998). Therefore, this paper hypothesises that the perceived social benefits of platform sociality makes users behave more altruistic. Thus:

H1a. Sociality is associated with higher levels of altruism.

Fear of misbehaviour is one of the largest obstacles owners express when it comes to the actual sharing of their cars. A fundamental difference from a traditional exchange is due to the two-sided nature of the ‘car sharing’ market (Belk, 2010) and is caused by the feeling of vulnerability of being taken

(13)

12

advantage of by the other party (Schor & Fitzmaurice, 2015). By placing trust in a social platform, one signals that another is expected to behave in a trustworthy manner, which can become a self- fulfilling prophecy. Features of social platforms such as rating systems and visibility of information aim to reduce uncertainty of other’s opportunism by providing information about their previous actions or behaviour (Dreyer, Lüdeke-Freund, Hamann, & Faccer, 2017; Frenken & Schor, 2017; Kuwabara, 2015). Public display and visibility that comments and peer ratings give, to the users’ impact on the platform and their actions, lead the user to a sense of responsibility and need for social approval towards the counterparts and their good (Bénabou & Tirole, 2006). Therefore, this paper hypothesises that the perceived social benefits of platform sociality makes users expect less misbehaviour. Thus:

H1.b Sociality is associated with lower levels of expected negative norms.

The mediating influence of peer trust.

Social interaction fosters a feeling of empathy and reduces social distance between the subjects (Mohlin & Johannesson, 2008) and additionally creates a sense of ‘social responsibility’ (Andreoni & Petrie, 2004). Trust of consumers in other users, is essential for platforms of collaborative services or goods (Bhattacherjee, 2002). Platform sociality aims to reduce uncertainty of others’ opportunism by providing information about actions or behaviour (Dreyer et al., 2017; Frenken & Schor, 2017; Kuwabara, 2015). Public display and visibility which platform sociality gives to the users’ impact on the owner and their actions, lead the user to a sense of responsibility and need for social approval towards the counterparts and their good (Bénabou & Tirole, 2006). Prosocial behaviour has been argued to create positive emotions, such as happiness, bonding and feelings of gratitude that motivate providers and users alike to remain committed over time and seek out additional sharing experiences (Bucher, Fieseler, & Lutz, 2016; Ikkala & Lampinen, 2015).

A low level of anonymity of the users reduce the amount of misbehaviour among users (Schaefers et al., 2016), combined with social bonds (Belk, 2010) users perceive it is also possible to perceive a sense of ‘communal identification’ or ownership (Schaefers et al., 2016). In a social dilemma situation without an authority, sociality structures are welcomed by community members, this implementation of a social platform successfully increases the level of cooperation (Hartl et al., 2016). The positive outcome that is rooted in the users’ ‘a leap of faith’ to other users, expecting its sharing partner will do right by them (Hurni & Huber, 2014), together with trust as an organizational principle (McEvily et al., 2003), diminishes the users’ natural impulse to protect themselves under conditions of uncertainty from opportunistic behaviours.

Accordingly, this paper expects the relationship between sociality and citizenship, and sociality and misbehaviour to be mediated by trust. Therefore, this paper argues that platform sociality leads to higher

(14)

levels of peer trust, which positively impacts citizenship behaviour and negatively impacts misbehaviour. Thus:

H2 The relationship between sociality and altruism is mediated by peer trust, such that the presence of sociality is associated with higher levels of peer trust, which leads to higher levels of altruism.

H3. The relationship between sociality and expected negative norms is mediated by peer trust, such that the presence of sociality is associated with higher levels of peer trust, which leads to lower expected negative norms.

(15)

14

2.5 How organizational platform power influences altruism and expected negative normative behaviour.

Hartl et al. (2016) found that consumers support authority and power structures because they regard humans as egoistic and therefore in need of regulation. Without the enforcement mechanisms, consumers might think that the system will break down and therefore see governance as a necessary device (Bardhi & Eckhardt, 2012). In order to increase the return rate of goods used in common, in the sharing platform, supporters of a power structures in their opinion underline that governance is needed (Hartl et al., 2016). Cooperation increases if a sanction system is perceived as fair (Verboon & van Dijke, 2011), whereas when these sanctions are not applied in a fair way can destroy altruistic cooperation almost completely (Fehr & Rockenbach, 2003). When relationships among participants are perceived as governed more strongly by communal sharing and equality matching, it promotes sharing citizenship behaviour amongst both user and provider in the form of altruism and conscientiousness, as well as it increases the willingness to keep participating (Stofberg et al., 2019). By exerting power, the platform could force their users to act according to the platform’s will (Bachmann, 2001), reducing the vulnerability of users and their uncertainty. When platforms are preoccupied with ensuring value capture, thus implementing a great set of rules and regulations, this might unnecessarily reduce the amount of total value for all actors (Fehrer, Woratschek, & Brodie, 2018). It becomes a self-fulfilling prophecy if platforms wield too much power (i.e. engage in excessive safeguarding and monitoring), they run the risk of signalling suspicion and a lack of faith in their user base, which triggers distrust (McEvily et al, 2003; Eckhardt et al., 2019). Therefore, this paper argues that organizational power makes users behave less altruistic and makes users expect more negative normative behaviour. Thus:

H4.a Power is associated with lower levels of altruism

H4.b Power is associated with higher levels of expected negative norms

The mediating influence of enforced compliance.

Empirical evidence generally supports the relevance of power and authority as determinants of tax compliance (Kogler et al., 2013; Wahl, Kastlunger, & Kirchler, 2010; Muehlbacher, Kirchler, & Schwarzenberger, 2011). Such that from the power of tax authorities, results in enforced compliance. Under circumstances of low authority and power, compliance will be at a minimum (Gangl, Hofmann, & Kirchler, 2015). One way to encourage cooperation is to sanction defection and to monitor activity (Kollock, 1998). Gangl et al. (2015) found that power structures favour enforced compliance and an antagonistic climate.

(16)

It can be argued that relying solely on trust in a confidence climate is far too optimistic in a social dilemma context, since there always will be citizens tempted to engage in egoistic profit maximizing activities and hence trusting authorities are not perceived as being able to enforce compliance. Yet, egoistic free-riders whom are not threatened by power and authority, can easily change the confidence climate, of the tax authorities’ interaction with taxpayers, into an antagonistic or service climate if they are perceived as hostile prosecution or if they trigger rational consideration of authorities’ intentions (Gangl et al., 2015). Users that encounter misbehaviour are more likely to mimic this behaviour themselves (Schaefers et al., 2016). Distrust in other users can decrease the trustworthiness of a platform environment and can become a self-fulfilling prophecy (Goshal & Moran, 1996).

Consumers may exploit the service or good, if they sense others will do the same, especially in a community where little to no regulating power exists (Jorgensen, Graymore, & O’Toole, 2009). Consumers support platform power and authority (Hartl et al., 2016), because without they might think that the system will break down and therefore see it as a necessary device (Bardhi & Eckhardt, 2012). In order to increase the return rate of goods used in common, in the sharing environment, supporters of a power structures in their opinion underline that governance is needed (Hartl et al., 2016). Vulnerability of users and their uncertainty can be reduced when platforms exert their power (Bachmann, 2001) and enforce compliance (Gangl et al., 2015). This paper proposes that enforced compliance, is not only influenced by the authority and power in the interaction between the tax authorities’ and taxpayers, but might also be applicable within the sharing economy. Thus:

H5. The relationship between power and expected negative norms is mediated by enforced compliance, such that the presence of power is associated with higher levels of enforced compliance, which leads to higher expected negative norms.

H6. The relationship between power and altruism is mediated by enforced compliance, such that the presence of power is associated with higher levels of enforced compliance, which leads to lower levels of altruism.

(17)

16

3. Conceptual framework and hypotheses

This study looks into the effects of power and sociality on consumers’ altruism and expected negative norms in the car sharing industry, independently of each other. The presence of power should lead to a feeling of enforced compliance, resulting in higher expected negative norms and lower levels of altruism. Whereas, the presence of sociality should lead to a feeling of peer trust, resulting in higher levels of altruism and lower expected negative norms. Table 1 presents an overview of the hypotheses. Figure 1 illustrates the conceptual framework with the corresponding hypotheses.

(18)

(19)

18

Table 1.

Overview of the hypotheses H1.a Sociality is associated with higher levels of altruism.

H1.b Sociality is associated with lower levels of expected negative norms.

H2. The relationship between sociality and altruism is mediated by peer trust, such that the

presence of sociality is associated with higher levels of peer trust, which leads to higher levels of altruism.

H3. The relationship between sociality and expected negative norms is mediated by peer trust, such that the presence of sociality is associated with higher levels of peer trust, which leads to lower expected negative norms.

H4.a Power is associated with lower levels of altruism

H4.b Power is associated with higher levels of expected negative norms

H5. The relationship between power and expected negative norms is mediated by enforced compliance, such that the presence of power is associated with higher levels of enforced compliance, which leads to higher expected negative norms.

H6. The relationship between power and altruism is mediated by enforced compliance, such that the presence of power is associated with higher levels of enforced compliance, which leads to lower levels of altruism.

(20)

4. Methodology

In this chapter first the design of the study will be presented. This study will use two survey instruments: a vignette study and survey questionnaire. The survey questionnaire measures individual levels of peer trust, enforced compliance and control variables (i.e. age), while the vignette study measures the influence of power and sociality on expected negative norms and altruism. Second, the design of the vignettes will be presented. Third, a pre-test is performed prior to research to check the credibility of the hypothetical situation and to ensure the vignettes communicate what they intend to communicate. Fourth, the survey instrument is explained. Fifth, the variable measures are presented. Sixth, a vignettes description of data collection and the sample is given. Followed, the method with which the vignette data is analysed, summarized in this chapter: by means of multiple regression. This thesis is part of a bigger study performed by 11 bachelor students of the University of Amsterdam.

4.1 Research design

The research is conducted through a vignette experiment as well as a survey questionnaire for data collection. A vignette experiment is utilized, to test the hypotheses and this consists of “a set of systematically varied descriptions of subjects, objects, or situations in order to elicit respondents’ beliefs, attitudes, or intended behaviours with respect to the presented vignettes” (Steiner, Atzmüller & Su, 2017). This technique allows researchers to manipulate the factors that are theorized to influence decision making, causality between an independent variable and a behavioural outcome can be established (Porter, 2001; Dülmer, 2007). In this study, this paper focus specifically on the influence of different social and authority punishment techniques deployed by sharing platforms (IV’s) on expected negative normative behaviour of a sharing community, as well as intentions to act altruistically on a platform (DV’s). The results were contrasted with a neutral vignette in which no punishment information was provided to measure the response of both punishment features on the paper’s DV’s.

A description of similar yet non-identical situations, will be presented to the respondents, which consists of varied independent variables. To test the hypotheses, a vignette study enables the ability to manipulate different independent variables in different circumstances and thus, enables to create survey designs that closely resemble real life issues and situations. A high internal validity is ensured by utilizing an experimental vignette study, as the corresponding questions to the vignettes are embedded in a realistic context (Steiner et al., 2017). Furthermore, this research design combined a survey method, with the advantages of the depth of a qualitive method, which has proven to be a successful method related to the sharing economy testing theories (Stofberg & Bridoux, 2019). Respondents were randomly assigned a

(21)

20

social condition and are presented a hypothetical situation in a vignette, that relates to the car sharing platform. To obtain demographic and social background information of the respondents, a traditional survey questionnaire is utilized, in addition to the vignette experiment.

4.2 Vignette design

The emergence of a new car sharing platform named SplitCar is presented in all vignettes. An essential to a good vignette study is that the vignettes are perceived to be both realistic and credible (Bridoux & Stoelhorst, 2016). To establish credibility the vignettes were built on actual punishment measures sharing platforms implement to evoke altruistic behaviour. For example, power features such as the provisions of fine, user blacklists and user expulsion were inspired from the practices of Turo. Social punishment mechanisms, such as controlling behaviour through social control and reviews were similarly based on real life practices of Snapcarr and Turo. Members of these platforms rate one another, interact and provide feedback to the platform in order to allow the platforms to control behaviour. In the B2C context social punishment mechanisms were slightly adjusted to account for the fact that users did not interact with owners of the car but rather with an employee or liaison of the company, who are voluntarily responsible for cars owned by the platform, following the business model of the sharing platform MyWheels.

During the research of this paper, the corona virus (covid-19) spread from Asia to over 188 countries worldwide (BBC, 2020b). It has drastically changed how freely consumers can move and influences the psychological factors of when consumers interact with the outside world. The corona virus is considered the biggest threat to humanity since the second world war. A well-known threat is the fear of contagion, which has academically been studied during earlier epidemics such as MERS and found to significantly alter the behaviour of consumers (Jung, Park, Hong, & Hyun). It is uncertain how this new virus will impact the sharing economy on the long term and a question that comes to minds is: will it survive? To deal with this changing reality and account for potential negative (or positive) intentions during this pandemic as well as after a vaccine is formulated, the Corona variable will present the following two scenarios; a future with a neutralized threat as well as a present day threat of the virus.

Similar to the study by Stofberg and Bridoux (2019) participants read a short introductory story that introduced the sharing concept, as well as addressed whether or not the threat of the corona virus was neutralized. This is done to provide sufficient background information as well as ensure respondents could realistically imagine themselves as prospective car sharing users. Moreover, in the ‘current corona threat’ context, the vignette outlined suggested measures actual platforms currently take to keep users safe: such as keeping a safe distance, the use of disinfectants and cleaning the car after use. In order to identify whether

(22)

the manipulation of the control variables Business model and Corona alongside Power and Sociality leads to a different behavioural response in terms of expected negative norms and altruism.

The vignette population is being shaped by the interaction ‘of all possible combinations of each dimension’ with the behavioural intention being researched. In this study the vignette population consists 2 for power, 2 for sociality, 2 for mediation, 2 for business model and 2 for the corona virus threat, by their corresponding dimensions (Jasso & Opp, 1997). This results in a vignette population of 2x2x2x2x2 is 32. (see table 2 for the dimensions and their levels in a list). The total amount of vignettes in this research is eighteen (see Appendix A), due to the following reasons; On a P2P platform the following scenarios were dropped since they were deemed unrealistic; high sociality, low market mediation, high power, corona and neutralized threat (2); low sociality, low mediation, high power, corona and neutralized threat (2); low sociality, low mediation, low power, corona and neutralized threat (2) as it infers that the platform is unwilling to connect members and users. On a B2C platform all the vignettes containing low mediation were deemed unrealistic, in both corona and neutralized threat scenarios (8) as it diminishes the ‘business’ role of this model. Thus, these designs are omitted and leave eighteen final vignette designs.

Table 2. Overview of the dimensions and their levels

Dimensions Levels

Independent variables

Power High Power

Low Power

Sociality High Sociality

Low Sociality

Mediation High Mediation

Low Mediation Control variables

Business Model Business to Consumer

Peer-to-Peer

Corona Threat of Corona

(23)

22

Vignette Population

First, the dimension Power is measured in two levels; high (e.g. SplitCar may fine you, the platform keeps a blacklist of users that damaged the vehicle) and low (e.g. normative behaviour is stipulated in the user agreement, the platform makes users responsible to report damage etc.). Second, the dimension Sociality is measured in two levels; high (e.g. you meet the owner of the car/ambassador of SplitCar in person.) and low (e.g. you never have to contact the owner/representatives from SplitCar directly). Third, the dimension

Business Model is measured in two levels; Peer to peer (e.g. car that is owned by another user) and Business

to consumer (e.g. car that is owned by SplitCar). Last, the dimension Corona is measured in two levels; threat of corona (e.g. current situation in society) and neutralized threat (e.g. there is a vaccine and a cure).

(24)

Table 3. Operationalization of the dimensions and their levels in the vignettes

Variables Description in Vignette

Power

High Power SplitCar is a sharing platform, so you don’t own the car you drive, but you have to treat it like it is your own car. If you don’t, other users may report you and SplitCar may fine you.

The platform keeps a blacklist of users that have not kept the vehicle clean, damaged it, returned it late, or with an empty fuel tank.

Low Power SplitCar is a sharing platform, so you don’t own the car you drive, but you have to treat it like it is your own car. This is stipulated in the user agreement. The platform makes users responsible to keep the vehicle clean, undamaged, returned it on time, and with a topped up fuel tank.

Sociality

High Sociality The first time you rent a car, you meet the owner of the car/ambassador of SplitCar in person. You'll probably exchange some friendly small talk and more importantly, information about the vehicle. Apart from renting a car, this also makes it an easy way to meet new people.

However, even if you only have contact online, SplitCar is still a social place. All SplitCar users have a profile with a bio and a picture, and you can connect with them, become friends, and chat. Moreover, everyone on the platform is encouraged to leave ratings and write reviews about their experiences. Low Sociality Whenever you rent a vehicle, you can open the car ‘keyless’ with your

smartphone, which will also show you information about the vehicle. You never have to contact the owner/representatives from SplitCar directly or meet them in person to receive the keys.

Business Model

Peer to Peer The platform is called SplitCar and it's a platform on which people can access and use a car that is owned by another user on SplitCar.

Business to Consumer The platform is called SplitCar and it's a platform on which people can access and use a car that is owned by SplitCar.

Corona

Threat of Corona Due to the Corona crisis, you try to stay home as much as possible, however, sometimes you have to travel. Fortunately, this is still allowed as long as you keep to social distancing regulations.

Neutralized Threat The year is 2021, the Corona crisis is over: there is a vaccine and a cure, so even on the rare occasion someone gets infected, symptoms are very mild and easy to treat. There is no more need for social distancing and there are no more

(25)

24

Pre-Test

An exploratory pre-test is conducted, in order to check whether the credibility and manipulations of the vignettes are perceived as intended. Through a randomized process the pre-test was distributed to 51 respondents. As a result from the pre-test, it was indicated that no components required improvements apart from minor layout details. The scenarios where then send to people in the students’ social circles, the respondents were asked a set of questions about the vignettes to investigate the clarity of the vignettes, one of these questions is “How is the contact with the owner (face-to-face/online)?” (see Appendix G). After this, the vignettes were altered mostly in layout and clarity of the communicated information.

4.3 Survey Instrument

A factorial online survey will be used to collect data, in order to test the hypotheses (Aguinis & Bradley, 2014). The ‘qualities of multivariate designs with survey techniques’ (Porter, 2001, p. 382) is combined with this technique. Respondents were asked to judge several similar but not identical situations in a factorial closed-ended survey. A short description is provided of a social situation or a person which contain specific references to what are thought to be important factors of the respondents’ processes of judgement making or decision making (Weibel, Rost, & Osterloh, 2007). The independent variables are varied within the different vignettes and the ‘target variables’ (Weibel et al., 2007), expected negative norms and altruism are asked. The factors that influence behavioural intentions are uncovered, because features in the different vignettes are systematically manipulated (Porter, 2001, Dülmer, 2007). The base language of the online survey is English.

4.4 Variables Measures

Applied in this research are the measures of pre-existing constructs which are adapted to the context of the sharing economy. Utilizing these existing constructs enables the internal validity and reliability of this research. The constructs for each variable are described and examples are provided.

Enforced compliance

The first mediating variable of this research is enforced compliance. This variable is measured based on Gangl et al. (2015). To measure enforced compliance, a 4-item construct is adopted with one of the items being: “When I make use of the car I see myself following the rules because I expect the fines for breaking the rules to be very high on SplitCar”. This construct is measures on a 7-point Likert scale (1=Strongly disagree; 7=Strongly agree).

(26)

Peer trust

The second mediating variable of this research is peer trust. This variable is measured based on Pavlou & Gefen (2004). To measure peer trust, a 4-item construct is adopted with one of the items being “In general, I expect the users on SplitCar to be reliable”. This construct is measured on a 7-point Likert scale (1=Strongly disagree; 7=Strongly agree).

Expected negative norms

In order to test the first dependent variable which is expected negative norms, the scale based on Schaefers et al. (2016) is used an 8-item scale with items such as “Other users will not notify SplitCar about a scratch they made”. This item is measured on a 7-point Likert scale (1=Strongly disagree; 7=Strongly agree).

Altruism

In order to test the second dependent variable which is altruism, the scale based on Stofberg et al. (2019) is used a 7-item scale with items such as “When participating I see myself willing to help other users, with any additional questions they might have, during my own time”. This item is measured on a 7-point Likert scale (1=Strongly disagree; 7=Strongly agree).

Control Variables

Research found evidence for factors influencing the likelihood of participation in the sharing economy, therefore the results are controlled for gender, age, education level, income, familiarity. Research found that the probability of millennials in an urban area participating in sharing, is for instance higher compared to older generations (Kats, 2017). Furthermore, Andreotti et al. (2017) indicate that studies have shown that higher educated users, high income and young users are more likely to participate in the sharing economy. Additionally, the results in this research are controlled for business model, covid-19 pandemic (also known as corona) and its risks as these variables are expected to influence the behaviour of participants in the sharing economy.

Realism and credibility

The realism and credibility of the vignettes measures to what extent the scenario is less abstract, realistic, and concrete. These two variables are measured with 2-questions on a 7-point Likert scale (1=Strongly disagree; 7=Strongly agree). The questions are as follows; “I had no problem imagining myself in the mentioned situation” and “I found the situation in the mentioned scenario realistic”.

(27)

26

4.5 Data Collection Procedure

The data collection was conducted within social circles of the students. The respondents were send a link to the survey questionnaire, via WhatsApp, Facebook, Instagram, LinkedIn and email, and were asked to participate. This link automatically directed them to the questionnaire. The vignette’s shown to the respondent were equally and randomly distributed. The order of items within the questionnaire was shuffled to prevent order effects (Jasso & Rossi, 1977). A total of 1671 respondents participated in -and completed- the questionnaire.

4.6 Sample

This sample with a total of 1175 respondents is recruited and used. Of these 1175 respondents 59% identified as woman, 41% as male and less than 1% as other, which means women participated slightly more in this study. A wide variety of age ranging from 18 to 80 (mean = 30, SD = 11,41), the majority is younger than 28 years old (66%). Moreover, 84.4% of the respondents lives in an urban area. The majority of the respondents is from the Netherlands (55.2%), another country is Italy (13.2%) and in total 88.9% is from Europe. The sample is highly educated with 92% of the respondents who have at least an applied university degree (hbo), of which the majority holds a bachelor or masters university degree, respectively 49.8% and 23%. A small number filled in high school (4.8%) or mbo (3.2%) as their highest education. Compared to the high education number, the income numbers are rather low. The biggest group of 52.5% has a gross income of less than € 1,500 per month. An explanation for this could be the fact that many students participated, who do not have a steady income. 21.5% earns between € 1,501 and € 3,500 per month. An eight of the sample (12.2%) earns more than € 3,500, of which 5,6% earns above € 5,500 per month.

(28)

Table 5. Description of Respondents on the Study

Variable Level N %

Gender Female 687 58.5

Male 484 41.2

Other 4 0.3

Area of Residence Rural 183 15.6

Urban 992 84.4

Education Low (Highschool / MBO) 94 8.0

Professional (HBO) 174 14.8

Academic (WO) 874 74.4

Others 33 2.8

Income Low (below €1500) 617 52.5

Medium (€1500 to €3500) 253 21.5

High (above €3500) 143 12.2

Prefer not to say 162 13.8

Range M SD

Age 18 to 80 30.02 11.4

4.7 Method of Analysis

To test for a mediated effect the SPSS extension PROCESS model 4 is used, which enables us to isolate the indirect effect for both mediators (Hayes, Preacher, & Meyers, 2011). Respondents who completed the survey ‘too fast’ would affect the correlation between two variables and would affect the quality of the responses. In addition to this, two test were conducted to detect computer-generated and fraudulent responses; Person-Total Correlation (PTC) and Mahalanobis Distance (MD) (Dupuis et al., 2020). Based on these tests, out of the 1671 responses 496 were deleted (duration<10, PTC<0 and MD p-value<.001), leaving 1175 respondents for the final sample. Dummy variables were created for the categorical variables, utilizing the dataset (see appendix C). The variables with multiple items were checked for its reliability through Cronbach’s alpha (α>.7) without possibility to improve for the following variables; enforced

compliance (α=.828), peer trust (α=.902) and expected negative norms (α=.903). One of the variables, altruism had possibility to improve and was split up in two separate variables; altruism (α=.788) and altruism corona (α=.742) (see appendix D). Afterwards, on the dataset multiple analyses were conducted

such as a manipulation check to ensure realism and credibility of the vignettes, identifying the distribution of the variables, computing correlations and running the SPSS extension PROCESS. In the next chapter the results of these analyses are presented.

(29)

28

5. Results

In this chapter the results of the data analyses are viewed. First of all, the distributions of the variables are outlined. Second, the manipulation check is shown. Third, the correlation matrix analysed. And fourth, the analyses of direct effects and mediations are outlined accordingly with the outcomes of the hypotheses.

5.1 Distribution

Both mediators enforced compliance and peer-to-peer trust and the dependent variables altruism and power are tested on skewness and kurtosis (see appendix E). None of these values have skewness or kurtosis problems. Further analysis shows that the results for these variables are not normally distributed: both the Shapiro- Wilk and the Kolmogorov-Smirnov tests have significant p- values (<0.05). However, the violation of normality is not expected to cause extensive problems with large sample sizes, >30 or 40 (Pallant, 2007).

5.2 Manipulation check

Firstly, it is checked whether the vignettes were perceived as credible and real, checking whether the vignettes communicate the correct underlying theories. Realism has a p- value > 0.05, meaning there is no statistically significant difference between the means per vignette. Credibility has a p- value < 0.05, meaning there is a statistically significant difference between the means per vignette, yet all vignettes have a credibility mean higher than 4. After conducting a post hoc test, it was found that only vignette number 8 showed a statistically significant lower mean (4.35), compared to the other vignettes (see appendix F).

Second, it is checked whether the vignettes communicate the correct underlying manipulations, presented in table 6. Power has a negative correlation with the manipulation check of consequences (r=-.120, p<.000).

Sociality has a positive correlation with the manipulation check of social (r= .244, p<.000), and negative

correlations with anonymity (r=-.509, p<.000). Market mediation has positive correlations with the manipulation check of anonymity (r=.169, p<.000), convenience (r=.122, p<.000) and assurances (r=.311, p<.000) which means these vignettes triggered the correct manipulation of market mediation.

(30)

5.3 Correlations

Table 7 illustrates there is no direct correlation between power and expected negative norms (r=.012, p=.689), another correlation hypothesized based on the theory is between power and altruism (r=-.024, p=.409), these correlations are statistically insignificant. This does not match the theoretical framework. However, as predicted by the literature, power is significantly correlated with enforced compliance (r=.067, p<0.05). This indicates that power is associated with higher levels of enforced compliance. The correlations between enforced compliance and expected negative norms (r=.031, p=.296) and altruism (r=.215, p<.000), and indicates enforced compliance to be insignificantly correlated with expected negative norms but to be significantly correlated with altruism.

The table also illustrates there is no direct correlation between sociality and peer trust (r=.044, p=.134). This does not match the theoretical framework. Additionally, the direct correlation between

sociality and expected negative norms (r=-.100, p<.01). This indicates that sociality is associated with lower

levels of expected negative norms. Another correlation hypothesized based on theory is between sociality and altruism (r=.091, p<0.01), which indicated that sociality is associated with higher levels of altruism. The correlations between peer trust and altruism (r=.429, p<.000) and between peer trust and expected

(31)

(32)

5.4 Control variables

The independent variables, dependent variables and mediators have some correlations with the control variables and demographics. First, corona is positively correlated with expected negative norms (r= -.064, p<.05), which means that in a corona situation respondents have less expected negative norms compared to a no corona situation. Second, risk to self is positively correlated with altruism (r=.123, p<.000), enforced

compliance (r=.068, p<.05) and corona (r=.160, p<.000). Risk to self is negatively correlated with peer trust

(r=-.090, p<.01). Third, risk to family is positively correlated with altruism (r=.114, p<.000), peer trust (r=.075, p<.05), enforced compliance (r=.111, p<.000) and risk to self (r=.290, p<.000). Fourth, familiarity is positively correlated with altruism (r=.099, p<.01) and peer trust (r=.083, p<.01). Fifth, female is negatively correlated with expected negative norms (r=-.099, p<.01), which means females have less expected negative norms compared to males and other genders. Additionally, female is positively correlated with peer trust (r=.117, p<.000), which means females have more peer trust compared to males and other genders. Sixth, low education positively correlates with altruism (r=.096, p<.01) and peer trust (r=.059, p<.05), which means people with a low education are more altruistic and have more peer trust compared to people with a higher education. Seventh, low income is negatively correlated to expected negative norms (r=-.061, p<.05), indicating people with a low income have lower expected negative norms compared to people with a higher income.

(33)

32

Figure 2. Hypotheses testing

*. Correlation is significant at the 0.05 level (2-tailed). **. Correlation is significant at the 0.01 level (2-tailed). ***. Correlation is significant at the 0.001 level (2-tailed).

(34)

Hypothesis 1.a

It is hypothesized that respondents who were presented a platform based on sociality, is associated with higher levels of altruism. The direct effect of sociality on altruism is Coeff. = .155, t = 2.518, p < .05. Therefore, Hypothesis 1.a is supported (see table 10).

Furthermore, four control variables are found to be predictors of altruism. Respondents who have a low education are associated with altruism of .317 units higher in comparison with respondents that have a higher education (Coeff. = .317, p<.01). Second, respondents who are one point more familiar with carsharing, are associated with altruism of .041 units higher (Coeff. = .041, p<.05). Third, respondents who are female are associated with altruism of .173 units lower in comparison with respondents that are male or identify as other (Coeff. =-.173, p<.01). Lastly, respondents who have one point more risk to family are associated with altruism of .095 units higher (Coeff. =.095, p<.01).

Hypothesis 1.b

It is hypothesized that respondents who were presented a platform based on sociality, is associated with lower expected negative norms. The direct effect of sociality on expected negative norms is Coeff. = -.171, t = -2.507, p < 0.05. Among the respondents expected negative norms is negatively influenced by sociality. Therefore, Hypothesis 1.b is supported (see table 10).

Additionally, six control variables are found to be predictors of expected negative norms. Respondents who have a low income are associated with expected negative norms of .187 units lower in comparison with respondents that have a higher income (Coeff. = -.187, p<.05). Second, respondents who are one point more familiar with carsharing, are associated with expected negative norms of .043 units higher (Coeff. = .043, p<.05). Third, respondents who have one point more risk to self are associated with expected negative

norms of .141 units higher (Coeff. =

.141, p<.01). Fourth, respondents in a corona situation are associated

with .174 units less expected negative norms compared to respondents in a no corona situation (Coeff.= -.174, p<.01). Fifth, respondents living in an urban area are associated with .279 units less expected negative

norms compared to respondents living in a rural area (Coeff. = -.279, p<.01). Lastly, one year increase of

(35)

34

Hypotheses 2.

The second hypotheses states that there is an indirect effect of sociality on altruism through peer

trust. This indicates that those respondents who were presented with a sociality platform experience a

significant increase in peer trust in comparison with respondents who were presented another platform (Coeff. = .128, t = 2.131, p < .05). This high peer trust is further associated with an increase in altruism (Coeff. = .475, t = 15.852, p < .000), independent of sociality. Since the bootstrap confidence interval is entirely above zero (indirect effect = .061, SE = .029, CI: .004 to .115), the indirect effect can be interpreted as significantly positive. In conclusion, a full mediation (Zhao, Lynch, & Chen, 2010) supports hypothesis 2 (see table 10).

In addition, five control variables are found to be predictors of peer trust. Respondents who are one point more familiar with car sharing platforms are associated with peer trust of .062 units higher in comparison with respondents that are not (Coeff. =.062, p<.000). Second, respondents who are associated with one point more risk to self are associated with peer trust of .140 units lower in comparison with respondents that are not (Coeff. =-.140, p<.000). Third, respondents who are female are associated with peer trust of .265 units higher in comparison with respondents that are male or identify as other (Coeff. =.265, p<.000). Fourth, respondents who are associated with one point more risk to family are associated with peer trust of .113 units higher in comparison with respondents that are not (Coeff. =.113, p<.01). Moreover, respondents who have low education are associated with peer trust of .230 units higher in comparison with respondents that have a higher education (Coeff. = .230, p<.05).

Hypotheses 3.

The third hypotheses states that there is an indirect effect of sociality on expected negative norms through

peer trust. This indicates that those respondents who were presented with a sociality platform experience a

significant increase in peer trust in comparison with respondents who were presented another platform (Coeff. = .128, t = 2.131, p < .05). This high peer trust is further associated with a decreased expected

negative norms (Coeff. = -.581, t = -17.507, p<.000), independent of sociality. Since the bootstrap

confidence interval is entirely below zero (indirect effect = -.074, SE = .036, CI: -.143 to -.005), the indirect effect can be interpreted as significantly negative. In conclusion, a full mediation (Zhao et al., 2010) supports hypothesis 3 (see table 10).

(36)

Hypothesis 4.a

It is hypothesized that respondents who were presented a platform based on power structures is associated with lower levels of altruism. The direct effect of power on altruism is Coeff. = -.062, t = -.925, p = .355. Therefore, Hypothesis 4.a is not supported (see table 10).

Hypothesis 4.b

It is hypothesized that respondents who were presented a platform based on power structures is associated with higher expected negative norms. The direct effect of power on expected negative norms is Coeff. = .014, t = .175, p = .861. Therefore, Hypothesis 4.b is not supported (see table 10).

Hypotheses 5.

The fifth hypotheses states that there is an indirect effect of power on expected negative norms through

enforced compliance. This indicates that those respondents who were presented with a power platform

experience a significant increase in enforced compliance in comparison with respondents who were presented another platform (Coeff. = .203, t = 2.634, p < .01). This high enforced compliance is further associated with an increase in expected negative norms (Coeff. = .015, t = .498, p = .619), independent of

power. Since the bootstrap confidence interval is not entirely above zero (indirect effect =.002, SE =.007,

CI: -.009 to .020), the indirect effect can be interpreted as insignificant. Therefore, Hypothesis 5 is not supported (see table 10).

Additionally, two control variables are found to be predictors of enforced compliance. One year increase of age is associated with .008 units less enforced compliance (Coeff. = -.008, p<.05). Furthermore, respondents who have one point more risk to family are associated with enforced compliance of .157 units higher (Coeff. = .157, p<.000).

Hypotheses 6.

The sixth hypotheses states that there is an indirect effect of power on altruism through enforced

compliance. This indicates that those respondents who were presented with a power platform experience a

significant increase in enforced compliance in comparison with respondents who were presented another platform (Coeff. = .203, t = 2.634, p < .01). This high enforced compliance is further associated with an

(37)

36

increase in altruism (Coeff. = .180, t = 7.103, p < .000), independent of power. Since the bootstrap confidence interval is entirely above zero (indirect effect = .037, SE = .016, CI: .011 to .072), the indirect effect can be interpreted as significantly positive. In conclusion, a full mediation (Zhao et al., 2010) rejects hypothesis 6 (see table 10).

(38)

(39)

38

Table 10. Outcomes of the hypotheses

H1.a Sociality is associated with higher altruism. Supported

H1.b Sociality is associated with lower expected negative norms. Supported H2 The relationship between sociality and altruism is mediated by peer trust, such

that the presence of sociality is associated with higher levels of peer trust, which leads to higher levels of altruism.

Supported

H3 The relationship between sociality and expected negative norms is mediated by peer trust, such that the presence of sociality is associated with higher levels of peer trust, which leads to lower expected negative norms.

Supported

H4.a Power is associated with lower levels of altruism Rejected

H4.b Power is associated with higher levels of expected negative norms Rejected H5. The relationship between power and expected negative norms is mediated by

enforced compliance, such that the presence of power is associated with higher levels of enforced compliance, which leads to higher expected negative norms.

Rejected

H6. The relationship between power and altruism is mediated by enforced

compliance, such that the presence of power is associated with higher levels of enforced compliance, which leads to lower levels of altruism.

Platform governance and engagement behaviour in the sharing economy

Platform governance and engagement behaviour in the sharing economy.

2

4

Table of Contents

6

8

10

12

14

16

18

20

22

24

26

28

32

.141, p<.01). Fourth, respondents in a corona situation are associated

34

36

38