The relationship between justice and repeat purchase intentions
mediated by trust: a quantitative study of online auctions
Tess Samsom - 11924047 Business Administration (MSc) Digital Business Track
University of Amsterdam Supervisor: Andrea Ganzaroli Date: 22-06-2018
STATEMENT OF ORIGINALITY
This document is written by Student Tess Samsom who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This study aims to extend the current literature on justice to buyer-seller relationships by examining the effects of justice perceptions on repeat purchase intentions, directly and indirectly through trust. Data was collected from 782 consumers of the largest online auction marketplace in the Netherlands. Different types of justice are assessed with two models, which compare the effects of distributive-, procedural-, interpersonal- and informational justice on with the effect of overall justice on repeat purchase intention. Consistent with recent theoretical advances, results show that the model that includes overall justice yields the highest total effect. This suggests that a focus on distinct types of justice may not provide a complete picture of how individuals experience decision outcomes. However, it is noteworthy that informational justice has a strong and significant effect on trust in this study, which implies that buyers highly value transparency in the current age of digitisation. Theoretical and practical implications are provided, as well as suggestions for future research.
Keywords: Distributive Justice, Procedural Justice, Interpersonal Justice, Informational Justice, Overall Justice, Trust in Intermediary, Repeat Purchase Intention, Online Auction.
Table of contents
1. Introduction ...6
2. Literature review ...9
2.1 Online auction markets...9
2.2 Repeat purchase intention ... 10
2.3 Justice theory ... 11
2.3.1 Distributive justice ... 11
2.3.2 Procedural justice ... 12
2.3.3 Interpersonal and informational justice (interactional justice) ... 13
2.3.4 Overall justice ... 14
2.4 Justice and repeat purchase intention ... 15
2.5 Trust ... 16
2.6 Justice and trust ... 19
2.7 Trust and repeat purchase intention ... 20
2.8 Conceptual models ... 21
3. Data and methods ... 23
3.1 Research design ... 23
3.2 Measures ... 24
3.2.1 Distributive-, procedural-, interpersonal- and informational justice ... 24
3.2.2 Overall justice ... 25
3.2.3 Trust in intermediary ... 25
3.2.4 Repeat purchase intention ... 25
3.2.5 Control variables ... 26 3.3 Data collection ... 26 4. Results ... 28 4.1 Preliminary analyses ... 28 4.1.1 Data cleaning ... 28 4.1.2 Recoding variables ... 29
4.1.3 Descriptive analysis ... 29
4.1.4 Correlation analysis ... 32
4.1.5 Variance inflation factor analysis... 33
4.1.6 Normality analysis ... 34
4.1.7 Reliability analysis ... 36
4.1.8 Exploratory factor analysis ... 39
4.2 Regression analyses ... 40 4.2.1 Model 1 ... 46 4.2.2 Model 2 ... 51 4.2.3 Control variables ... 52 5. Discussion ... 54 5.1 Theoretical implications ... 54 5.2 Managerial implications ... 57
5.3 Limitations and future research ... 58
6. Conclusion ... 60
References ... 62
Appendix ... 70
List of tables
Table 1. Demographics of the sample ... 29
Table 2. Means, standard deviations and correlations ... 27
Table 3. Collinearity statistics ... 33
Table 4. Normality check ... 34
Table 5. Exploratory factor analysis ... 24
Table 6. Model 1 summary: distributive justice ... 41
Table 7. Direct, indirect and total effects of model 1: distributive justice ... 41
Table 8. Model 1 summary: procedural justice ... 42
Table 9. Direct, indirect and total effects of model 1: procedural justice... 42
Table 10. Model 1 summary: interpersonal justice ... 43
Table 11. Direct, indirect and total effects of model 1: interpersonal justice ... 43
Table 12. Model 1 summary: informational justice ... 44
Table 13. Direct, indirect and total effects of model 1: interpersonal justice ... 44
Table 14. Model 2 summary ... 45
Table 15. Direct, indirect and total effects of model 2 ... 45
List of figures Figure 1. Conceptual model 1 ... 22
Figure 2. Conceptual model 2 ... 22
Figure 3. Results of research model 1 ... 53
1. Introduction
Since the emergence of the Internet as a channel of commerce, online auction activity has significantly increased. The Internet has enabled online auction firms to reduce the costs of reaching customers, increase operational efficiency through real-time inventory management and minimise transaction fees. However, the rise of e-commerce has also lowered entry barriers and minimised switching costs for buyers, ultimately leading to intensified competition in the online auctions market (Cui, Lai & Lowry, 2016). Intensified competition in the online auctions market, accompanied by growing competition from other e-commerce channels, has exerted increasing pressure on managers of online auction firms to reconsider their business model and its effects on customer behaviour (Chen, Yen, Kuo & Capistrano, 2016). As the retention of current customers is found to cost less than obtaining new customers, online auction firms primarily aim to establish a sustainable competitive advantage through an increase in buyers’ intentions to repeat purchases (Reichheld & Schefter, 2000; Gommans, Krishnan & Scheffold, 2001). Furthermore, the power asymmetry relationship between buyers and seller leads to the possibility of opportunistic behaviour from sellers in online auction markets (Greenberg & Cropanzano, 2001). Therefore, buyers’ perceived justice may offer a means by which to explain and understand buyers’ intention to continue to bid for products from sellers in an online auction marketplace in the future (Chiu, Huang & Yen, 2010).
Little research has been dedicated to justice as a variable to initiate repeat purchase intentions. The impact of justice has been primarily examined within the context of organisational relationships. Organisational justice research focuses on the effects of justice on various decision outcomes in employee-organisation relationships, such as satisfaction, commitment and turnover intentions. This study focuses on the decision outcome to continue to bid for products from an online auction website. Moreover, as existing literature identifies
intentions, this study tests the direct and indirect effects of different types of justice on repeat purchase intentions through trust.
The proposed research models predict that trust mediates the relationship between different types of justice and repeat purchase intentions. A literature review will be composed that narratively reviews the volume of theoretical work on the constructs of justice, trust and repeat purchase intentions. In addition, a digital survey will be designed to collect information from the target sample, which consists of consumers who have made repeat purchases from the largest online auction firm in the Netherlands. Thereafter, the relationships of interest will be analysed using statistical software. This allows for findings to be elaborated upon and conclusions to be drawn.
In terms of theory building, this study aims to extend the justice literature from employee-organisation relationships to buyer-seller relationships. It also attempts to explain paramount antecedents of trust and repeat purchase intentions in online auctions. Moreover, results may expand current theoretical understandings of individuals justice experiences and their reactions on them. These insights may benefit managers in their search for ways to increase revenues. Valuable information can be provided to managers on which to base their investment decisions in order to drive profitability through increased customer loyalty.
This research is structured as follows. Section 2 provides a detailed background of theories on online auction markets, repeat purchase intention, justice and trust, followed by the proposed conceptual model. Section 3 presents the research design and the measures that are adapted from existing literature to form the survey. In addition, the data collection process is described. In section 4, preliminary analyses are conducted as well as final regression analyses. Thereafter, section 5 discusses the theoretical- and practical implications of this study. Also,
limitations and suggestions for areas of future research are provided. In the last section, the conclusion sums up the aim, approach and findings of this research.
2. Literature review
In this section, the characteristics of online auction markets are listed. Thereafter, the theoretical construct of repeat purchases is described, followed by the introduction of four types of justice that have contributed to the current dominant insight of overall justice. Subsequently, the relationship between justice and repeat purchase intentions is scrutinized. Finally, the notion of trust is conceptualized and the literature on the justice-trust relationship, as well as the trust-repeat purchase intention relationship, is examined. The section concludes with the presentation of two conceptual models.
2.1 Online auction markets
Brown and Morgan (2006) describe the online auctions market as a two-sided market; the greater number of buyers and sellers exchanging goods on the platform, the more valuable the market. On the one hand, online sellers enjoy increased profits as the number of bidders rises. On the other hand, online buyers experiencesprice competition and an increased variety of products as the number of sellers rises. According to Hsu and Lu (2007), online auction markets also exhibit network effects, as the benefits for both buyers and sellers increase exponentially when the number of online users reaches a point of critical mass.
Gregg and Walczak (2010) distinguish between online auctions and traditional e-commerce in four ways. (1) In online auctions, many similar or even identical products are offered for sale by sellers at any point in time and, therefore, the marketplace is highly competitive. (2) Sellers in an online auction share a marketplace that is owned by a larger auction intermediary. This constrains sellers to distinguish themselves from others that operate in the same market. (3) As buyers bear the responsibility of setting prices according to their
willingness to pay, the prices that sellers receive depend on factors such as reputation and trust. (4) The online environment makes it easy for sellers to set-up and run an auction and, thereby, pretend to be reliable when they are not. This creates uncertainty from the buyers’ perspective.
2.2 Repeat purchase intention
Literature on purchase intentions is rooted in the theory of reasoned action (TRA) by Ajzen and Fishbein (1980). According to TRA, the most important determinant of a person’s actual behaviour is behavioural intent. The model, originating from the field of social psychology, proposes that the attitude of a person towards a certain behaviour is determined by his beliefs about the outcome. Hence, if a positive outcome from performing certain behaviour is expected, a person will decide to undertake actual behaviour (Pavlou, 2003). Therefore, online sellers consider the willingness of buyers to purchase to be an important variable of interest (McKnight & Chervany, 2002).
Many sources of previous e-commerce research do not make a distinction between buyers’ initial purchase intention and repeat purchase intention. Yet, Qureshi et al. (2009) argue that initial purchase intention and repeat purchase intention are conceptually different and should, therefore, be studied separately. This finding is supported by Parthasarathy and Bhattacherjee (1998), who state that initiating transactions with new customers may cost up to five times as much as retaining existing ones. Moreover, retained customers carry economic value since they have spending capacity, are more resistant to competitive messages, are less sensitive to price and can be served at lower cost (Reichheld & Sasser, 1990; Balabanis, Reynolds & Simintiras, 2006). Also, retained customers have a higher likelihood of generating positive word-of-mouth (Khalifa & Liu, 2007). Increasing customer retention by as little as 5%
These implications of customer retention for cost savings and profitability suggest that this research should focus on buyers’ repeat purchase intentions (Chen et al., 2016). Hence, this research focuses on repeat purchase intention, defined as an individual’s willingness to make another purchase from the same company based on previous experience (Kim, Galliers, Shin, Ryoo & Kim, 2012).
2.3 Justice theory
2.3.1 Distributive justice
Until 1975, the main area of focus in the justice literature has been distributive justice. Most of the research on distributive justice is derived from initial work conducted by Homans (1961) and Adams (1965). Homans (1961) introduced distributive justice as a term to describe the perceived fairness of decision outcomes. According to Homans (1961), individuals experience distributive justice when they believe their inputs are being fairly rewarded. Adams (1965) confirmed this finding by emphasizing that individuals are not concerned by the absolute level of outcomes per se, but whether those outcomes were fair compared to the inputs provided. Drawing upon equity theory in a social exchange framework, Adams (1965) suggested that distributive justice can be determined by calculating the ratio of an individuals’ inputs (e.g. money, time and effort) to the resulting outcome (e.g. product and/or service) and to compare that ratio with the ratio of relevant others. Inequity, consequently, exists when the perceived inputs and outcomes are incongruous with the inputs and/or outcomes of comparison individuals.
Whereas Adams (1965) proposed an equity basis to determine the fairness of decision outcomes, researchers have elaborated upon other values that may underlie distributive justice
(Deutsch, 1975; Leventhal, 1976; Frazier, Spekman & O’neal, 1988). Deutsch (1975), for example, proposed equality and needs as alternative bases for the perceived fairness of decision outcomes. Equality refers to the degree to which individuals receive the same outcome regardless of their inputs. On the other hand, needs that determine the perception of fair outcomes concern the degree to which outcomes fit the needs of the individual who receives the outcome. Altogether, the notions of equity, equality and needs are referred to as theories of distributive justice, as they share emphasis on the fairness of outcomes that individuals receive.
2.3.2 Procedural justice
In the 1980s, the focus in the justice literature shifted towards procedural justice. Thibaut and Walker (1975) proposed a general theory of procedural justice by examining disputes dealt with in the legal process. In their study, procedural justice is defined as the perceived fairness of the processes by which outcomes are distributed among parties to an exchange. Thibaut and Walker’s (1975) focus on procedural justice in the legal context has, consequently, spurred research on the study of process in non-legal settings. Leventhal (1980) developed a theory of procedural justice in organisational settings, which centres around six criteria that should by met in order for a procedure to be perceived as fair. Procedures should be (1) applied consistent across time and person, (2) deprived of personal bias on the part of decision makers, (3) based on accurate information, (4) accompanied by a system to appeal decisions, (5) reflect the interests and concerns of the relevant stakeholders and (6) executed in a moral and ethical manner. Thus, procedural justice reflects the perceived fairness of decision outcomes and the degree to which they are consistent, unbiased, accurate, correctable, representative and ethical (Colquitt et al., 2013). Based on the constructs of distributive and procedural justice, Cropanzano and Folger (1991) presented a two-component model of justice.
The authors suggest that ‘a full understanding of fairness cannot be achieved by examining the two constructs separately’. Therefore, the interaction between outcomes and procedures needs to be considered.
2.3.3 Interpersonal and informational justice (interactional justice)
So far, the study of organisational justice was primarily concerned with the two-factor model of distributive- and procedural justice. In the early 1990s, however, the social side of justice was introduced to expand the theory of justice. Bies and Moag (1986) examined the role of interpersonal treatment in procedures, named interactional justice. In their research, interactional justice is considered to be a subset of procedural justice and refers to the interpersonal treatment people receive as processes are carried out. Contrary to the perspective of Bies and Moag (1986) and several other researchers who view interactional justice as a subset of procedural justice (Tyler & Bies 1990; Niehoff & Moorman, 1993), others have treated interactional justice as a third type of justice (Aquino, 1995; Skarlicki & Folger 1997; Colquitt 2001). Skarlicki and Folger (1997), for example, examined retaliation against the organisation from a three-way interaction of distributive-, procedural- and interactional justice.
Greenberg (1993) introduced yet another perspective to the foregoing debate about interactional justice by suggesting a four-factor model of justice. According to the author, interactional justice is made up of two social aspects: interpersonal justice and informational justice. Interpersonal refers to the concern for buyers regarding the direct decision outcomes they receive and can be viewed as a facet of distributive justice. Informational justice, on the other hand, refers to the extent to which individuals are provided with information or rationale behind procedures, and can be viewed as a facet of procedural justice (Chiu et al., 2010). Colquitt (2001) demonstrated support for this four-factor structure of justice by treating
distributive, procedural, interpersonal and informational justice as distinct dimensions in a structural equation model.
2.3.4 Overall justice
More recent theoretical advances have questioned the benefits of focusing exclusively on specific types of justice and provide insight into the notion of overall fairness (Jones & Martens, 2009; Bobocel, 2013). Fairness heuristic theory proposes that a focus on distinct forms of justice may not provide a complete picture of how individuals experience decision outcomes. Hence, an assessment of overall justice adds an important construct and expands our theoretical understanding of individuals justice experiences and their reactions to them (Lind, 2001; Shapiro 2001; Van den Bos & Lind, 2002). Overall justice perceptions are defined by Holtz and Harold (2009) as global evaluations of the fairness of an event, based on both the individual’s own experience and those of others.
Consistent with recent insights into overall fairness, Cropanzano and Ambrose (2015) question the separate operationalisation of the dimensions of justice when the same pattern of findings for these dimensions is hypothesised. In addition, the authors acknowledge the drawbacks of modelling the four types of justice. Troublesome levels of multicollinearity are generated in structural models as the dimensions of justice are highly correlated. In particular, correlations may be high when all four types of justice are measured in the same time frame and from the same source. As a consequence of multicollinearity, standard errors will be inflated and bouncing betas will potentially be created. The findings of Colquitt and Shaw (2005) suggest that the problem of multicollinearity can be addressed by modelling overall justice as a higher order factor that drives scores on the procedural-, distributive-, interpersonal-
the data from 16 independent samples reasonably well. The loadings of the four types of justice were high and statistically significant, ranging from 0.64 to 0.86.
2.4 Justice and repeat purchase intention
In the existing literature, little is known about the relationship between justice and repeat purchase intentions. The effects of justice are predominantly assessed within the context of organisational relationships. Social exchange theory proposes that the feeling of justice is a key antecedent of trust in organisational employment relationships (Colquitt, 2011). Moreover, literature on buyer-seller relationships shows that trust is a crucial enabling factor in online transactions (Pavlou & Gefen, 2004). Following from the previous findings, this study offers an extension of the justice literature from employee-organisation relationships to buyer-seller relationships by examining and comparing the effects of distributive-, procedural-, interpersonal-, informational- and overall justice on repeat purchase intention. The line of reasoning behind the proposed extension is that both organisational relationships and buyer-seller relationships are characterised by information asymmetry. Due to fear of opportunism, employees and buyers rely on perceived feelings of justice and trust to avoid mistreatment (Chiu et al., 2010). Therefore, a similar pattern is expected for the justice - repeat purchase intention relationship as for the trust - repeat purchase intention relationship. Literature demonstrates that trust is a significant positive predictor of customers’ repurchase intentions (Chiu et al., 2009). Hence, positive relationships between different types of justice and repeat purchase intentions are hypothesised as well.
H1: There is a positive relationship between distributive justice and repeat purchase intention
H3: There is a positive relationship between interpersonal justice and repeat purchase intention
H4: There is a positive relationship between informational justice and repeat purchase intention
H5: There is a positive relationship between overall justice and repeat purchase intention
2.5 Trust
The importance of trust as enabling factor in the formation of buyers’ repeat purchase intention leads to the examination of the theoretical construct of trust in this study. Many researchers have provided definitions for the concept of trust in online environments. Geyskens, Steenkamp, Scheer and Kumar (1996) define trust as a ‘belief or expectation that the word or promise by the seller can be relied upon and the seller will not take advantage of the consumer’s vulnerability’. Moreover, Gambetta (1998), Bhattacharya, Devinney and Pillutla (1998) and McKnight and Chervany (2002) define trust as the ‘subjective assessment of one party that another party will perform a particular transaction according to his or her confident expectations, in an environment characterized by uncertainty’. Following Pavlou and Gefen (2004), this research uses a trust-transference logic in which trust in an intermediary builds buyer trust in sellers. An online intermediary is a third-party institution that uses a web-based infrastructure to facilitate transactions among buyers and sellers in its online marketplace. Intermediary functions include the balancing and integrating of the, often conflicting, buyer and seller needs (Sarkar, Butler & Steinfield, 1995). Because of the perceived positive association with the intermediary, buyers who trust the intermediary should also trust the sellers. On the other hand, if trust in the intermediary is lost this will result in an
erosion of trust in the seller (Pavlou & Gefen, 2004). Consistent with Pavlou and Gefen (2004), this research defines trust in the intermediary as ‘the subjective belief with which a buyer believes the intermediary will institute and enforce fair rules, procedures, and outcomes in its marketplace competently, reliably, and with integrity, and, if necessary, will provide recourse for buyers to deal with seller opportunistic behaviour’.
Although researchers have found many factors affecting the trustworthiness of sellers, Mayer, Davis and Schoorman (1995) identify ability, benevolence and integrity to encompass the major factors in existing literature. Ability is described as ‘that group of skills, competencies and characteristics that enable a party to have influence within some specific domain.’ Moreover, benevolence is referred to as the extent to which a seller is believed to want to do good to the buyer, aside from an egocentric profit motive. Lastly, integrity involves the buyer’s perception that the seller adheres to a set of principles that the buyer finds acceptable. The generic topology of Mayer et al. (1995) is still used by recent researchers as a basis for their studies (Gregg & Walczak, 2010; Hsu, Chunag & Hsu, 2014; Wu, et al., 2014).
In traditional offline transactions among strangers, Resnick and Zeckhauser (2002) have identified eight factors of trust creation. (1) Individuals have the opportunity to inspect the goods, as most transactions are conducted locally. (2) Buyers have frequent interaction with the same seller and, therefore, learn whom they can trust. (3) Given that a seller’s sales are concentrated in a local market, buyers learn from their peers and seller reputations are easily developed. (4) Seller reputations are borrowed from other contexts (e.g. pillars of the community), which provides status. (5) Taking the leading auction houses Sotheby’s (since 1744) and Christie’s (since 1766) as an example, reputations are built over many years. (6) Reputations are borrowed from others (e.g. celebrities). (7) New products signal reputation as they benefit from established brand names. (8) Significant expenditures indicate the reliability
of the seller (e.g. building a sophisticated store). These factors of trust formation are not available in the online auction market.
In comparison to traditional offline transactions, in online auction markets the buyer-seller relationship reflects an imbalance of power as a result of two information problems; adverse selection and moral hazard (Pavlou, Liang & Xue, 2007). A situation of adverse selection originates with the nature of a seller, which is either honest or dishonest (e.g. an honest seller advertises its goods according to its true quality, whereas a dishonest seller fails to provide the true quality of the good). Thereafter, the seller and buyer agree upon a contract and, consequently, the seller behaves according to its nature. In the case of moral hazard, a buyer and seller begin with symmetric information and agree to a contract. However, the seller decides to undercut quality of a good in order to maximise profit while the buyer is unable to observe its behaviour (Li, 2010). Thus, the fact that buyers do not have complete information to assess product conditions leads to the classic lemon problem, first posed by Akerlof in 1978, in which the seller has an incentive to act opportunistically (Ba, 2001; Ba & Pavlou, 2002; Li, 2010; Chiu et al., 2010; Gregg & Walczak, 2010). Due to fear of opportunism in online transactions, trust has been identified as prominent enabling factor in online auctions (Gefen & Straub, 2003).
Common sources of opportunistic behaviour by sellers in online auctions are the reception of payment without delivering the purchased item (Choi, Stahl & Whinston, 1997), delivery of a good that is different from the one described in the product description, e.g. fake or defective (Pavlou & Gefen, 2005), and product delivery delay (Javenpaa, Tractinsky & Vitale, 2000). Other sources of opportunistic behaviour are high handling and shipping costs (Chiu et al., 2010), refusal to accept payment (contract default) (Pavlou & Gefen, 2005) and
offering a return policy but then failing to acknowledge product guarantees (Mishra, Heide & Cort, 1998).
2.6 Justice and trust
Lewicki, Wiethoff & Tomlinson (2005) state that the existing body of theoretical and empirical work evidently proposes a strong relationship between justice and trust. This volume of literature includes social exchange theory (Blau, 1964) and fairness heuristic theory (Lind, 2001). According to social exchange theory, perceptions of distributive-, procedural-, interpersonal- and informational justice are central to the development of employees’ trust in their supervisors. Konovosky and Pugh (1994) subsequently find in an empirical study of hospital employees that trust in the supervisor mediates the relationship between distributive justice, procedural justice and organisational citizenship behaviour. More recently, Kernan and Hanges (2002) tested survivor reactions to reorganisation and demonstrated that procedural-, interpersonal- and informational justice are predictors of perceived managerial trustworthiness. With regard to fairness heuristics theory, Kim and Leung (2007) note that perceived overall fairness also has a strong relationship with various outcomes that are beneficial to employees and organisations, one of which is trust. More specifically, Jones and Martens (2009) state that overall fairness perceptions determine whether individuals trust organisational authorities. Individuals are especially likely to let overall fairness perceptions determine trust when uncertainty about their assessment of trust prevails (Jones & Martens, 2009).
Given the strong relationship between different types of justice and trust posed by theories on social exchange and fairness heuristics, Turel, Yuan and Connelly (2008) examined the impact of justice perceptions on trust in online buyer-seller relationships during the transaction process. The study finds significant positive effects of distributive-, procedural-,
interpersonal- and informational justice on trust. Moreover, research by Chiu et al. (2010) reports positive relationships between overall justice and trust. Hence, positive relationships between different types of justice and trust in the intermediary are hypothesised in this study.
H6a: There is a positive relationship between distributive justice and trust in intermediary H7a: There is a positive relationship between procedural justice and trust in intermediary H8a: There is a positive relationship between interpersonal justice and trust in intermediary
H9a: There is a positive relationship between informational justice and trust in intermediary H10a: There is a positive relationship between overall justice and trust in intermediary
2.7 Trust and repeat purchase intention
Many researchers have examined the influence of trust on repeat purchase intentions in the online transaction process. Trust plays a vital role in the formation of buyer expectations, as it assures buyers that the seller is both able and willing to deliver the goods and services purchased (McKnight, Choudhury & Kacmar, 2002). Trustworthiness gives buyers some perception of control over a potentially uncertain transaction (Pavlou, 2003). Therefore, trust is considered to be an effective mental shortcut for buyers to cope with uncertainty in complex exchange relationships (Chiu et al., 2010). The positive influence of trust on repeat purchase intention has been tested empirically by many researchers (Hsu, Chang & Chuang, 2015). To illustrate, Zhou, Lu and Wang (2009) found that consumers’ trust and satisfaction are important factors increasing buyers willingness to purchase again on e-commerce websites. In addition, Qureshi et al. (2009) demonstrate that a returning buyer’s trust in an online seller is positively related to his/her intention to repurchase from the online seller. According to the authors, acknowledgement of this finding is crucial as it provides valuable information for sellers on
which to base investment decisions in trying to drive repurchase behaviour. Finally, the results of a study on online behaviour continuance models by Chiu, Hsu, Lai and Chang (2012) indicates that trust has a significant positive effect on repeat purchase intention in online stores. Based on the previously mentioned research, this study hypothesises a positive relationship between trust in intermediary and repeat purchase intention.
H6b-H10b: There is a positive relationship between trust in intermediary and repeat purchase intention
2.8 Conceptual models
To summarise the relationships elaborated upon in the literature review, Figure 1 and 2 propose the conceptual models that will be tested in this study. Model 1 predicts that trust mediates the relationship between distributive justice, procedural justice, interpersonal justice, informational justice (separate independent variables) and repeat purchase intentions (dependent variable). Consistent with more recent theoretical advances, model 2 predicts that trust mediates the relationship between overall justice (independent variable) and repeat purchase intentions (dependent variable). To rule out potentially spurious relations, this study controls for age (in years), gender (0 = male, 1 = female, 2 = gender neutral), e-commerce experience (in years) and experience in online auctions (in years). In summary, this study aims to answer the following research question: What are the effects of justice on repeat purchase intentions in online auctions, both directly and indirectly through an increase in trust in intermediary?
Figure 1. Conceptual model 1
Figure 2. Conceptual model 2
Distributive justice H1 (+) H10a (+) Trust in intermediary Repeat purchase intention Procedural justice Interpersonal justice Informational justice H2 (+) H3 (+) H4 (+) Trust in intermediary H5 (+)
Overall justice purchase Repeat
intention H6a (+) H6-H9 H7a (+) H8a (+) H9a (+) H10b (+) H10 H6b (+) H7b (+) H8b (+) H9b (+)
3. Data and methods
In this section, the research design of this study is presented. Thereafter, measures are adapted from literature and used to measure survey items. Finally, the data collection process is described.
3.1 Research design
This research uses deductive reasoning to test for theories on justice, trust and repeat purchase intentions. As source of primary data, a digital survey is designed to collect information from the target sample. The survey is of explanatory nature, thereby studying the online auction setting in order to explain relationships between variables. Moreover, the research is observational and is conducted on a cross-sectional basis for reasons of a limited time horizon, thus examining the variables of justice, trust in intermediary and repeat purchase intentions as they exist in a defined population at a single point in time. A pilot test has been administered to 20 persons to minimise the likelihood of respondents having problems with responding to questions. In addition, the pilot test allowed for an assessment of the questions’ validity and the reliability of the data that was going be collected (Saunders, Lewis & Thornhill, 2012). The resulting feedback has been used to improve the survey. To illustrate, the pilot test showed that the term ‘transaction policy’ was perceived to be unclear as no reference was made to a transaction policy on the website of online auction intermediary BVA Auctions, whose customers were going to be addressed in the survey. Therefore, the term ‘transaction policy’ has been replaced by the term ‘revocation right’, which is frequently mentioned on the website of BVA Auctions. Another adjustment that has been made to the survey as a result of the pilot test concerned measurement items RPI1 – RPI3 (see appendix A). The direct translation of these items to Dutch proved to be ambiguous. Hence, the terms have been adapted in order for respondents to understand the nuances between these questions.
The research model will be tested with data from buyers of the online auction intermediary BVA Auctions. BVA auctions was chosen because it is the largest online auction marketplace in the Netherlands. The sampling frame is the list of registered consumers of the BVA Auctions, who have been addressed via e-mail. As the members of the target population have an equal and independent chance of being selected to participate in this survey, the research has been conducted using a probability simple random sample. Moreover, the survey was administered in Dutch. The back-translation procedure was followed to translate the measurement items.
3.2 Measures
3.2.1 Distributive-, procedural-, interpersonal- and informational justice
Colquitt’s (2001) scale, adapted to refer to online auctions (Chiu et al., 2010), was used to measure distributive justice (four-items; e.g. “I got what was fair compared to the effort and time I spent in bidding” and Cronbach’s α = 0.95), procedural justice (four-items; e.g. “The company responds to requests or questions in a timely manner” and Cronbach’s α = 0.89), interpersonal justice (four-items; e.g. “The company has treated me in a polite manner” and Cronbach’s α = 0.95) and informational justice (four-items; e.g. “The company has provided me with accurate information about products” and Cronbach’s α = 0.95). Participants responded to survey items using a seven-point Likert-type scale, ranging from 1 = “strongly disagree” to 7 = “strongly agree”.
3.2.2 Overall justice
Overall justice was measured using Ambrose and Schminke’s (2009) three-item measure for personal justice experience, including “Overall, I have been treated fairly by the company” (Cronbach’s α = 0.93). Participants responded to survey items using a seven-point Likert-type scale, ranging from 1 = “strongly disagree” to 7 = “strongly agree”.
3.2.3 Trust in intermediary
Trust in intermediary was measured using Pavlou and Gefen’s (2004) three-item measure, including “As an auction intermediary, the company has high integrity” (Cronbach’s α = 0.96). Measurement item TR3 (see appendix A) has, however, been adapted in consultation with my thesis supervisor. In the study of Pavlou and Gefen (2004) TR3 is a direct measure of trust: “As an auction intermediary, the company can be trusted at all times”. The purpose of questions TR1-TR3, however, is to measure trust indirectly through three different measures. Hence, an adjustment of TR3 was deemed necessary. Items TR1-TR2 measure contractual trust through integrity and knowledgeability. To complement contractual trust, item TR3 has been adapted to measure moral trust. Participants responded to survey items using a seven-point Likert-type scale, ranging from 1 = “strongly disagree” to 7 = “strongly agree”.
3.2.4 Repeat purchase intention
Repeat purchase intention was measured using Khalifa and Liu’s (2007) three-item measure, including “I expect to repurchase from the company in the near future” (Cronbach’s α = 0.84). Participants responded to survey items using a seven-point Likert-type scale, ranging from 1 = “strongly disagree” to 7 = “strongly agree”.
3.2.5 Control variables
To rule out potentially spurious relations, this study controls for age (in years), gender (0= male, 1 = female, 2 = gender neutral), e-commerce experience (in years) and experience in online auctions (in years). These control variables are commonly used in studies on repeat purchase intentions in online auctions (Chiu et al., 2010; Chen et al., 2016; Wu, Cheng & Yen, 2014).
3.3 Data collection
The primary data was collected using a digital survey that was distributed through Qualtrics. In February 2018, contact has been sought with the marketing director of BVA Auctions to explore the possibilities of disseminating a digital survey among its customers. The marketing director gave approval for distribution of the survey and, consequently, the survey has been designed. In the week of 16 – 22 April, the target population was selected in cooperation with the marketing department. First, a selection has been made on the basis of opt-in. Only consumers who have explicitly and demonstrably given permission to receive e-mails from BVA Auctions could be addressed. The second selection criterion is home delivery auctions, which allows for questions to be asked such as “I received the products from BVA in a timely manner” (measurement item DJ4). Some auction formats (e.g. foreclosure auctions) allow buyers to obtain the product directly at site and would, therefore, not be suitable to answer such survey questions. Third, consumers were selected who have been the highest bidder in more than one home delivery auction, thereby classifying as customers with repeat purchase intentions. Finally, only customers have been addressed who satisfy the previous selection criteria and have, additionally, made their last purchase in the previous three months (February, March and April of 2018). As a result, the target population consisted of 12,340
consumers of BVA Auctions. These consumers have been addressed via e-mail in BVA format containing an anonymous link to the survey in Qualtrics. After a week, a reminder e-mail has been sent to the target population. The final number of respondents amounted to 782, which is a response rate of 6.3%. This amount of respondents exceeded the ideal sample size of 630, which was calculated on the basis of a 99% confidence level and a 5% margin of error. No respondents are discarded from the data file because every participant has completed the survey.
4. Results
In this section, analyses are conducted of the collected data and the subsequent results are demonstrated. Data cleaning methods, descriptive statistics, a correlation matrix and variance inflation factors are provided, as well as normality checks. Thereafter, reliability analyses are run and an exploratory factor analysis is presented. Finally, simple mediation regression analyses are run and the results of these analyses are interpreted.
4.1 Preliminary analyses
4.1.1 Data cleaning
After collection of the data through Qualtrics, the dataset of this sample was exported to the program IBM SPSS Version 25 for further analysis. Before testing the hypothesised conceptual models, the dataset was cleaned. First, a frequency check was run in order to see how many times each of the scores occurs in the data and to, consequently, identify possible errors. Not all values fell into the category of the variables. To illustrate, the measurement item distributive justice had values of 1, 30, 31, 33, 35, 39 and 40 instead of a seven-point Likert-type scale, ranging from 1 to 7. The error probably existed due to a conversion problem from Qualtrics to SPSS. Hence, the variable was transformed and recoded into the appropriate Likert-type scale values. Another frequency check demonstrated that no additional errors were found in the dataset.
Second, an outliers check was performed to recognise cases that have extremely high or extremely low values in one variable. Outliers can bias mean, standard deviation, skewness and kurtosis (Barnett & Lewis, 1994). An outliner is identified when z > | 3 |. Standardised variables were saved and, consequently, frequencies were run on these variables. The z-values
for all variables are between -3 and 3, so there were no outliers that had to be excluded from further analyses.
4.1.2 Recoding variables
This study does not contain any counter-indicative items, which means that there are no survey items that are phrased so that an agreement with the item represents a low level of the measured construct. Hence, there is no need to recode the survey items by reverse scoring it. Nevertheless, several variables were recoded into groups as they were recorded at the highest measurement level. To illustrate, the control variable age was recoded into groups (under 25, 26-35, 36-45, 46-55, older than 56) as well as e-commerce experience and auction experience (less than 5, 6-10, 11-15, 16-20, more than 21).
4.1.3 Descriptive analysis
Table 1. Demographics of the sample
Demographics N = 782 Gender Male Female Gender neutral Frequency 613 164 3 Percentage 78.4% 21.0% .4% Age < 25 26 – 35 36 – 45 46 – 55 > 56 11 66 160 227 313 1.4% 8.4% 20.5% 29.2% 40.3% E-commerce experience < 5 years 6 – 10 years 11 – 15 years 16 – 20 years 170 350 164 63 21.7% 44.8% 21.1% 8.1%
> 21 years 29 3.7%
Online auction experience
< 5 years 6 – 10 years 11 – 15 years 16 – 20 years > 21 years 506 217 35 13 2 64.7% 27.7% 4.5% 1.7% 0.3%
From the demographics in Table 1, it can be concluded that men are overrepresented in this survey (78%). Hence, this study may be susceptible to gender bias. In addition, some age differences are visible among participants, yet relatively old participants (56 years and older) represent the largest group of respondents (40%). In terms of e-commerce experience, most participants (45%) had 6 to 10 years of experience with buying products online. Finally, more than half of the respondents (65%) had less than 5 years of experience with online auctions. This is line with expectations that, for many participants, the years of experience in e-commerce exceed the years of experience in online auctions. Most consumers try out shopping online on retail websites to get an understanding of reasonable prices for products before daring to participate in a bidding process on an online auction website. The winner’s curse may occur on online auction websites as participants in online auctions are unable to physically inspect the items for sale and are, consequently, uncertain about the value of the auction item. This means that the winner of the auction with the highest estimate may pay a greater price than what the item is worth (Hou, Kuzma & Kuzma 2009). Therefore, there is more risk involved with online auctions than on retail websites.
Table 2. Means, standard deviations and correlations
Variables M SD 1 2 3 4 5 6 7 8 9 10 11
1. Gender .22 .42 -
2. Age 51.63 11.92 .04 -
3. E-commerce experience 10.51 5.68 .00 .13 -
4. Online auction experience 5.47 3.85 .01 .16** .44** -
5. Distributive justice 5.74 1.21 .04 .09* .08* .64 (.82) 6. Procedural justice 5.34 1.24 .00 .11** .01 .03 .47** (.90) 7. Interpersonal justice 5.92 1.2 -.01 .06 .07 .02 .54** .72** (.97) 8. Informational justice 5.49 1.08 .02 .12** .05 .05 .65** .70** .66** (.81) 9. Overall justice 5.96 1.11 -.03 .07* .06 .03 .62** .63** .78** .71** (.93) 10. Trust in intermediary 5.64 1.13 -.02 .09** .05 .05 .59** .68** .70** .77** .78** (.88)
11. Repeat purchase intentions 6.16 1.16 -.00 .02 .04 .07* .60** .57** .62** .64** .69** .67** (.96) *. Correlation is significant at the .05 level (2-tailed)
4.1.4 Correlation analysis
A correlation matrix was compiled with the scale means of the variables distributive justice, procedural justice, interpersonal justice, informational justice, overall justice, trust in intermediary and repeat purchase intention, as presented in Table 2. The correlation analysis studies the relationship between variables and gives insight on the quantification of the relationships between the variables of interest. Several correlations are found among the control variables age, commerce experience and online auction experience. Especially e-commerce experience and online auction experience show a tendency to a positive relationship (r = .44, p < .01), according to the rules of thumb to interpret linear relationships between two variables from Field (2013). This means that participants with a relatively long experience in buying products online also tend to have a long experience with buying through online auctions. Furthermore, age significantly correlates with distributive justice (r = .09, p < .05), procedural justice (r = .11, p < .01), informational justice (r = .12, p < .01), overall justice (r = .07, p < .05) and trust in intermediary (r = .09, p < .01). This would imply that older participants perceive higher levels of justice and trust. However, the correlations are so small that this may be an indication of the absence of relationships among these variables.
Furthermore, significant and high positive relationships are found between repeat purchase intention and different types of justice: distributive justice (r = .60, p < .01), procedural justice (r = .57, p < .01), interpersonal justice (r = .62, p < .01), informational justice (r = .64, p < .01) and overall justice (r = .69, p < .01). This can indicate the presence of the suggested relationships in hypotheses 1 to 5. Also, significant and high positive relationships are found between trust in the intermediary and different types of justice: distributive justice (r = .59, p < .01), procedural justice (r = .68, p < .01), interpersonal justice (r = .70, p < .01), informational justice (r = .77, p < .01) and overall justice (r = .78, p < .01). This may indicate
the presence of the suggested relationships that are tested in hypotheses 6a to 10a. Finally, a significant and high positive relationship are found between trust in the intermediary and repeat purchase intentions (r = .67, p < .01). This suggests the presence of the relationship in hypotheses 6b to 10b. Overall, the findings in the correlation matrix support the hypothesis 6 to 10 and may suggest the presence of a mediation effect.
4.1.5 Variance inflation factor analysis
Table 3. Collinearity statistics
Tolerance VIF
Effect of justice on repeat purchase intention
Distributive justice Procedural justice Interpersonal justice Informational justice Overall justice .52 .39 .31 .34 .32 1.92 2.57 3.19 2.90 3.16
Effect of justice on trust
Distributive justice Procedural justice Interpersonal justice Informational justice Overall justice .52 .39 .31 .34 .31 1.93 2.60 3.24 2.95 3.22
Effect of trust on repeat purchase intention
Trust in intermediary .33 3.03
As the variables of interest display high correlations, this may be an indication of multicollinearity. A variance inflation factor (VIF) analysis is conducted to test for problematic
levels of multicollinearity that could potentially affect the results. The VIF indicates to what extent a particular variable contributes to multicollinearity issues within the dataset. The higher the number of the VIF, the bigger the multicollinearity problem caused by that particular variable. The VIF is the reciprocal of the tolerance value (VIF = 1 / tolerance). A general rule of thumb is that numbers in excess of 10 indicate a strong multicollinearity problem (Hair et al., 1995). As can be seen in Table 3, this study does not exhibit multicollinearity problems and, therefore, no variables have to be removed from further analyses.
4.1.6 Normality analysis
Table 4. Normality check
Skewness SE Kurtosis SE Distributive justice -1.64 .09 2.80 .18 Procedural justice -.89 .09 .78 .18 Interpersonal justice -1.98 .09 4.62 .18 Informational justice -1.38 .09 2.60 .18 Overall justice -2.18 .09 6.12 .18 Trust in intermediary -1.43 .09 2.63 .18
Repeat purchase intention -2.37 .09 6.78 .18
To examine the extent to which the data was distributed symmetrically around the centre of all scores, normality checks were conducted. Tests for skewness and kurtosis are displayed in Table 4. In testing for skewness and kurtosis, H0 predicts a normal distribution and H1 predicts a non-normal distribution. Normality is assumed within the bounds of the 95% confidence interval. Thus, if zero falls within the lower boundary (statistic – 1.96 * std. error)
As zero does not lie within the confidence interval of any variable, the skewness and kurtosis tests cannot be accepted for the variables distributive justice, procedural justice, interpersonal justice, informational justice, overall justice, trust in intermediary and repeat purchase intention. However, tests for skewness and kurtosis are considered to be powerful tests that often reject H0, thereby confirming non-normality. According to empirical criteria (West, Finch & Curran, 1995), acceptable values for normal distribution lie between -1 and +1. Relying on this rule of thumb, the variable procedural justice is normally distributed, while the other variables still display non-normality. Additional histograms and Q-Q plots and demonstrated that the most frequent scores of all variables were grouped towards the higher end of the distribution and show a sharp peak. Thus, all variables were negatively skewed and leptokurtic. This implies that the majority of the respondents reported relatively high levels of distributive justice, procedural justice, interpersonal justice, informational justice, trust in intermediary and repeat purchase intentions. The Kolgomorov-Smirnov and Shapiro & Wilk tests of normality have a significance level of .000, which is not p. >.05. This means that the null hypothesis of normal distribution cannot be accepted. The absence of normal distribution can be explained by literature. It is found that a higher perceived feeling of justice leads to higher levels of trust (Chiu et al., 2010) and trust, in its turn, contributes to repeat purchase intentions (Zhou et al., 2009). With regard to online auctions, it may be the case that the customers of an online auction firm have registered because they intend to participate in multiple auctions. Customers who do not wish to participate in auctions anymore can deregister from the website. It can, thus, be more difficult to include customers with low levels of perceived justice, less trust and a low intention to repurchase in the sample. Therefore, participants that score low values on the variables may be underrepresented.
To normalise the distribution, the variables were transformed. First, due to the display of moderate negative skewness (values -.5 to -1), the variable procedural justice was
transformed with Square Root (X* = Ö (K – X), where K is the highest value of the variable X + 1). Second, due to the display of substantial negative skewness (values -1 to -2), the variables distributive justice, interpersonal justice, informational justice and trust were transformed with Log (X* = Log10(K – X), where K is the highest value of the variable X + 1). Finally, due to the display of extreme negative skewness (values <-2), the variables overall justice and repeat purchase intention were transformed with Reciprocal (X* = 1/(K-X), where K is the highest value of the variable X + 1). The variables are reflected as a consequence of transformation, which means that the direction of the result interpretation has to be reversed as well. The technique of variable transformation has corrected for distributional problems. Now, all variables reach acceptable skewness and kurtosis values (between -1 and +1).
4.1.7 Reliability analysis
To test the extent to which data collection techniques yield consistency, reliability analysis was conducted (Saunders, Lewis & Thornhill, 2012). The distributive justice scales have a Cronbach’s Alpha of .83 and, therefore, a high reliability. The procedural justice scales a Cronbach’s Alpha of .90 and, therefore, also a high reliability. The interpersonal justice scales have the highest Cronbach’s Alpha, which measures up to .97. The informational justice scales have a Cronbach’s Alpha of .81, thereby demonstrating a high reliability as well. The overall justice scales have a Cronbach’s Alpha of .93 and, thus, a high reliability. The trust in intermediary scales have a high reliability by obtaining a Cronbach’s Alpha of .88. Finally, the repeat purchase intention scales show a Cronbach’s Alpha of .96. In conclusion, all measurement items display Cronbach Alpha’s in the range of .83 to .97. This is well above the threshold of α > .70. Therefore, the data collection techniques yield consistency and analytic procedures would produce consistent findings if they were repeated on another occasion
Furthermore, the corrected item-total correlations indicate that all scales of the variable distributive justice range from .53 to .73. For the scales of the variable procedural justice, the corrected item-total correlations range from .68 to .83. In addition, the corrected item-total correlations of the procedural justice scales range from .89 to .94. The corrected item-total correlations of the informational justice scales range from .58 to .68. Moreover, the corrected total correlations of the overall justice scales range from .80 to .88. The corrected item-total correlations of the trust in intermediary scales range from .70 to .80. Finally, the corrected item-total correlations of the repeat purchase intention scales range from .91 to .94. In conclusion, all corrected item-total correlations are well above the threshold of > .30. Hence, the variables in the dataset have a good correlation with the total score of the scale.
Finally, the delta of the Cronbach’s Alpha if the items of distributive justice were deleted ranges from .00 to .08. The delta of the Cronbach’s Alpha if the items of procedural justice were deleted ranges from .00 to .05. In addition, the delta of the Cronbach’s Alpha if the items of interpersonal justice were deleted from .00 to .01. The delta of the Cronbach’s Alpha if the items of informational justice were deleted ranges from .05 to .02. The delta of the Cronbach’s Alpha if the items of overall justice were deleted ranges from .00 to .07. The delta of the Cronbach’s Alpha if the items of trust in intermediary were deleted ranges from .00 to .08. Finally, the Cronbach’s Alpha if the items of repeat purchase intentions were deleted ranges from .02 to .03. In conclusion, none of the items would substantially affect reliability in case of deletion (∆ are below .10). Hence, no variables are removed from further analyses.
Table 5. Exploratory factor analysis
Rotated factor loadings
Item PJ DJ IPJ/OJ RPI
I got what I paid for at the BVA marketplace -.06 .75 -.10 -.00
I got what was fair compared to the effort and time I spent in bidding -.08 .67 -.09 -.00
The products I received from BVA have the same quality as advertised .04 .81 .08 -.00
I received the products from BVA in a timely manner .12 .41 -.09 -.11
BVA Auctions responds to requests or questions in a timely manner .45 -.03 0.30 -.08
BVA Auctions complies with revocation rights .82 -.07 -.06 .05
BVA Auctions has fair policies and practices to handle problems or disputes .84 -.03 -.06 .01
BVA Auctions seriously considers buyers’ objections or suggestions .84 -.10 -.07 -.04
BVA Auctions has treated me in a polite manner .04 .01 .90 .04
BVA Auctions has treated me with sincerity .07 .03 .88 .03
BVA Auctions has treated me with respect .02 -.05 .95 -.02
BVA Auctions has treated me with friendliness .04 -.05 .99 -.02
BVA Auctions has provided me with accurate information about products .23 .61 .05 -.05
BVA Auctions has provided me with adequate information about revocation rights or any changes in policies .56 .20 -.01 -.07 BVA Auctions has provided me with order processing details in a timely manner .28 .34 -.07 -.13 BVA Auctions seems to tailor its communications to my individual specific needs .64 .14 .07 -.06
Overall, I have been treated fairly by BVA Auctions .00 .26 .53 -.20
In general, I can count on BVA Auctions to be fair .11 .24 .43 -.22
In general, the treatment I receive from BVA Auctions is fair -.01 .24 .51 -.16
As an auction intermediary, BVA auctions has high integrity .29 .29 -.16 -.19
BVA Auctions is a competent and knowledgeable auction intermediary .23 .32 -.20 -.19
BVA Auctions does its best to satisfy customers, even beyond its contractual obligations .51 .16 -.12 -.13
I anticipate to repurchase from BVA Auctions in the near future -.03 .00 -.03 .96
It is likely that I will repurchase from BVA Auctions in the near future -.00 -.05 -.00 .97
I expect to repurchase from BVA Auctions in the near future -.02 -.02 .04 .96
Eigenvalues 13.72 1.90 1.26 1.11
4.1.8 Exploratory factor analysis
A principal axis factoring analysis (PAF) was conducted on the scales for the dimensions of distributive justice, procedural justice, interpersonal justice, informational justice, overall justice, trust in intermediary and repeat purchase intention. The Kaiser-Meyer-Olkin measure verified the sampling adequacy for the analysis (KMO = .957). Bartlett’s test of sphericity χ2 (231) = 18789.280, p < 0.001, indicated that correlations between items were sufficiently large for PAF. Four components had eigenvalues over Kaiser’s Criterion of 1. This indicates that four factors can be obtained, which in this case are the dimensions distributive justice, procedural justice, interpersonal justice or overall justice and repeat purchase intention. In combination these factors explained 71.93% of the variance. Examination of the scree plot revealed levelling off after the fourth factor in agreement with Kaiser’s criterion. Therefore, four factors were retained and rotated with an Oblimin with Kaiser normalization rotation. Table 5 shows the factor loadings after rotation. The items that cluster on the same factors suggest that factor 1 represents distributive justice, factor 2 procedural justice, factor 3 a combination of interpersonal justice (higher factor loadings) and overall justice (lower factor loadings) and factor 4 repeat purchase intention. Two items of informational justice show high cross-loadings on the factor procedural justice, while one other item of informational justice has high cross-loadings on the factor distributive justice. This could be due to the similarity of content amongst the different items of justice (e.g. questions about revocation rights). In addition, one item of trust in intermediary shows high cross-loadings on the factor of procedural justice / overall justice. This could be due to the adaptation of the measurement item to include moral trust (see ‘Measures’ section), thereby diverging from the literature.
The findings of the exploratory factor analysis suggest that the factors informational justice and trust in intermediary should be excluded from further analysis. However, these findings should not be examined in isolation. The reliability analyses show that the collection
techniques for these constructs yield high consistency: the informational justice and trust in intermediary scales have a Cronbach’s Alpha of .81 and .96, respectively. Moreover, a solid substantiation is provided in the literature for the conceptual model that is used in this study. According to the deductive approach of this research, theory should drive statistics rather than the other way around. Therefore, the factors are not excluded for further analyses. A confirmatory factor analysis may have been more appropriate in this context, since the conceptual model already had a strong conceptual underpinning. Yet, the lack of access to statistical analysing software, such as AMOS and LISREL, prevented this study from conducting such analysis
4.2 Regression analyses
After obtaining insights from the dataset by examining descriptive statistics, correlations, variance inflation factors, normality, reliability and factor loadings, the hypotheses were tested through simple mediation analyses. To test the hypotheses, an external macro developed by A.F. Hayes (2012), named Process (Version 2.16.3), is used as add-on to the statistical analysing software IBM SPSS (Version 25). For this study, Model 4 of Process is used to test the hypothesized relationships. Several regressions are performed to examine the differential effects of distributive-, procedural-, interpersonal-, informational- and overall justice on repeat purchase intentions. Moreover, the suggested mediation effect of trust in intermediary is examined after controlling for gender, age, e-commerce experience and experience with online auctions.
Table 6. Model 1 summary: distributive justice
Consequent
Trust (M) Repeat purchase intentions (Y)
Antecedent Coeff. SE p Coeff. SE p
Distributive justice (X) a1 .55 .03 <.01 c1’ .30 .03 <.01
Trust (M) --- --- --- b1 .50 .03 <.01
Gender -.12 .08 .11 -.01 .07 .83
Age .00 .00 .23 -.01 .00 .02
E-commerce experience -.00 .01 .80 -.01 .01 .22
Online auction experience .00 .01 .72 .01 .01 .12
constant i1 2.35 .17 <.01 i2 1.90 .20 .<.01
R2 = .35 R2 = .52
F(5,759) = 81.48, p = .00 F(6,758) = 135.32, p = .00
Table 7. Direct, indirect and total effects of model 1: distributive justice
Effect SE p LLCI ULCI
Direct effect c1’ .30 .03 <.01 --- ---
Total effect c1 .58 .03 <.01 --- ---
Boot SE Boot LLCI Boot ULCI
Table 8. Model 1 summary: procedural justice
Consequent
Trust (M) Repeat purchase intentions (Y)
Antecedent Coeff. SE p Coeff. SE p
Procedural justice (X) a1 .62 .03 <.01 c1’ .21 .03 <.01
Trust (M) --- --- --- b1 .53 .04 <.01
Gender -.05 .08 .48 .02 .07 .83
Age .00 .01 .60 -.01 .01 .03
E-commerce experience .01 .01 .18 -.00 .00 .71
Online auction experience .00 .01 .80 .01 .01 .11
constant i1 2.19 .18 <.01 i2 2.26 .01 <.01
R2 = .46 R2 = .48
F(5,758) = 129.18, p = .00 F(6,757) = 117.08, p = .00
Table 9. Direct, indirect and total effects of model 1: procedural justice
Effect SE p LLCI ULCI
Direct effect c1’ .21 .03 <.01 --- ---
Total effect c1 .54 .03 <.01 --- ---
Boot SE Boot LLCI Boot ULCI
Table 10. Model 1 summary: interpersonal justice
Consequent
Trust (M) Repeat purchase intentions (Y)
Antecedent Coeff. SE p Coeff. SE p
Interpersonal justice (X) a1 .65 .02 <.01 c1’ .29 .03 <.01
Trust (M) --- --- --- b1 .47 .04 <.01
Gender -.04 .07 .55 .02 .07 .81
Age .01 .00 .04 -.00 .00 .09
E-commerce experience -.00 .01 .53 -.01 .01 .71
Online auction experience .01 .01 .21 .02 .01 .23
constant i1 1.52 .20 <.01 i2 1.95 .21 <.01
R2 = .49 R2 = .50
F(5,758) = 142.83, p = .00 F(6,757) = 126.61, p = .00
Table 11. Direct, indirect and total effects of model 1: interpersonal justice
Effect SE p LLCI ULCI
Direct effect c1’ .29 .03 <.01 --- ---
Total effect c1 .60 .03 <.01 --- ---
Boot SE Boot LLCI Boot ULCI
Table 12. Model 1 summary: informational justice
Consequent
Trust (M) Repeat purchase intentions (Y)
Antecedent Coeff. SE p Coeff. SE p
Informational justice (X) a1 .80 .02 <.01 c1’ .34 .04 <.01
Trust (M) --- --- --- b1 .44 .04 . <.01
Gender -.09 .06 .16 .00 .07 .95
Age .00 .00 .88 -.01 .00 .02
E-commerce experience .00 .01 .94 -.00 .01 .44
Online auction experience .00 .01 .71 .01 .01 .11
constant i1 1.21 .17 <.01 i2 2.10 .20 <.01
R2 = .59 R2 = .49
F(5,759) = 222.76, p = .00 F(6,758) = 122.40, p = .00
Table 13. Direct, indirect and total effects of model 1: interpersonal justice
Effect SE p LLCI ULCI
Direct effect c1’ .34 .04 <.01 --- ---
Total effect c1 .69 .03 <.01 --- ---
Boot SE Boot LLCI Boot ULCI
Table 14. Model 2 summary
Consequent
Trust (M) Repeat purchase intentions (Y)
Antecedent Coeff. SE p Coeff. SE p
Overall justice (X) a1 .79 .02 <.01 c1’ .46 .04 <.01
Trust (M) --- --- --- b1 .34 .04 <.01
Gender .01 .06 .16 .05 .07 .43
Age .00 .00 .83 -.00 .00 .06
E-commerce experience -.00 .01 .14 -.01 .01 .29
Online auction experience .01 .01 .78 .02 .01 .44
constant i1 .72 .18 .44 i2 1.67 .20 .06
R2 = .60 R2 = .53
F(5,759) = 223.38, p = .00 F(6,758) = 141.54, p = .00
Table 15. Direct, indirect and total effects of model 2
Effect SE p LLCI ULCI
Direct effect c1’ .46 .04 <.01 --- ---
Total effect c1 .73 .03 <.01 --- ---
Boot SE Boot LLCI Boot ULCI
