Expanding customer engagement: the role of negative engagement, dual valences and contexts

Purpose – The purpose of this study is to operationalise and measure the effects of negative customer engagement (CE) in conjunction with positive CE. Both valences are explored through affective, cognitive and behaviour dimensions, and, in relation to the antecedent of involvement and outcome of word-of-mouth (WOM). It also explores the moderating influence of service context by examining engagement within a social service versus a social networking site (SNS). Engagement with the dual focal objects of a service brand and a service community are also examined. Design/methodology/approach – Structural equation modelling is used to analyse 625 survey responses. Findings – Involvement is a strong driver of positive CE, and positive CE has a strong effect on WOM. These findings are consistent across the “brand” and “community” object, suggesting positive CE is mutually reinforced by different objects in a relationship. Positive CE is also found to operate consistently across the service types. Involvement is a moderately negative driver of negative CE, and negative CE is a positive driver of WOM. These relationships operate differently across the objects and service types. Involvement has a stronger inverse effect on negative CE for the social service, diverging from assumptions that negative CE is reflective of highly involved customers. Interestingly, negative CE has a stronger effect on WOM in the social service, highlighting the active and vocal nature of customers within this service context. Originality/value – To the best of the authors’ knowledge, this is the first paper to quantitatively measure positive and negative valences of engagement concurrently, and examine themoderating effect of dual objects across contrasting service types.


Introduction
Today's service environment is marked by dynamic change. Organisations are faced with increasing pressure to continually transform their practice in response to technological advancement, intensified global competition, and enhanced customer and stakeholder participation in service encounter. As a result, service relationships are no longer simple dyadic exchanges (Vargo and Lusch, 2008). They instead involve an ecosystem of interactions between engagement actors who are involved in a network of value creation (Chandler and Lusch, 2015;Anderson et al., 2013). This expanded network perspective offers service organisations' the opportunity to engage in a continuous process of value enhancement through interactive exchange (Chandler and Lusch, 2015). It also provides organisations with a greater understanding of the focal engagement objects that consumers interact with and the specific way in which these may contribute to a positive or negative service experience Jaakkola and Alexander, 2014;Hollebeek and Chen, 2014).
Customer engagement (CE) as an inherently interactive and reciprocal concept has become increasingly important as a mechanism by which to understand how service relationships are developed, maintained and enhanced (Chandler and Lusch, 2015;Jaakkola and Alexander, 2014). CE is defined as "a psychological state that occurs by virtue of interactive, co-creative customer experiences with a focal agent/object (e.g. a brand) in focal service relationships" (Brodie et al., 2011, p. 260). As a manifestation of the consumer's participation, connection and investment in a service organisation's offerings (Brodie et al., 2013;Hollebeek and Chen, 2014;Vivek et al., 2012), CE has been closely linked to organisational profit (Kumar et al., 2010), positive referral  and future purchase/usage intention .
To date, the literature has favoured research on positively valenced CE, which rewards organisations through a customer's affective commitment, brand equity, trust, self-brand connections, customer retention, loyalty, profitability and positive word-of-mouth (WOM) (Sashi, 2012;van Doorn et al., 2010;Bowden, 2009;Islam and Rahman, 2016). Yet, service relationships are not always positive as customers can be negatively engaged within a service relationship, and "negative brand relationships are in fact more common than positive relationships, with an average split across categories of 55 per cent/45 per cent" (Fournier and Alvarez, 2012, p. 253). Negative CE captures "consumers' unfavorable brandrelated thoughts, feelings and behaviours during focal brand interactions" (Hollebeek and Chen, 2014, p. 63). It has a detrimental impact on brand reputation and value through customers' negative WOM, brand switching, avoidance, rejection and potential retaliation and revenge behaviours (Hollebeek and Chen, 2014;Lievonen et al., 2017).
Recent research suggests that the intensity, velocity and direction of CE is not a static phenomenon (Pansari and Kumar, 2017;Palmatier et al., 2013). Service relationships may well contain the existence of both positive and negative manifestations of engagement (Hollebeek and Macky, 2019) and the extent to which engagement manifests from positive to negative expressions can significantly impact organisational performance (Palmatier et al., 2013). This interplay of valences may result in customer relationships fading or consequently terminating (Bowden et al., 2015). Engagement may also be differentially valenced (positive/negative) towards unique service elements in a relationship (Hollebeek and Macky, 2019;Bowden et al., 2017). Despite this, research on negative engagement remains nascent. There is, therefore, a pressing need to operationalise negative engagement including its dimensions and implications for service organisations, as well as the way it manifests towards focal objects within a service relationship (van Doorn et al., 2010;Hollebeek and Chen, 2014;Lievonen et al., 2017). The first contribution EJM 54,7 of this paper is to advance the nascent literature on negative engagement by operationalising and measuring negative CE through its affective, cognitive and behavioural dimensions.
Additionally, little is known of whether CE is consistent across service types or best when applied at a context-specific level (Islam and Rahman, 2016;Calder et al., 2016). The broad application of CE studies across commercial services has created an "urgent need in the service literature to account more fully for the influence of context and experience on CE" (Chandler and Lusch, 2015, p. 9). Research is required to expand conceptualisations of CE that are firstly, applicable to the social services, and secondly, transferable across diverse contexts (Islam and Rahman, 2016;Bowden et al., 2015). The second contribution of this paper is, therefore, to examine the contextual nature of engagement by examining CE across two contrasting service types, namely, a social service (Local Governments) and social networking sites (SNS) (Facebook, LinkedIn and Twitter) to explore whether engagement valences are context-specific.
The engagement literature to date has focused on single object research, which may "obscure the relevance of other objects" (Dessart et al., 2016, p. 400). It is suggested that differing valences of engagement may manifest differently towards multiple engagement objects within the one service (e.g. customers, brand communities, online platforms) and, that this duality may subsequently impact the overall relationship Dessart et al., 2015Dessart et al., , 2016Bennett and Bove, 2002;Li et al., 2017). The third contribution of this paper is, therefore, to examine the formation of engagement towards dual engagement objects of the service brand, and the service community. In doing so this paper examines the way, which positive and negative engagement manifest in relationship to differing engagement objects Brodie et al., 2013;Dessart et al., 2015Dessart et al., , 2016.
The paper proceeds as follows. We first present a detailed overview of nature, characteristics and dimensions of negative engagement providing the basis for the development of the conceptual model for the paper. We then provide a theoretical examination of the literature on involvement and its interconnection with both positive and negative engagement and their affective, cognitive and behavioural dimensions leading to the hypotheses presented in our study. This is followed by an examination of the relationship between positive and negative engagement and WOM recommendation and the respective hypotheses for these relationships. To test the empirical model and hypotheses, we employ structural equation modelling. We then compare the model across our two service types, namely SNS and local government services, as well as their respective dual engagement objects, namely the service brand and the service community, for each service.
Negative customer engagement Negative CE captures negative contributions and co-destruction within a service relationship (Plé and Cáceres, 2010;Dolan et al., 2016). Negatively engaged customers can act as brand adversaries who are highly committed to the relationship, yet in ways that detract value from the exchange (Hollebeek and Macky, 2019;Hollebeek and Chen, 2014;Dolan et al., 2016). This unfolds through a customer's intentional efforts to damage: a brand, customers or group of customers through the spread of negative WOM; co-opting others to adopt an attitudinal and/or behavioural position about a provider; brand switching, avoidance and rejection; and retaliation and revenge behaviours Hollebeek and Chen, 2014;Dolan et al., 2016;Juric et al., 2015). Negative CE is, therefore, important to identify and understand, as experiencing negative emotions, thoughts and behaviours can be distressing for customers, and the strong emotional dimension carries the Expanding customer engagement risk of spilling over to damage other customer segments Surachartkumtonkun et al., 2015;Juric et al., 2015).
Given that negative engagement has the very real potential for a destructive impact on service value (Hollebeek and Chen, 2014), negative engagement is of significant interest to academics and practitioners seeking to optimise engagement and restore positive engagement within the service relationship. This study contributes to the literature on negative engagement by operationalising negative engagement and its dimensions (van Doorn et al., 2010;Hollebeek and Chen, 2014). There is a need to test scales to capture its operation Pansari and Kumar, 2017). Further, it addresses the co-existence of positive and negative engagement within two types of service relationships (Hollebeek and Macky, 2019). Finally, it examines how negative CE may manifest through focal engagement objects "as a result of dynamic, iterative processes of relational exchange" (Juric et al., 2015, p. 281).

Developing a framework for negative engagement
Our study adheres to the tri-dimensional framework of negative CE (Hollebeek and Chen, 2014;Juric et al., 2015), however, we argue that negative engagement is not simply the reversal of positive CE, but manifests through its own unique characteristics (Juric et al., 2015). While negative CE may share the same drivers and dimensions (affect, cognition and behaviour) as positive CE, how these dimensions are measured and operate are ultimately distinct (Juric et al., 2015). This perspective is in line with Rissanen et al. (2016) and Dolan et al. (2016), who suggests positive and negative CE share the same high degree of involvement but are driven and manifest in different ways. The reasons for adopting this tridimensional framework to measure positive and negative CE are threefold. Firstly, the combination of these dimensions has, thus, far been the most widely-accepted across literature on positive (Dessart et al., 2015;Sim and Plewa, 2017;Leckie et al., 2016) and negative valences of CE (Hollebeek and Chen, 2014;Bowden et al., 2017). Secondly, affect, cognition and behaviour can vary in both valence (positive to negative) and magnitude (strong to weak) (Hollebeek and Chen, 2014;Dessart et al., 2015;Bowden et al., 2017). Thirdly, these dimensions are broad enough in scope to contain different subdimensions germane to the operation of each valence within the specific context in which they are applied (Dessart et al., 2015;Khuhro et al., 2017;Hollebeek and Chen, 2014).
Regarding the affective dimension of negative CE, we suggest this includes feelings of anger and dislike customers hold towards a service relationship. Anger is a strongly held negative emotion aroused by and directed at the misbehaviour of others (Bougie et al., 2003). Feelings of anger and hostility arise in response to unfulfilled service expectations when a customer's sense of autonomy and efficacy is violated (Smith, 2013;Park et al., 2013). Anger can be particularly detrimental to service organisations as it prompts customers to display punitive actions towards a brand, which aligns with the highly active nature of negative CE (Smith, 2013). For example, Lievonen et al.'s (2017) typology of negative stakeholder engagement features "revenge-seeking" and "trolls", who hold extremely strong, negative emotions towards an organisation. For revenge-seekers, these emotions prompt hostile thoughts, malice and brand sabotage, whereas trolls retaliate through spreading sadistic and often false claims about an organisation (Huefner and Hunt, 2000;Lievonen et al., 2017). Miller et al. (2012) also explore the emotional components of abusive relationships, which occur when customers feel powerless, under-valued and exploited; and adversarial relationships, which fester due to value incongruence and a strong hatred of the product/ service brand. EJM 54,7 The cognitive dimension of negative CE is suggested to be the degree of interest and attention paid to negative information about a service brand/community. This is in line with prior research finding negatively engaged customers dedicate higher levels of cognitive processing when reading, evaluating and reacting to negative brand information (Juric et al., 2015;Hollebeek and Chen, 2014). Frow et al. (2011) identify ten cognitive triggers that antagonise customers, which range from: provider dishonesty, information misuse and privacy invasion; to: unfair customer favouritism, misleading or lock-in contracts and exploitation. These triggers represent the more cognitive dimensions of negative CE, as they require evaluative judgement of one's treatment during a service encounter (Bowden et al., 2015;Hollebeek and Chen, 2014).
Finally, we posit the behavioural component of negative CE to manifest through collective complaint and anti-brand activism. Research on negative CE in public sector organisations suggests that negatively engaged stakeholders display their anger and dislike towards an organisation through public venting, boycotts and protests (Luoma-aho, 2015). This supports research, which finds that customers retaliate due to dissatisfying service experiences via negative WOM, personal attacks, demonstrations, e-mail campaigns or temporary boycotts (Romani et al., 2013;Huefner and Hunt, 2000). Appraisals of unfairness and dissatisfaction can prompt customers to then behaviourally display co-destructive behaviours, such as venting anger and frustration; spreading negative WOM; starting conflicts with members of a brand's online community; and creating alias online accounts to spread brand hatred using multiple profiles as a way of recruiting other members (Dolan et al., 2019;Gebauer et al., 2013;Smith, 2013).
In the mainstream CE literature, van Doorn et al. (2010) suggest the behavioural manifestations of negative CE include organising public actions against a firm and spreading negative WOM (e.g. blogging). More recent research focuses on typologies of CE behaviours segmented by intensity (low-high) and valence (positive-negative). For example, there is a growing body of research focusing on customers' negative behaviours ranging from "detachment" (low valence and intensity) through to "co-destruction" (high valence and intensity) (Dolan et al., 2016;Dolan et al., 2019). Lievonen et al. (2017) suggest that negative engagement behaviour may manifest through a range of behaviours including; negative voice, blaming, customer revenge and retaliation, justice-seeking behaviour, brand sabotage behaviour, co-opting others, anger activism, anti-brand sites, hate speech, trolling and hate holder behaviour. The behavioural dimension of negative CE is, therefore, suggested to have a collective element in that customers seek the support and reinforcement of others when complaining against a service brand/community. Conversely, the affective component of positive CE includes the feelings of pride, happiness, enjoyment, excitement and positivity customers experience during focal consumer/brand interactions Dessart et al., 2016). These positive emotions develop and transpire over the trajectory of a service relationship as opposed to discrete events (Dessart et al., 2016, p. 35). The cognitive dimension centres on a customer's positive mental states during and after interacting with engagement objects (Vivek et al., 2012;Dessart et al., 2016). This can include the level of positive attention, interest and reflection paid to a brand or its service community (Vivek et al., 2014;. The behavioural dimension captures a customer's level of participation, energy and passion towards various engagement objects (Vivek et al., 2012). It can also include the time spent using/interacting with an object and the degree to which they share and endorse this to others Dessart et al., 2016). We suggest that both positive and negative valences of engagement may occur within the one service relationship. The next Expanding customer engagement section presents our empirical model, and hypotheses. Our research model is presented in Figure 1.

Model development
The literature surrounding the antecedents and consequences of CE is disparate. Involvement and WOM are chosen as the focal antecedent and consequent of engagement respectively given their clear grounding in the literature (Islam and Rahman, 2016;Hollebeek and Chen, 2014;Verleye et al., 2014). This supports Hollebeek and Chen (2014), who conceptualise positive/negative CE to share common drivers, dimensions and outcomes, but operate as two extremes of a positive/negative continuum. While other relational drivers of CE have been explored within commercial service contexts, including satisfaction, identification trust, participation, self-brand congruity and flow Leckie et al., 2016;De Vries and Carlson, 2014), little is known of the drivers of CE within a social service (Bowden et al., 2016;Luoma-aho, 2015) and on social media platforms (Sashi et al., 2019). Involvement is chosen as the focal antecedent given its relevance and application to each engagement valence, and, in light of its role as a driver of negative (Rissanen et al., 2016;Hollebeek and Chen, 2014) and positive (Islam and Rahman, 2016;De Vries and Carlson, 2014) engagement in prior literature. Customers are also more likely to be cognisant of their involvement with service relationships compared with more abstract concepts of identity, self-brand congruity or flow (Harrigan et al., 2018;. For example, a positively engaged customer may experience high levels of involvement through depth of processing, emotional reactions and service usage, whereas a negatively engaged customer may experience the same, heightened levels of involvement but in a way that leads to the co-destruction of service value (Hollebeek and Chen, 2014;Rissanen et al., 2016;Harrigan et al., 2018;Islam and Rahman, 2016;Leckie et al., 2016).
Several outcomes of CE have also been proposed, including trust, loyalty, satisfaction, self-brand connections, affective commitment, brand value Pansari and Kumar, 2017;Maslowska et al., 2016;France et al., 2016). WOM is chosen as a more suitable and relevant outcome of CE given its applicability to negative (Azer and Alexander, 2018;  (Sashi et al., 2019) valences of CE. Customers may express their negative engagement through complaints, and efforts to warn other customers (Azer and Alexander, 2018;Juric et al., 2015), whereas positive engagement may be evidenced through recommendations, positive brand recounts and advocacy (Sashi et al., 2019).
Involvement as driver of positive and negative customer engagement Involvement has been noted as an important driver of positive and negative CE (Harrigan et al., 2018;Rissanen et al., 2016). Involvement is defined as a customer's "perceived relevance of the object based on inherent needs, values and interests" (Zaichkowsky, 1985, p. 342). Customers that are highly involved with a service relationship are more likely to be cognitively, emotionally and behaviourally invested with the brand (Harrigan et al., 2018;Bowden, 2009). The evidence for involvement as a driver of positive CE is sound, however, to date no studies quantitatively explore involvement as a driver of negative CE. Yet, several qualitative studies position negative CE to be driven by high levels of involvement (Rissanen et al., 2016;Hollebeek and Chen, 2014).
Customers who are highly involved in a relationship are more likely to have stronger reactions when service failures occur due to their inflated expectations, which if not met, can prompt strong negative emotional reactions (Lau and Ng, 2001;Gebauer et al., 2013). Therefore, heightened involvement can influence the affective aspect of negative CE, which manifests through emotions such as hate, contempt, dislike and resentment (Romani et al., 2013;Hollebeek and Chen, 2014). Conversely, involvement can influence the affective dimension of positive engagement, as involved customers are more emotionally bonded with service brands, and thus, demonstrate higher affective commitment to a service relationship (Leckie et al., 2016;Bowden, 2009). Involvement may emphasise negative cognitive reactions, as highly involved customers are more likely to seek, endorse, and pay attention to negative information about a focal service organisation (Kottasz and Bennett, 2014;Harrigan et al., 2018). This is also true for positive cognitive engagement, as customers are more likely to interpret and comprehend marketing communication favourably when they are highly involved with the service relationship (Leckie et al., 2016;Dwivedi et al., 2016).
Lastly, we expect involvement to drive the behavioural dimensions of CE. According to Harrigan et al. (2018, p. 389) the "intangible resources accrued via involvement, such as information, affiliation, and status may be motivation for consumers to engage with the brand". For example, involvement may motivate destructive behaviours as customers spread negative WOM and publicly complain to warn other customers, and, gain social reinforcement after service failures (Dolan et al., 2019;Hennig-Thurau et al., 2004;Dolan et al., 2016). Conversely, involved and positively engaged customers are more likely to interact in online brand communities, and co-create value during service encounters as these behaviours provide valuable resources through affiliation with others (Harrigan et al., 2018;De Vries and Carlson, 2014;Dwivedi et al., 2016). Considering the above, we suggest involvement to drive both positive and negative valences of CE, given the increased importance that service relationships, and their community, hold for highly involved and invested consumers.
We propose: H1. Involvement is a positive driver of positive CE.
H2. Involvement is a positive driver of negative CE.

Expanding customer engagement
Word-of-Mouth as an outcome of positive and negative customer engagement WOM is defined as "informal, person-to-person communication between a perceived noncommercial communicator and a receiver regarding a brand, a product, an organisation or a service" (Harrison-Walker, 2001, p. 63). The changing landscape of the customer experience has expanded WOM beyond physical conversations to encompass electronic word-of-mouth (eWOM), which is spread through a range of digital platforms including online communities and social media (Islam and Rahman, 2016;Hollebeek and Chen, 2014;Hennig-Thurau et al., 2004). Negatively engaged customers may be motivated to discuss their negative experiences to warn others, vent or gain revenge on a service organisation/community, whereas positively engaged customers may seek to share feelings of happiness and passion towards their service organisation (Dolan et al., 2019;Islam and Rahman, 2016;Hollebeek and Chen, 2014). Customers are more likely to perceive the choices and opinions of other customers as authoritative information, compared to brand-initiated communication (Azer and Alexander, 2018). Customers may abide by a shared responsibility of informing others to reduce service risk and alleviate customers' reliance on traditional marketing communications (Hollebeek and Macky, 2019;Azer and Alexander, 2018). Therefore, understanding how positive/negative CE drives WOM is of crucial importance (Harrison-Walker, 2001;Vivek et al., 2012).
Negative engagement has been found to result in a range of WOM activities including blogging, online reviews, complaints and negative recommendations (van Doorn et al., 2010;Azer and Alexander, 2018). These can be triggered by a range of factors, including poor customer service, dissatisfaction and unethical brand behaviour (Juric et al., 2015;Azer and Alexander, 2018). Interestingly, Hollebeek and Chen (2014) found negative CE to be a stronger driver of WOM compared to positive CE, highlighting the asymmetrical impact of negative versus positive engagement on brand outcomes. This may be due to the stronger emotional valence and subsequent contagion effect of negative CE, which including feelings of hate, dislike and contempt towards an organisation and its reputation can prompt negative WOM (Lau and Ng, 2001;Dolan et al., 2016). However, positively engaged customers also engage in WOM in ways that see them acting as unofficial "spokespersons" when broadcasting their positive experiences with others (Sashi et al., 2019;Pansari and Kumar, 2017;Vivek et al., 2012). For example, positively engaged customers recommend and praise favoured service relationships, and often encourage others to interact with the focal brand, as well as the wider brand community (Islam and Rahman, 2016;Vivek et al., 2012).
We propose: H3. Positive CE has a positive effect on WOM.
H4. Negative CE has a positive effect on WOM.
Moderating role of service context To date, research has applied engagement at a context-specific level, and little is known of the contextually contingent nature of engagement (Islam and Rahman, 2016;Vivek et al., 2012). There are calls for research to move away from exploring only those "extraordinary" events had by customers in commercial (usually hedonic) services, to consider the more mundane experiences customers have within the social and public sector . According to Hollebeek et al. (2016, p. 590), "the undertaking of academic engagement-based research across a broad range of online and offline environments is imperative". As such, there is an opportunity to consider CE in contexts where the option to "buy-into" the relationship is not given . This study explores EJM 54,7 negative and positive CE across two contexts, namely, social networking sites (SNS) and in contrast, a social service, which is operationalised as local governments. Social media has empowered customers to become "active collaborators in an interactive value formation process" (Dolan et al., 2019, p. 35). However, not all engagement on social media is positive and customers are increasingly turning to online platforms to express their negative brand opinions and experiences (Dolan et al., 2019;Hogreve et al., 2019). Prior research has found that negative CE is contagious within online platforms, which can damage an organisation's reputation and lead to distrust, switching behaviour and negative WOM Dolan et al., 2019;Hogreve et al., 2019;Rissanen et al., 2016;Juric et al., 2015). It has been found to manifest through negative engagement emotions such as feelings of irritation toward other social media users and imbalanced community intimacy (Azer and Alexander, 2018). Negative cognitive engagement is suggested to manifest through frustration with site functionality and time-consuming activities (Dolan et al., 2019;Azer and Alexander, 2018). Negative behavioural engagement manifests through valuedestructing behaviours such as, for example, public actions against a firm and spreading negative WOM (Azer and Alexander, 2018;Dolan et al., 2019).
In contrast to the autonomous and hedonic nature of social media networking sites is local government services. Local governments have "not received research attention in the service literature in proportion to their importance in people's lives [. . .] despite the fact that cities have always provided extensive community service systems" (Freund et al., 2013, p. 38). These services include: parks and recreation services; community safety; provision of local healthcare; youth and aged care; art and cultural services; local road maintenance; residential and commercial development services; and sanitary and waste services (Freund et al., 2013). Negative engagement in the public sector literature is defined as a "premeditated, activated and dedicated behaviour" (Bowden et al., 2016). It is found to manifest through protests, boycotts, negative WOM and revenge-seeking against government service providers (Bowden et al., 2016). As such it is defined as a: [. . .] goal-directed process that involves citizens' active and persistent expressions of negativity towards some aspects of their local government, which has a detrimental effect on the service relationship and the value derived from the relationship" (Bowden et al., 2016, p. 262).
Preliminary qualitative research has found that negative engagement within government services undermines the status, recognition and organisational legitimacy of a local government (Artist et al., 2012) reducing cohesiveness, participation and social capital among a community (Bowden et al., 2016;Luoma-aho, 2015).
The second contribution of this paper is, therefore, to examine the formation of positive and negative engagement towards differing service types, namely, social networking sites and local governments as two distinct service types. While SNS provides customers with highly customisable, flexible and hedonic service experiences, local governments are highly centralised, process-orientated and bureaucratic organisations that provide "one-to-many" services with little to no customisation (Freund et al., 2013;Kaplan and Haenlein, 2010). To date, no studies have quantitatively examined the manifestation of positive and negative CE in relation to significantly different service types. We suggest that the way customers engage with these types of services is governed by different dynamics that may moderate the process of positive and negative CE. We propose that the interrelationships between involvement, positive CE, negative CE and word of mouth will be positive within both an SNS and local government service context, but that differences may exist in the salience of these interrelationships. We examine the moderating effect of SNS and local government on Expanding customer engagement our empirical model by using a multi-group analysis of structural invariance to examine the differences between groups. We propose that: H5. The interrelationships between involvement, positive and negative engagement and WOM will be positive but moderated by service context.

Customer engagement across dual objects
In addition, recent research suggests that customers can engage with multiple focal engagement "objects", simultaneously within a service relationship Maslowska et al., 2016;Chandler and Lusch, 2015). Importantly, these objects (e.g. a focal service, consumer community, staff, service intermediaries) can exert different influences on the type of engagement experienced (Dessart et al., 2015(Dessart et al., , 2016Schamari and Schaefers, 2015). Yet, research largely adopts a single-object focus, and few studies consider the interplay of dual objects and how they affect CE valences or how dual objects influence positive and negative CE concurrently (Dessart et al., 2016(Dessart et al., , 2015Sim and Plewa, 2017;Bowden et al., 2017). Yet, it is recognised that customers engage with dual objects simultaneously in ways that create and diminish service value (Hollebeek and Macky, 2019;Chandler and Lusch, 2015;Dessart et al., 2015Dessart et al., , 2016. The focus on single object research may "obscure the relevance of other objects, casting doubt on the validity of the research models" (Dessart et al., 2016, p. 400), and there is a need to understand dual engagement objects (e.g. customers, brand communities, online platforms) as the nature and quality of a customer's interaction with these foci can significantly impact the overall relationship (Dessart et al., 2015(Dessart et al., , 2016Bennett and Bove, 2002;Li et al., 2017). is interrelated with their engagement with the focal brand. Specifically, a "spillover" effect existed whereby customers' positive CE with their OBC further enhanced their engagement with the brand, yet, their negative CE within the OBC detracted from brand engagement. This highlights the fluidity of CE, as the type of CE experienced in one interaction can spill over to add or detract value from their engagement with other objects . Importantly, it reveals that customers may be negatively engaged with an object (i.e. brand community) irrespective of their positive engagement with the service brand overall. The third contribution of this paper is, therefore, to addresses the gap in the literature concerning the formation of engagement towards dual objects of the service brand and the service community. In doing so, this paper examines the way in which positive and negative engagement manifest in relationship to differing objects Brodie et al., 2013;Dessart et al., 2015Dessart et al., , 2016: H6. The interrelationships between involvement, positive and negative engagement and WOM will be positive but moderated by service object.

Method
An online, dual-focus self-administered voluntary survey was distributed to Australian service consumers. A total sample of 625 participants was achieved across the two contexts and the respondent criteria were as follows: approximately equal numbers male and female, EJM 54,7 aged over between 18 and must live in an Australian local government area or for the social media sample, must have used either Facebook, LinkedIn or Twitter at least once in the past two weeks. Roughly equal quotas were achieved for the two service types (325 respondents for the local government service and 300 respondents for the social media context), and within each context, equal quotas were achieved for the two objects due to the questionnaire being repeated per respondent. Therefore, 325 respondents completed the survey for both the community and brand object for the local government services and 300 respondents completed the survey for both objects in the social media context. The social media context experienced a skew towards Facebook as the self-selected platform choice. This skew was somewhat expected, as Facebook is the preferred SNS for customers to engage with brands (Dessart et al., 2016). The respondent profile can be seen in Table I. Involvement, positive CE and WOM were all measured using scales that have been validated within existing CE literature. Involvement was measured using Zaichkowsky's (1985) bipolar semantic differential scale, which captures the relevance and importance of a brand, object or relationship to the customer  WOM was measured using Harrison-Walker's (2001) 'word-of-mouth activity' scales, which measure the frequency with which customers mention and discuss a brand with others (Vivek et al., 2014). Positive CE was measured using a hybrid of scales from seminal CE studies by  and Vivek et al. (2014). The affective dimension was captured using Hollebeek et al.'s (2014) "affection" scales, and the cognitive and behavioural dimensions were measured using Vivek et al.,'s (2014) "conscious attention" and "enthused participation" scales, respectively. The scales for negative CE were created from existing scales in related areas of marketing literature. The affective dimension was measured using Romani et al. (2013) scales on "anger" and "dislike", the cognitive dimension was captured using Kottasz and Bennett's (2014) "depth of processing" scales, and the behavioural dimension measured using Romani et al. (2015) "anti-brand activism" scale.

Validation of measures
All measures were subjected to exploratory factor analysis, which was conducted in SPSS 24.0. Cronbach's alpha was examined (Hair et al., 2006) along with average variance extracted. Construct validity was established through tests of discriminant and convergent validity. Convergent validity was established for all constructs, which had parameter estimates above the 0.50 criterion, and displayed significance at the 6 1.645, p > 0.05 level. In addition, the average variance extracted estimates demonstrated that the measurement scales accounted for a greater proportion of explained variance than measurement error as the AVE statistics were above the >0.50 criterion value. Discriminant validity was examined according to Fornell and Larcker's (1981) stringent tests to establish separation between latent constructs. Discriminant validity was established for all latent constructs across all four samples as per Appendix 2. The bivariate correlations and descriptive statistics can be found in Appendix 3. The correlations ranged from positive to negative, (À0.223) to (0.635), with the lowest correlation (À0.001). Some correlations failed to be significant at the 0.01 level. The data analysis then followed the two-step procedure recommended by Anderson et al. (1987) including estimation of the measurement model followed by estimation of the structural model. The additional confirmatory factory analysis was undertaken in AMOS prior to testing the structural model. The measurement model indicated good fit and all items retained served as strong measures for their respective constructs (x 2 = 352, df = 153, p 0.000, RSMEA = 0.064, GFI = 0.90, CFI = 0.96, IFI = 0.96). An RSMEA under 0.07 is considered to be good model fit by Steiger (2007) and Hooper et al. (2008).

Results
The hypothesised relationships in the model were tested using structural equation modeling. Goodness of fit statistics indicated that a reflective, second-order model fitted the data adequately (GFI = 0.937, CFI = 0.968, IFI= 0.968. RMSEA = 0.06). A competing model was tested to observe the direct effect of involvement on WOM. This resulted in moderate . Therefore, the first model where involvement has an indirect effect on WOM through positive and negative CE demonstrates better fit and is retained for this article. This is also supported theoretically, as involvement measures the degree of relevance and importance a service holds to a consumer, whereas engagement entails more interactive, collaborative and passionate dimensions, which would prompt customers to discuss their service experience with others ( The squared multiple correlation for the structural equation is 0.610. (Figure 2) A second purpose of this paper is to understand the potential moderating effect that different service contexts have on the salience of the constructs and to compare these across two object types, which explored in H5 and H6, respectively. An unconstrained baseline model both within and between each context is established. To test the invariance (equal Expanding customer engagement weights) across the service brand and community objects, a multigroup analysis of structural invariance is conducted (Byrne, 2004). The Chi-square statistic was used to assess measurement invariance, that is, if the chi-squared difference was not significant between the unconstrained and constrained model then measurement invariance was established. The results of the invariance tests are presented in Table II.
Measurement invariance was established across all groups. This indicates that the meaning of the latent constructs in the model is similar between the engagement objects within each context and between the same object across contexts.
Having satisfied the conditions necessary at the measurement level we then proceeded to test for structural invariance (Byrne, 2004). The analysis indicated three of the four comparisons were not invariant at the structural model level. The only model to have a nonsignificant difference (p = 0.38) was within the SNS context comparing across the "brand" and "community" objects [ Table III panel (b)]. The results of this analysis are shown in Table III (panels a, b, c and d). The analysis indicated that involvement had a consistent effect as a driver of positive CE across the dual objects within the social service and the SNS. Positive CE had a consistent effect as a driver of WOM across the dual objects within each service. The process of positive CE was also consistent when comparing the "brand" object in the social service with the "brand" object for the SNS, and when comparing the "community" object across both service types. This suggests that the process of positive CE operates the same way same regardless of object or service type.
Involvement also had a consistent effect as on positive negative CE across the dual objects within the social service and the SNS. Negative CE also had a consistent effect as a driver of WOM across the dual objects within each service. However, the effect of involvement on negative CE was stronger for the social service context compared to the SNS. That is, involvement was a stronger (negative) driver of negative CE for the "brand" and "community" object within the local government service context compared with SNS. Additionally, the effect of negative CE on WOM was not consistent across the contexts when comparing the 'brand' object of the social service with the "brand" object of the SNS. Negative CE was a stronger driver of WOM towards the "brand" object in the social service compared with the SNS.

Discussion and implications
The literature on consumer engagement to date has predominantly focussed on an examination of positive manifestations of engagement. Yet, research has demonstrated that negative brand relationships are in fact more common than positive relationships (Fournier and Alvarez, 2012). From this perspective, expressions of engagement may be both positive or negative depending on the prevailing relationship climate and conditions (Hollebeek and Chen, 2014;Naumann et al., 2017) significantly impacting upon organisational performance (Palmatier et al., 2013;Bowden et al., 2015). This study contributes to the nascent literature on negative engagement through an examination of its dimensions, operationalisation and expression within contrasting service types, and its manifestation towards specific focal engagement objects within a service relationship (van Doorn et al., 2010;Hollebeek and Chen, 2014;Lievonen et al., 2017).
The first theoretical contribution of this study relates to the conceptualisation, operationalisation and role of negative CE. The findings for negative CE are complex and contribute new insights into its operation. Involvement was found to have a weak, negative effect on negative CE, and negative CE holds a positive relationship with WOM. The finding of involvement as a "non-driver" of negative CE diverges from existing theoretical assumptions that claim negative CE is displayed by highly involved and invested customers x 2 , df/x 2 , p-value reported from measurement weights assuming unconstrained model to be correct   Table III.
Expanding customer engagement (Rissanen et al., 2016;Islam and Rahman, 2016). While it is plausible that highly involved customers would be more motivated to display negative CE in response to service failures, the findings of this article suggest involvement is not a relevant driver, and further investigation is needed of antecedents of negative CE.
Conversely, negative CE has a moderately positive influence on WOM, which supports prior literature claiming negative CE to result in highly active cues (Rissanen et al., 2016;Islam and Rahman, 2016). This further confounds the finding of involvement as a non-driver of negative CE, as customers evidently experience a degree of emotional, cognitive and behavioural engagement with the service that prompts them to discuss their concerns with others. Regardless, finding negative CE to drive WOM is important, as customers now have more power to influence those in their social network, especially when aided by social media platforms.
In addition, this study offers several new theoretical insights into the dual existence of positive and negative valences of CE. The results of the overall research model find involvement to be a strong driver of positive CE, which supports prior research finding customers with higher levels of interest and involvement in a relationship to be more positively engaged So et al., 2016). Involvement is one of the most conceptually relevant antecedents of positive CE, thus, the results of this article provide much needed empirical evidence of its role in motivating positive CE. While WOM is conceptually explored as a key outcome of positive CE, empirical evidence to support this relationship is limited. The findings of this article support Islam and Rahman (2016) and Hollebeek and Chen (2014) by revealing positive CE to have a strong, positive impact on WOM, suggesting positively engaged consumers are motivated to discuss and share their service experiences with others. This finding is important, as recent literature considers WOM to be a more significant outcome of engagement in public and health services (van Doorn et al., 2010;Verleye et al., 2014). The overall research model, therefore, sheds new theoretical light on the nuances of positive and negative CE. While previous research has framed these valences as "two sides of the same coin", this article provides an alternative view that sees positive and negative CE as more unique and individual concepts requiring different drivers and outcomes. Involvement had markedly different effects on positive/negative CE, suggesting it cannot be used as a universal driver for both valences. This supports recent research by Mittal et al. (2018, p. 189) that claims, "an asymmetry exists between the negative and the positive aspects of engagement". Further research is needed into the drivers of negative CE and should examine constructs that are neutral in valence (e.g. relating to the customer's' level of time an investment in the relationship) and negatively valenced (e.g. frustration and irritation). Exploring a variety of neutral and negatively valenced drivers may help clarify whether positive/negative CE can share a universal antecedent(s) or if they should be conceptualised as their own unique processes.
The second theoretical contribution of this study is in its contribution to an understanding of the way in which engagement manifests towards multiple focal engagement objects through a dual-focus scale capturing positive and negative object engagement. As per Table III (panel a) and (panel b), the relationship between involvement and positive and negative CE, and their impact on WOM, is consistent across each object when compared within each service context. While prior qualitative research finds the nature of CE to differ across objects (Dessart et al., 2015(Dessart et al., , 2016Roskruge et al., 2013) our findings reveal positive and negative CE to be more holistic and consistent across each object type. This finding aligns with recent research by Bowden et al. (2017) that finds a "spillover" between customers' positive/negative CE with online brand communities, and, the focal brand. This EJM 54,7 suggest that consumers' interactions with different objects may mutually reinforce their positive/negative CE, as consumers view their engagement with each object considering their aggregate experiences with the service relationship.
The third theoretical contribution of this study relates to the contextual application of CE through a cross-examination of CE across a social service (Local Government); and social media SNS platform. This provides much-needed insight into how context-specific factors affect the development of positive and negative CE across a range of service types with varied organisational structures and market environments .
Negative CE is revealed to exhibit context specificity. For the social service, involvement had an inverse effect on negative CE. This may be attributed to the "choice-constraining" contextual factors present in local governments, which can reduce the level of voluntarism and participation typically associated with CE in commercial contexts . Given the high barriers to exit, local government customers may exist in a "masterslave" type of relationship whereby they become negatively engaged when they cannot exit due to social, economic or legal barriers (Miller et al., 2012). As such, negative levels of involvement may exacerbate negative CE within these types of service relationships. Conversely, the operation of positive CE was consistent across both contexts, which suggests that the way in which customers develop positive engagement is similar across a range of service types. These findings highlight the salience of involvement as a key driver of positive engagement, suggesting customers' degree of personal relevance and interest in a service category serves as strong motivation behind engagement.
Negative CE was also found to have a positive influence on WOM, supporting prior research finding negative CE to result in complaint behaviour and negative recommendations (Hollebeek and Chen, 2014;Juric et al., 2015). This suggest customers will discuss their local government experiences with others despite having negative levels of involvement with the relationship. One explanation for this contrasting result is that directing anger and dissatisfaction towards social services can provide an avenue for social bonding and identification among customers (Luoma-aho, 2009). Further, customers are more likely to attribute blame to organisations that have a high degree of control over service design and delivery, such as highly bureaucratic or monopolistic services (Bougie et al., 2003). This, in turn, may heighten customers' willingness to firstly assign blame to their local government, and then discuss their negative service experiences with others.
Within the social media context, involvement has no effect on negative CE for either the brand or community object. These findings may also be attributed to contextual factors surrounding SNS. Namely, Facebook, Twitter and LinkedIn operate as voluntary "opt-in" services that allow customers to forge social and professional connections, create and share content, self-brand and discuss shared interests with others (Miller, 2017;. As such, customers may join SNS already having a positive disposition towards the relationship. This contrasts with the social service, whereby customers are forced into the exchange regardless of whether they perceive a need or want for it. Further, once customers join SNS, the types of interactions they have on these platforms largely revolve around creating and maintaining a sense of belongingness with others as opposed to engaging in oppositional discourse (Miller, 2017;Kaplan and Haenlein, 2010). This is because customers mainly use SNS to interact with like-minded people, and seek the approval and reinforcement of others (Miller, 2017). As such, users of SNS often exist in a "bubble" whereby the content they are exposed to simply reinforces their existing perceptions and attitudes. Therefore, while customers may discuss and share content that causes negative emotional, cognitive and behavioural reactions, the function of SNS means that interactions that foster solidarity and belongingness naturally take precedence over ones that spark Expanding customer engagement negative discourse and reactions (Miller, 2017). Considering this, finding involvement to have no effect on negative CE seems plausible as customers are unlikely to become negatively engaged due to heightened involvement on these platforms.
Negative CE had a moderately positive influence on WOM for both objects in the social media context, suggesting that customers were only mildly motivated to discuss their negative thoughts, feelings and behaviours towards these platforms or the communities they host with others. This may also be due to the nature of SNS, which customers use on a personal and autonomous basis. Unlike the social service context, customers have the autonomy to control and limit their exposure to negative experiences while using SNS. For example, Facebook users can "hide" pages that cause negative reactions, whereas Twitter users can "unfollow" accounts. Thus, even when customers encounter distressing or frustrating content while using these platforms, their ability to control the duration and exposure to this may lessen their motivation to discuss it with others. This contrasts with the social service context, whereby customers are essentially trapped, and may, thus, turn to WOM to cope with their negative engagement towards the service relationship.

Managerial implications
Several managerial implications can also be drawn from the findings. Firstly, this research provides strategic insight into the drivers and outcomes of positive and negative engagement. The overall results for negative CE suggest service managers should aim to lessen and contain this segment to prevent the further co-destruction of service value. While investing resources into negatively engaged customers may appear counterintuitive, it is important given the strong relationship between negative CE and WOM. That is, negative CE is not experienced alone but shared via customers discussions with others. Therefore, focussing on containing negative CE is important, as it entails strong negative emotions and collective complaint behaviour (e.g. blogging, public activism), which carries the risk of creating a "contagion effect" onto other customer segments. This may be done through: identifying the segment; understanding the motivations for their negative CE; responding to areas of service failure as specifically as possible through creating an open and transparent dialogue; and ultimately, trying to re-involve these customers. While challenging, attempting to involve these customers in positive service encounters is crucial, as negative CE can damage organisational reputation and create a negative perception among customers.
The overall findings for positive CE suggest service managers should encourage and reward customers for their involvement not only with the host organisation but also within customer-managed communities. Although community interactions do not directly involve the host brand, service organisations should still play a role in facilitating and rewarding customers' engagement with others. In turn, service organisations are more likely to be rewarded via customers' advocacy and frequent discussion of their positive brand/ community experiences with others. Generating positive WOM is particularly crucial for service organisations, as customers use the recommendations and experiences of others to frame their own expectations. Finding the process of positive CE to be consistent across diverse service objects and contexts also benefits service managers, who can strategise for positive CE in a uniform way across dual engagement foci, and draw from a multitude of service contexts when creating their own strategies for fostering a positively engaged customer base.
Our findings also highlight the need for service managers to monitor and manage the interactions their customers have across multiple touch points in their service relationships. Both positive and negative CE are found to be consistent across the dual "brand" and EJM 54,7 "community" objects within each service type. While this benefits service firms when positive CE is reinforced between these dual objects, it is also detrimental when the effects of negative CE towards the brand object spill over to diminish their engagement with their community object, and vice versa. Service managers should, therefore, consider how the multiplicity of objects affect customers' engagement overall.
Finally, this study offers strategic insight for managers in two diverse service environments. A negatively engaged customer base can undermine the status, recognition and organisational legitimacy of a social service organisation. Service managers should be cognisant of the effect that negative involvement can have on driving negative CE towards both the organisation and community object. Social services should emphasis their role as relevant, useful and important services to try and re-involve their negatively engaged customers. This is especially crucial for monopolistic or forced services, as bureaucratic constrictions and regulations can heighten customers' negative reactions towards a service relationship. Services of this nature must invest resources into managing the negatively engaged segment, as these customers are unable to "exit" the relationship in the same way they would a commercial service. Further, negative CE has the strongest effect on WOM for the social service "brand" object, suggesting that customers are highly motivated to discuss their negative experiences with the organisations in this context. As such, social services may be particularly vulnerable to not only harboring a trapped negatively engaged customer base but also being exposed to the detrimental effects of these customers' WOM, which can greatly damage organisational reputation and may influence other customer segments. Within the social service context, service managers should focus on promoting and rewarding involvement to drive positive CE, as positively engaged customers are more likely to be supportive of the planning, strategy and creation of municipal services. Further, positively engaged customers are more likely to spread WOM about their brand/community experiences, which can help reinforce a sense of shared identity, satisfaction and happiness among a municipal area.
Managers of SNS should be cognisant of the strong effect that positive CE has on WOM for both the host brand and community object. Strategies should focus on facilitating and encouraging customers' involvement in online communities not only to maintain their positive engagement but to also encourage customers to attribute the social bonds and connections they form while interacting with these communities to the host brand. Although object type was not found to moderate negative CE within the SNS context, a stronger relationship was found between negative CE and WOM for the community object. As such, host platforms might consider adopting a greater role in monitoring and observing the nature and valence of customers' interactions and engagement within their communities. Managers of SNS should be especially cognisant of the effect of negative CE and WOM, as the detrimental effect of WOM can be compounded by the ease and speed in which negative information can be disseminated to a wide audience on virtual platforms.
In summary, this study has contributed to several important areas of the engagement literature by providing a multi-valenced, dual object and cross contextual exploration of CE. Given that no empirical investigations into negative CE exist, this study also represents a major contribution to the nascent literature on negatively valenced engagement. The findings of this article highlight the need for CE research to continue to progress a more expanded perspective that considers the broader range of service contexts, objects and dimensions involved in the process of CE.

Limitations
The findings of this article should be considered with several limitations in mind. Firstly, this study provides an empirical exploration into positive and negative CE within local Expanding customer engagement government and social media services. Future research could expand the contextual application of CE to cross-examine its operation in a wider range of service types. Future research may consider a wider range of object types, even extending beyond a dual focus to capture customers' engagement with multiple (three, four, etc.) objects simultaneously. This may better reflect the dynamic nature of the service ecosystem, which involves customers interacting with several actors within the focal service relationship. Future research should explore whether the WOM resulting from customers' positive engagement has ongoing effects on a service relationship. For example, WOM may exist in a "feedback" loop to drive and/or reinforce positive CE for new or existing customers. This may be especially important within online platforms, whereby customers act an unofficial "advocates" and can, thus, influence perceptions throughout their social networks. To this end, adding a "social" dimension to the existing tri-dimensional CE framework may better capture the type of engagement experienced during customers' interactions with other actors in their service ecosystem. This may be particularly important when exploring CE in both a social service and online social networking site whereby customers use their service experiences as a means of validation, connection and social enhancement. Future research may consider how involvement works in conjunction with other complimentary antecedent factors to reinforce a customer's positive engagement, such as category knowledge and participation. In addition, future research may need to consider drivers that encompass aspects of a service interaction that can lead to negative CE, such as time, degree of personal contribution to the service, attachment, irritation and community intimacy (Heinonen, 2017;Palmatier et al., 2017). Future research also should extend the contrasting service types used in this study to explore the role of involvement as a motivation behind positive CE across a range of both forced and "opt-in" service types. To this end, future research should also consider whether a feedback loop exists where WOM drives or reinforces negative CE for new and/or existing customers. This is especially important within SNS, which can exacerbate dysfunctional and harmful customer behaviours.