Negative brand engagement represents a pervasive and persistent feature of interactivity in online contexts. Although existing research suggests that consumer negativity is potentially more impactful or detrimental to brands than its positive counterpart, few studies have examined negative brand-related cognitions, feelings and behaviours. Building on the concept of brand engagement, this study aims to operationalise negative online brand engagement.
This paper presents the results of nine studies that contributed to the development and validation of the proposed scale. Building on the concept of engagement, Studies 1–3 enhanced the construct conceptualisation and generated items. Study 4 involved validation with an academic expert panel. The process of measure operationalisation and validation with quantitative data was completed in Studies 5–8. Finally, the scale's nomological validity was assessed in Study 9.
The results confirm the multidimensional nature of negative online brand engagement. The validated instrument encompasses four dimensions (cognition, affection, online constructive behaviour and online destructive behaviour), captured by 17 items.
Progress in understanding and dealing with negative online brand engagement has been hampered by disagreements over conceptualisation and the absence of measures that capture the phenomenon. This work enhances managerial understanding of negativity fostering strategies that protect brand engagement and improve firm performance.
1. Introduction
In recent years, engagement has emerged as an important concept in marketing, garnering increasing research interest (Hollebeek et al., 2022). While definitions and perspectives vary, the dominant conception asserts that consumer engagement has an interactive nature and is a context-dependent construct encompassing consumers' cognitive, affective and behavioural investment in specific interactions with a focal engagement object (Hollebeek et al., 2023). Brands are a major focal object of engagement (Algharabat et al., 2018; Wang et al., 2022) in various online and offline contexts, (Dessart et al., 2019; Hollebeek, 2011a, 2011b; Hollebeek et al., 2022; Morgan-Thomas et al., 2020). Brand engagement may take in a bipolar fashion (Naumann et al., 2020). Positive brand engagement occurs when consumers' brand-related cognitions, affections and behaviours are favourable to the brand (Dessart et al., 2016), while negative brand engagement denotes unfavourable or disapproving sentiments and behaviours (Do et al., 2020; Hollebeek and Chen, 2014; Stathopoulou et al., 2017). Negative brand engagement offers unique insights into consumer–brand interactions that are undesirable and harmful for brands (Hollebeek et al., 2022, 2023).
Contrasted with significant interest in positive engagement (Cao et al., 2021; Hollebeek et al., 2022; Tuškej and Podnar, 2018), few studies explore negativity (Azer and Alexander, 2020a, 2020b; Bitter and Grabner-Kräuter, 2016; Rahman et al., 2022). Yet, compared with positive brand engagement, negativity can be more widespread (Rissanen and Luoma-Aho, 2016), potentially more impactful (Bowden et al., 2017; Rodrigues and Pinto Borges, 2021) and detrimental to brands and consumers (Naumann et al., 2020), particularly in the interactive online environment (Hollebeek et al., 2022). Consumers may develop various negative brand feelings, such as anger and frustration (Naumann et al., 2020). Their interactivity may include behaviours, such as sharing regret and publically deriding (Azer and Alexander, 2020a), harming brand performance (Naumann et al., 2017a).
The connectivity of the online context, particularly social media (Schultz and Peltier, 2013; Barger et al., 2016), fosters online brand engagement by affording communication and interaction at scale (Baldus et al., 2015; Parihar and Dawra, 2020). Contrasted with other brand-related contexts (Morgan-Thomas et al., 2020), online brand engagement is generally characterised by higher levels of participation (Swaminathan et al., 2020) and engagement interactivity (Hollebeek et al., 2023). In essence, the online environment fosters new kinds of engagement practices affecting the phenomenon's nature and essence (Hollebeek et al., 2014, 2023; Morgan-Thomas et al., 2020). Negative engagement in online environments, such as social media, is widespread (Dessart et al., 2020; Liao et al., 2023) and has an idiosyncratic nature (Lievonen et al., 2022).
Despite some progress, several gaps persist in the emergent literature on negative engagement. Most existing conceptions of negativity are either theoretical (Do et al., 2020) or qualitative (Heinonen, 2018), highlighting difficulties in drawing conceptual boundaries or empirically capturing the phenomenon (Bowden et al., 2017). Although confirmatory research on negative engagement has begun to emerge (Azer and Alexander, 2020a, 2020b; Bitter and Grabner-Kräuter, 2016; Kulikovskaja et al., 2023; Naumann et al., 2020; Obilo et al., 2021; Rahman et al., 2022) efforts to capture negative engagement have tended to rely on either proxies (Bitter and Grabner-Kräuter, 2016; Arora and Chakraborty, 2021; Labrecque et al., 2022) or partial measures (Azer and Alexander, 2020a; 2020b). Despite its profound consequences (Baldus et al., 2015; Kumar, 2021, 2022), little is known about the actual domain and boundaries of the concept. Very few studies have robustly conceptualised it and disagreements persist with some authors highlighting the behavioural dimension (Dolan et al., 2016) and others suggesting multiple dimensions and incorporating consumers' unfavourable brand-related thoughts, emotions and behaviours (Do et al., 2020). As recent calls for further research illustrate (Hollebeek et al., 2023; Naumann et al., 2020), scholars are yet to adequately define the concept in general (Bowden et al., 2017; Hollebeek and Chen, 2014) and, specifically, to acknowledge variations with reference to particular objects (i.e. the brand) or contexts (e.g. online).
To address these gaps, this study aims to refine the definition of negative online brand engagement and develop a scale to capture it. Building on research on both positive and negative engagement, the study deploys a robust scale development process to offer an empirical operationalisation. The adopted four-step procedure reports the negative online brand engagement definition and item generation (Step 1), item purification (Step 2), reliability and validity (Step 3) and nomological network and discriminant validity (Step 4).
The study's contribution is threefold. Theoretically, the proposed conceptualisation and operationalisation of negative online brand engagement enhances understanding of negativity and its relationship with other concepts, namely brand disloyalty and happiness. This concurrently addresses extant gaps in the body of knowledge (Heinonen, 2018), calls for theoretical rigour (Hollebeek et al., 2023) and paves the way to future research on negativity (Hollebeek et al., 2022, 2023) contributing novel insights into the emerging negative engagement literature in marketing (Hollebeek and Chen, 2014; Naumann et al., 2020). Practically, this work enhances managerial knowledge regarding strategies to control negative effects and improve firm performance (e.g. sales growth and superior profitability) (Hollebeek et al., 2014; Rahman et al., 2022).
2. The need for new conceptualisation and operationalisation of negative online brand engagement
Multiple reasons justify the need for refinement of current approaches to negative online brand engagement. The research on consumers' negative engagement in marketing is, only now, emerging and limited (Table I), and past studies tend to be conceptual (Do et al., 2020) or qualitative (Heinonen, 2018). Only rarely do the definitions refer to the negative nature of the engagement (see Table I for exception) and they tend to neglect the differences between online and offline contexts (Naumann et al., 2017a, 2017b, 2020).
Existing definitions of negative engagement with various objects
| Articles | Paper type | Objects | Construct | Context | Definitions |
|---|---|---|---|---|---|
| Van Doorn et al., 2010 | Conceptual | Brand | Customer engagement behaviour | N/A | Customers' behavioural manifestation toward a brand or firm, beyond purchase, resulting from motivational drivers |
| Hollebeek and Chen, 2014 | Qualitative | Brand | Brand engagement | Online | Consumers' unfavourable brand-related thoughts, feelings and behaviours during brand interactions |
| De Villiers, 2015 | Qualitative | Brand | Consumer brand engagement | Online | Consumers' negatively valenced cognitions, emotions and behaviours toward the brand, which can be active or passive |
| Dolan et al., 2016 | Conceptual | Brand | Social media engagement behaviour | Online | Consumers' unfavourable brand-related behaviours during interactions |
| Bitter and Grabner-Kräuter, 2016 | Quantitative | Brand | Customer engagement behaviour | Online | Behavioural manifestations of customer engagement on social networking sites |
| Rissanen and Luoma-Aho, 2016 | Qualitative | Organisation | Consumer engagement | Online | Negative behavioural manifestations such as protests and sharing negative information about the organisation |
| Naumann et al., 2017a | Qualitative | Service provider, community | Consumer engagement | Offline | Consumers' strong negative thoughts, feelings and behaviours towards their service provider |
| Naumann et al., 2017b | Qualitative | Organisation, community | Customer engagement | Offline | A negatively valenced manifestation of engagement consisting of cognitive, emotional and behavioural components |
| Bowden et al., 2017 | Qualitative | Brand, brand community | Consumer engagement | Online | A consumer's negatively valenced cognitive, emotional and behavioural investments during or related to interactions with focal objects or agents |
| Azer and Alexander, 2018 | Qualitative | Brand | Customer engagement behaviour | Online | Negative engagement behaviours include discrediting, deriding, expressing regret, endorsing competitors, dissuading and warning |
| Heinonen, 2018 | Qualitative | Interests | Consumer engagement | Online | Community members' cognitive, emotional and behavioural investments in a specific area of interest |
| Do et al., 2020 | Conceptual | Brand | Customer engagement behaviour | Offline | A customer's unfavourable thoughts, feelings and behaviours towards a service brand or provider result from negative critical events that cause perceived threats to customers |
| Naumann et al., 2020 | Quantitative | Brand, brand community | Customer engagement | Online, offline | Consumers' unfavourable thoughts, feelings and behaviours towards the dual focal objects |
| Azer and Alexander, 2020a | Quantitative | Service provider | Negative customer engagement behaviour | Online | Customer contributions of resources such as knowledge, skills, experience and time negatively affect other actors' knowledge, expectations and perception about a focal service provider |
| Azer and Alexander, 2020b | Quantitative | Service provider | Negatively valanced engagement behaviour | Online | Customers beyond the transactional negative behavioural manifestations |
| Rahman et al., 2022 | Quantitative | Brand | Negative customer engagement | Online | Customers' motivation to invest time and resources to bring disappointing service experiences to the attention of relevant authorities in the form of formal complaints to negatively affect other actors' service perception about the firm in question |
Source: Authors' own work
To date, a handful of quantitative papers (Azer and Alexander, 2020a, 2020b; Bitter and Grabner-Kräuter, 2016; Kulikovskaja et al., 2023; Naumann et al., 2020; Rahman et al., 2022) have focused on consumers' negative engagement. The concept has been typically captured by proxy measures, for example, posting behaviour including negative posts (Bitter and Grabner-Kräuter, 2016; Arora and Chakraborty, 2021; Labrecque et al., 2022) or the number of complaints (Rahman et al., 2022). Others have used experimental designs with stimuli to demonstrate negative behaviours and investigate consumers' reactions to this material (Azer and Alexander, 2020a, 2020b), omitting the negative engagement itself. One study offered a more holistic approach and deployed an adapted scale that included items from a variety of sources but lacked thorough conceptualisation, development or validation (Naumann et al., 2020).
Only one study developed a psychometric measure of negative engagement. Obilo et al. (2021) have suggested a scale encompassing four dimensions, two on engagement activities (content engagement, co-creation) and two on engagement valence (advocacy and negative engagement). Although useful, the instrument builds on a behavioural tradition conceiving engagement as behaviour (Algesheimer et al., 2005). The scale is, therefore, misaligned with the dominant conception of engagement as a multidimensional concept that includes cognition and emotions (Hollebeek et al., 2022, 2023). Other concerns include specificity (the scale does not exclusively focus on negative engagement) and discriminant validity (the content engagement and negative engagement dimensions are potentially related) (Rahman et al., 2022). Given the above, further refinements to the measurement seem justified.
A closer look at the current treatment of negative engagement from other studies reveals several shortcomings. To begin with, insufficient attention has been paid to conceptual boundaries and some confusion persists between negative consumer engagement and related concepts such as disengagement. For instance, past qualitative studies identify both active and passive negative engagement but regard the latter as disengagement (Naumann et al., 2017a, 2017b). However, other studies report differences between disengagement and passive negative engagement, where the former focuses on the absence of engagement and ending the relationship with focal objects (Bowden et al., 2015; Florenthal, 2019; Rissanen and Luoma-Aho, 2016), while the latter reflects a lower level of engagement (Dolan et al., 2019; Schamari and Schaefers, 2015; Shahbaznezhad et al., 2021).
Concept dimensionality is also an issue (see Table I). Some scholars have taken a unidimensional approach, defining negative engagement as consumers' unfavourable behavioural manifestations during interactions (e.g. Bitter and Grabner-Kräuter, 2016; Dolan et al., 2016). For example, Rahman et al. (2022) operationalise negativity through complaints, thus disregarding negative behaviour of a different nature (Arora and Chakraborty, 2021; Labrecque et al., 2022). Few empirical studies view negative engagement as a multidimensional construct incorporating cognitive, affective and behavioural aspects (e.g. Bowden et al., 2017; Naumann et al., 2020). The multidimensional view that aligns negative engagement with the mainstream conception of positive engagement (Dessart et al., 2016; Hollebeek et al., 2023) seems to represent, now, the dominant approach to negativity. Nonetheless, the multidimensional view is yet to be robustly operationalised.
The absence of robust operationalisation is relevant for several reasons. Construct definition plays a fundamental role in empirical research. The development of a coherent, robust and generalisable theory rests on clearly defined constructs (Bergkvist and Eisend, 2021; Gilliam and Voss, 2013; MacKenzie, 2003). The need for clarity is particularly acute for multidimensional constructs, such as engagement (Hollebeek et al., 2023). When redefining concepts, it is important to avoid two pitfalls: (1) defining the construct solely in terms of its antecedents or outcomes and (2) relying on illustrative examples (MacKenzie, 2003). Unfortunately, some past studies have fallen foul of these principles (e.g. Rissanen and Luoma-Aho, 2016; van Doorn et al., 2010). Clearly, a fresh perspective is needed.
Existing measures of related constructs are a natural starting point for new measure development, and the concept of consumer engagement serves as a useful referent, since existing approaches are inadequate (Frikha, 2019; Haws et al., 2023). Engagement has been typically conceptualised as a three-dimensional construct encompassing cognitions, emotions and behaviours (Barger et al., 2016; Hollebeek et al., 2023) which are directed at an object of engagement, typically a brand (Azer and Alexander, 2020a, 2020b; Rahman et al., 2022). Further, engagement involves a subject of engagement, typically customer or consumer (Heinonen, 2018; Naumann et al., 2020) and occurs in a particular context, for example online (Dessart et al., 2015; Barger et al., 2016) or offline (Naumann et al., 2017b, 2020).
A closer examination of engagement scales rules out adapting or mirroring instruments as possible solutions. Although similarities exist (Table II), there are significant challenges. Considering sentiments, for instance, the affective dimension focuses on emotions with only positive valence (e.g. Dessart et al., 2016; Mirbagheri and Najmi, 2019). The cognitive dimension has been primarily approached as cognitive processing; the items do not reflect its long-term, enduring characteristics (e.g. Dessart et al., 2016; Hollebeek et al., 2014) and do not capture negative thinking (Lourenço et al., 2022). The affective dimension focuses only on positive emotions (e.g. Dessart et al., 2016; Mirbagheri and Najmi, 2019). The indicative behaviours are not related to negative often destructive engagement behaviours (e.g. Azer and Alexander, 2018; Hollebeek et al., 2014; Naumann et al., 2020; Wolter et al., 2023). In addition, many positive statements have no negative equivalent, further hampering adaptation (e.g. learning) (Dessart et al., 2016). Lack of context specificity also matters: although cognitions and affections may not be context specific, studies have shown that engagement behaviours do differ significantly depending on the context (Díaz et al., 2017; Moon et al., 2021).
Indicative articles that reported a quantitative scale development for positive engagement
| Article | Focus | Construct | Dimensions of positive engagement and number of items | ||
|---|---|---|---|---|---|
| Cognitive | Affective | Behavioural | |||
| Algesheimer et al., 2005 | Community | Community engagement | - | - | Community engagement (4 items) |
| Sprott et al., 2009 | Brand | Brand engagement | - | Brand engagement in self-concept (8 items) | - |
| Jahn and Kunz, 2012 | Fanpage | Customer engagement | - | - | Fanpage engagement (5 items) |
| Hollebeek et al., 2014 | Brand | Consumer brand engagement | Cognitive processing: a consumer's level of brand-related thought processing and elaboration in a particular consumer/brand interaction. (3 items) | Affection: a consumer's degree of positive brand-related affect in a particular consumer/brand interaction. (4 items) | Activation: a consumer's level of energy, effort and time spent on a brand in a particular consumer/brand interaction. (3 items) |
| Vivek et al., 2014 | Brand | Customer engagement | Conscious attention: the degree of interest the person has or wishes to have in interacting with the focus of their engagement. (3 items) | Enthused participation: the zealous reactions and feelings of a person related to using or interacting with the focus of their engagement. (4 items) | Social connection: enhancement of the interaction based on the inclusion of others with the focus of engagement. (3 items) |
| Dijkmans et al., 2015 | Social media activities | Online engagement | Familiarity with social media activities (1 item) | - | Online following of these activities (1 item) |
| Dessart et al., 2016 | Brand, brand community | Consumer engagement | Set of enduring and active mental states that a consumer experiences. Attention (2 items); Absorption (4 items) | Summative and enduring level of emotions experienced by a consumer. Enthusiasm (3 items); Enjoyment (3 items) | Behavioural manifestations towards an engagement partner. Sharing (3 items); Learning (3 items); Endorsing (4 items) |
| Schivinski et al., 2016 | Brand-related social-media content | Consumer engagement | - | - | Consumption, Contribution, Creation (17 items) |
| Kumar and Pansari, 2016 | Customer, firm | Customer engagement | - | - | Customer purchase, Referral, Influencer, Knowledge behaviour (16 items) |
| Mirbagheri and Najmi, 2019 | Social media activation campaigns | Consumer engagement | Attention: the extent to which a consumer concentrates on, is attentive to, thinks about, and is absorbed or engrossed in a social media activation campaign. (4 items) | Interest and enjoyment: the extent to which consumers become interested in, or excited about a social media activation campaign, as well as the extent to which they derive pleasure and joy from their experiences with it. (4 items) | Participation: consumers' willingness to spend effort and time during the campaign on activities (4 items) |
| Lourenço et al., 2022 | Brand | Brand engagement | Cognitive (3 items) | Emotion (3 items) | Behaviour (3 items) |
Source: Authors' own work
In sum, the concept of negative online brand engagement requires further refinement and, for several reasons, a full process of measure development is called for. Existing definitions suffer from issues concerning conceptual boundaries and have not taken into account the differences in object contexts. The measurement of positive engagement does provide instruments that can be easily adjusted to capture negativity, even when it has brands as focal objects and the online environment as a context. Consequently, further work is needed to improve both the definition and the measurement of negative online brand engagement. The empirical work that follows addresses these tasks.
3. Scale development process
Following well-established procedures for scale development (Churchill, 1979; DeVellis, 2017; MacKenzie et al., 2011), the study follows a four-step process to reconceptualise negative online brand engagement. Each step consists of several activities (Table III).
Scale development process (four-step)
| Steps | Methods | Data | Results | |
|---|---|---|---|---|
| Step 1: Definition of negative online brand engagement, Item generation | Activity 1 | Literature review on (a) negative and (b) positive consumer engagement | For the period 2000–2020 all papers on negative engagement (31 articles) and 314 articles on positive engagement were selected using a systematic approach (Table IV) | Construct definition, dimensionality and dimensions definitions (Table VI) Generation of the initial set of 171 items |
| Activity 2 | Online observation (Study 1) | A total of 654 Amazon reviews (73,110 words). | ||
| Activity 3 | Semi-structured interviews with moderators of online anti-brand communities (Study 2) | 10 moderators produced 40,231 words of transcription (Table V) | ||
| Activity 4 | Semi-structured interviews with members of online anti-brand communities (Study 3) | 15 community members produced 54,506 words of transcription (Table V) | ||
| Step 2: Item purification | Activity 1 | Face validity | Ask six anti-brand community members if the item pool is relevant to, can be useful and appropriate for what it intends to measure. | Reduced the 171 to 160 items |
| Activity 2 | Research team meeting | Thirteen 60-min-long face-to-face meetings between co-authors. | Reduced the 160 to 61 items | |
| Activity 3 | Academic experts panel (Study 4) | 29 academic researchers in branding | Reduced the 61 to 32 items | |
| Step 3: Reliability and validity | Activity 1 | Pre-test (Study 5) | 20 marketing researchers at the UK-based University | All 32 items were retained |
| Activity 2 | Pilot study (Study 6) | A convenience sample (author's network and snowballing) generated 41 usable responses | No reliability or any other issues concerning administration or response | |
| Activity 3 | Item reduction – Exploratory Factor Analysis (Study 7) | Calibration sample collected from Social Media anti-brand communities (N = 205) (Table VII) | Reduced the 32 to 27 items (Table VIII) | |
| Activity 4 | Item reduction – Confirmatory factor analysis (Study 8) | Both Calibration sample collected from Social Media anti-brand communities (N = 205) and Validation sample collected via snowballing in Social Media (N = 205) (Table VII) | Reduced the 27 to 17 items The 17-item scale exhibited a good fit and properties (Tables IX and X) | |
| Step 4: Nomological network, Discriminant validity | Activity 1 | Negative online brand engagement relationships with brand disloyalty and happiness (Study 9) | Validation sample collected from snowballing in Social Media (N = 205) (Table VII) | Nomological validity supported indicating reliable and validity with good fit and properties (Tables XI) |
Source: Authors' own work
3.1 Step 1: Definition of negative online brand engagement and item generation
Step 1 includes four activities.
Step 1 (Activity 1): Literature review: negative consumer engagement and positive consumer engagement. The process began with the review of negative engagement (Heinonen, 2018; Hollebeek and Chen, 2014; Hollebeek et al., 2022). Focusing on journal articles in academic refereed journals and using five keywords (negative engagement, negative consumer engagement, negative customer engagement, negative brand engagement and negative online engagement), the search generated a total of 31 articles published at the time of review (2022).
Given the limited volume of studies on negative engagement, the literature review was extended to consumer engagement, (Table IV). The search followed a narrow search strategy (Arora and Chakraborty, 2021) focusing on 68 journals ranked in the top 25 per cent in the category “marketing” and 109 journals ranked in the top 20 per cent in the category “strategy and management” in the Scopus 2020 CiteScore ranking, one of the most comprehensive and extensively used ranking instruments (Hollebeek et al., 2023; Pech and Delgado, 2020). The approach enabled a systematic search (Davis et al., 2014; Snyder, 2019; Siddaway et al., 2019) while controlling for the large volume of studies (Hollebeek et al., 2022) and the number of scales reporting positive engagement with various focal objects (Ferreira et al., 2020; Hollebeek et al., 2023).
Criteria for selecting articles on positive consumer engagement
| Marketing | Strategy and management | |
|---|---|---|
| Inclusion criteria #1 – Data source | Scopus top 25% | Scopus top 20% |
| Inclusion criteria #2 – Keywords | consumer engagement OR customer engagement OR positive engagement OR online engagement OR brand engagement OR cognitive engagement OR emotional engagement OR affective engagement OR behavioural engagement | |
| Inclusion criteria #3 – Time period | 2000–2020 | |
| Inclusion criteria #4 – Language | English | |
| Retreated | 320 | 87 |
| Exclusion criteria #1 – Articles with key field mistakes or missing author names | 29 | 43 |
| Retained articles | 291 | 44 |
| Exclusion criteria #2 – Articles outside the marketing/branding areas | 3 | 7 |
| Retained articles | 288 | 37 |
| Exclusion criteria #3 – Articles not focusing on engagement or consumers as the engagement subject | 3 | 8 |
| Retained articles | 285 | 29 |
Source: Authors' own work
The search deployed nine keywords (Table IV) developed in three stages: (1) key to positive consumer engagement relevant and the three most widely identified engagement dimensions (i.e. cognitive, affective and behavioural) themes (Dessart et al., 2019; Morgan-Thomas et al., 2020); (2) assessment of the potentially relevant keywords, considering the brand as the engagement object and the online focus; and (3) discussion of the 11 generated keywords with two marketing experts (two were removed because of the irrelevance of consumer engagement). The adopted exclusion criteria ensured that selected articles relevant to the scope of this research were retained. The search generated 314 articles.
Step 1 (Activity 2): Online observation (Study 1). The next stage of the development process involved an online observation study. Observation reveals negative online brand engagement behaviour in a real setting generating findings to inform the guides for Study 2. Following Kozinets' (2010) recommendations for site selection (i.e. being active, having recent and regular communications) and to ensure the robustness of findings, the observation concerned consumers' negative online reviews of Samsung and Apple on one of the world's largest e-commerce marketplaces, Amazon. Samsung and Apple were chosen because (1) they are both targets of active negative online engagement, with more negative reviews on Amazon (2019) than other brands attracting negative comments (e.g. Sony, HP, Nike, Starbucks and Nestlé); and (2) they had anti-brand groups on Facebook with more than 1,000 members. Contrasted with other brands that also attract negativity (e.g. Sony, HP, Nike, Starbucks and Nestlé), after an Amazon search, Samsung and Apple had more negative online reviews on Amazon in 2019. Data collection focused on negative reviews, i.e. reviews with rankings of 1* or 2* on a five-point scale (1* = least satisfied, 5* = highest satisfied) written between July and September 2019. In addition to the standard text, the data captured textual paralanguage (e.g. emoji), use of all capital letters and punctuation marks for emphasis (e.g. “ABSOLUTE RUBBISH!!!!!!!!!”) and interjections (e.g. “umm”, “hmm”) to reflect deep sentiments. The data set included a total of 481 reviews (63,450 words) for Samsung and 173 (9,660 words) for Apple products. The data collection stopped when information saturation was reached for each brand and in total (i.e. no new themes or coding emerged for 30 posts) (Creswell, 2007; Fusch and Ness, 2015). The data set included a total of 481 reviews (63,450 words) for Samsung products and 173 reviews (9,660 words) for Apple products, all with rankings of 1* or 2* on a five-point scale (1* = least satisfied, 5* = highest satisfied) written from July to September 2019.
Data analysis used thematic coding with existing literature on consumer engagement providing the list of initial codes. Line-by-line coding began with the existing but generated new codes when the existing concepts insufficiently captured the meaning (Clarke and Braun, 2017). NVivo software was used to record codes and themes and then Excel to summarise and organise findings.
Step 1 (Activities 3 and 4): Semi-structured interviews with moderators (Study 2) and members (Study 3) of online anti-brand communities. Interviews aimed to further inform the construct definition, its dimensionality and item generation. To secure negative online brand engagement, participants, 10 moderators (Study 2) and 15 community members (Study 3), were recruited from anti-brand communities on a social media platform (Table V). Facebook was chosen because it is the largest and most widely used social media platform internationally (Lee et al., 2018). Within anti-brand communities, the study targeted members who were highly involved (e.g. high visit frequencies, more time spent on the group) and demonstrated negative online brand engagement (Wong et al., 2018), with moderators typically more involved than members with high anti-brand community activities knowledge. Moderators were contacted first because of their deep anti-brand engagement, their role in the communities and ability to help with member recruitment.
Qualitative phase: respondents' demographics (10 moderators and 15 members of social media hosted anti-brand communities)
| No. | Name | Gender | Nationality | Age group | Employment | Facebook group | Number of words (transcript) | Interview duration (min) |
|---|---|---|---|---|---|---|---|---|
| 1 | Moderator 1 | F | US | 36–45 | At the hotel | I Hate Walmart With A Passion | 4,322 | 36 |
| 2 | Moderator 2 | M | UK | 26–35 | Engineer | We hate BT broadband speed/Openreach | 5,198 | 41 |
| 3 | Moderator 3 | F | UK | 56–65 | Manager | Nestlé Boycott | 1,669 | 13 |
| 4 | Moderator 4 | M | US | 26–35 | Financial advisor | Boycott Disney's Star Wars | 4,771 | 35 |
| 5 | Moderator 5 | F | UK | 56–65 | Lecturer | Nestlé Boycott | 4,352 | 28 |
| 6 | Moderator 6 | M | UK | 56–65 | Retired | BT broadband sucks | 5,828 | 39 |
| 7 | Moderator 7 | M | UK | 26–35 | Vehicle repair | I Hate Apple | 3,805 | 27 |
| 8 | Moderator 8 | M | Kuwait | 26–35 | Manager | I hate Google (page) | 3,272 | 23 |
| 9 | Moderator 9 | M | UK | 36–45 | Engineer | Boycott Amazon the tax avoiding pricks | 1963 | 17 |
| 10 | Moderator10 | M | UK | 26–35 | Insurance | Apple Sucks (page) | 5,051 | 35 |
| 11 | Member 1 | F | US | 46–55 | Bus driver | I Hate Walmart With A Passion | 3,587 | 30 |
| 12 | Member 2 | F | US | 36–45 | Social worker | I Hate Walmart With A Passion | 5,281 | 42 |
| 13 | Member 3 | M | UK | 26–35 | Fun expert | I Hate Apple | 4,615 | 37 |
| 14 | Member 4 | F | Canada | 66–75 | Retired | I Hate Walmart With A Passion | 4,692 | 53 |
| 15 | Member 5 | M | US | 36–45 | Engineer | I Hate Apple | 5,158 | 36 |
| 16 | Member 6 | M | UK | 66–75 | Retired | Nestlé Boycott | 2,848 | 24 |
| 17 | Member 7 | M | Denmark | 36–45 | Engineer | I Hate Apple | 4,712 | 39 |
| 18 | Member 8 | M | UK | 26–35 | Self-employed | BT broadband sucks! | 2,774 | 22 |
| 19 | Member 9 | M | UK | 26–35 | Recycling officer | Nestlé Boycott | 2,892 | 24 |
| 20 | Member 10 | M | US | 46–55 | Disabled | Nestlé Boycott | 5,904 | 58 |
| 21 | Member 11 | F | UK | 56–65 | Library assistant | Nestlé Boycott | 5,700 | 48 |
| 22 | Member 12 | M | US | 46–55 | IT tech | I Hate Apple | 960 | by text |
| 23 | Member 13 | M | US | 18–25 | At grocery store | Nestlé Boycott | 2,472 | 23 |
| 24 | Member 14 | M | Singapore | 26–35 | Software consultant | I Hate Apple | 3,871 | 30 |
| 25 | Member 15 | F | UK | 36–45 | Stay-at-home mother | Nestlé Boycott | 1,049 | by text |
Source: Authors' own work
Both types of participants received invitations to take part in the study via private messages on Facebook or email. Interviews were conducted primarily through video conferencing, except for two with participants with hearing issues where text was used. Normal discussion, interjections (e.g. umm, hmm), non-standard English, voice tone (pitch) changes and vocalizations were recorded and considered in the analysis when relevant. Following data collection and analysis, five respondents were contacted to assess the accuracy of their interview transcripts and the interpretation of the quotes. The interviews with moderators and members produced a total of 40,231 and 54,506 words of transcription, respectively (Table V).
Data analysis deployed thematic analysis to generate codes and themes and involved back-and-forward iterations between the literature and data to ensure the credibility and thoroughness of data analysis (Clarke and Braun, 2017). It followed six steps: familiarising data, generating initial codes, searching for themes, reviewing themes, defining and naming themes, and producing the report (Braun and Clarke, 2006). NVivo and Excel were used when developing and organising codes and themes.
Step 1-Outcomes. The literature review (Step 1 – Activity 1) and qualitative findings (Step 1 – Activities 2–4) informed the refinement of the negative online brand engagement concept (Table VI). Considering the online context and the brand focus, negative online brand engagement is defined, here, as consumer negatively valenced brand-related cognition, affection and online behaviour. The multidimensional view aligns with previous conceptions (Bowden et al., 2017; Naumann et al., 2020) and emerges from the qualitative findings. Specifically, the findings highlight differences among the three dimensions (Table VI), which have not been discussed in previous literature (e.g. Hollebeek and Chen, 2014). They reveal the idiosyncrasy of the online context, supporting the distinct focus on negative online engagement. The qualitative findings also enhance precision illuminating sub-dimensions and, thus, enhancing previous conceptualisations (e.g. Bowden et al., 2017; Naumann et al., 2020). An important outcome of Step 1 is a list of 171 items that potentially capture the various dimensions (Table VI).
Dimensions of negative online brand engagement
| Dimension | Sub-dimension | Literature influencing the definition | Supporting quotes from Studies 1–3 | Number of items | |||
|---|---|---|---|---|---|---|---|
| Step1 Generated | Step 2 Activity 1 | Step 2 Activity 2 | Step 2 Activity 3 | ||||
| Cognitive dimension: the level of a consumer's negatively valenced brand-related attention and thinking | Attention: the extent of a consumer being negatively attended to the brand in the online environment | Vivek et al., 2014; Hollebeek et al., 2014; Dessart et al., 2015, Dessart et al., 2016 | ’As the years progressed, I started to realise that Google is now attracted to the way Apple does which kind of puts me off' (Moderator 10, 31). – Study 2 ’The first impression of the Tab A in white was cheap and nasty.' (Samsung 1) – Study 1 | 32 | 30 | 10 | 4 |
| Thinking: the extent of a consumer considering negatively of a brand in their mind | Hollebeek and Chen, 2014; Dessart et al., 2015, 2016; Fang, 2017; Stathopoulou et al., 2017; Naumann et al., 2017a | ’They thought that the overall storytelling was so poor, that it needs to be recognized that it was poor storytelling. And it ruined their Star Wars experience' (Moderator 4, 30). – Study 2 ’Samsung limits the watch too much and forces you to use what they want. This isn't right' (Samsung 2). – Study 1 | 18 | 14 | 7 | 5 | |
| Affective dimension: the degree of a consumer's negative feelings and emotions toward the brand | Diversity of negative feelings: the collection of consumers' overall negative feelings about the brand | Hollebeek and Chen, 2014; Raïes et al., 2015; Baldus et al., 2015; Dessart et al., 2015; Naumann et al., 2017a | ’And I am also disappointed in the way that they behave.… I think they are holding the industry back in a lot of ways' (Member 5, 44). – Study 3 ’I had my doubts as soon as I powered it on but now after a panel failure after just 3 months I am worried about its longevity' (Samsung 4). – Study1 | 42 | 40 | 23 | 6 |
| Negative emotion demonstration: the extent of consumers consciously surface their negative emotions | Hollebeek and Chen, 2014; Raïes et al., 2015; Baldus et al., 2015; Dessart et al., 2015; Naumann et al., 2017a | ’ABSOLUTE RUBBISH!!!!!!!!! BROKE AFTER A WEEK OF USING THEM! WHAT….A….WASTE….OF….MONEY  ’ (Apple 1). – Study 1 ’The connection is TERRIBLE!!!!!!!!!!!!!!!!!!' (Samsung 5). – Study 1 | 20 | 20 | 11 | 7 | |
| Behavioural dimension: the consumer's negatively valenced constructive and destructive behaviours to a brand in the online environment | Online constructive behaviour: consumers' positively oriented online actions to solve the brand's problem considering one's own concern as well as those of the brand | Romani et al., 2013; Naumann et al., 2017a, Naumann et al., 2017b; Kim and Lim, 2020 | ’I've written many multi-paragraph essays in the group, so if Apple is monitoring the group or anything like that, then they have certainly gotten my opinion in that respect' (Member 5, 44). – Study 3 ’The screen locked and the waiting circle log in the middle of a black screen kicked in..and ran for 12 h during which time I could not do anything' (Apple 2). – Study 1 | 34 | 32 | 5 | 5 |
| Online destructive behaviour: consumers' negatively oriented online actions to harm the brand considering one's own concerns | Plé and Cáceres, 2010; Gebauer et al., 2013; Dolan et al., 2016; Naumann et al., 2017a, Naumann et al., 2017b; Zhang et al., 2018 | ’Sometimes, people can be a little aggressive but it is I HATE Walmart with a passion, so people have that passion…' (Moderator 1, 39) – Study 2 ’It ended up with a logo that said, killer Kit Kat and several members change their pro-Facebook profile to the killer profile and posted on Nestle's Facebook page' (Moderator 3, 61). – Study 2 | 25 | 24 | 5 | 5 | |
| Total | - | - | 171 | 160 | 61 | 32 | |
Source: Authors' own work
3.2 Step 2: Item purification
The second step of the development process involved purificantion and reductions of the 171 items. Three measures ensure the consistency, clarity and parsimony of the item pool.
First, using the definitions of the suggested sub-dimensions, face validity was assessed by six members of online anti-brand communities (four females) identified in Step 1. Taking into account relevance, usefulness and appropriatness reduced the pool of items from 171 to 160 within the three dimensions and six sub-dimensions of negative online brand engagement (Table VI).
Second, each proposed item was evaluated for clarity and alignment with the dimension and definition via thirteen 60-min-long face-to-face meetings between co-authors. The analysis identified inaccuracies, redundancies, repetition and overlaps, particularly within the cognitive and affective dimensions. These evaluations led to further purification of the scale with poorly rated items being removed, reducing the items from 160 to 61 (Table VI).
Third, an academic expert panel (Study 4) evaluated the items. Specifically, the concept definition, dimensions and surviving 61 items were examined by a panel of academic experts who acted as judges (DeVellis, 2017; Rossiter, 2002). A total of 68 branding experts from 19 different countries were contacted and 29 responded. Using a Qualtrics-based survey with structured and open questions, the experts commented on the construct definition, proposed dimensions' and sub-dimensions' structure and definitions. They also rated the suitability of the specific 61 items in terms of clarity and reflection of the sub-dimension's definition by using five-point Likert scale questions (1 = strongly disagree to 5 = strongly agree).
All expert respondents supported the definition and the suggested dimensionality. A threshold of 75 per cent (above 3.75) for clarity and reflection score was used to retain items (Hardesty and Bearden, 2004). Conseqently, 29 items that did not meet the threshold were removed. A total of 32 items, 9, 13 and 10 for the cognitive, affective and behavioural dimensions, respectively, were retained (Table VI).
3.3 Step 3: Reliability and validity
The third step of the scale development process was to address the reliability and validity of the developed scale. Quantitative data from an online questionnaire were collected using the retained 32 anchored on seven-point Likert scales (1 = strongly disagree to 7 = strongly agree).
The instrument was pre-tested for clarity with 20 marketing researchers at a UK-based university using Qualtrics (Study 5). Comments included (1) a few minor issues (e.g. wording and grammatical errors); (2) restructuring suggestions to minimise fatigue and confusion; (3) the inclusion of encouraging statements (e.g. you are doing great); and (4) the replacement of some attention check questions because they required knowledge that some respondents might not have (e.g. The sun rotates around the earth). Thus, several adjustments were made.
To detect any possible issues with the questionnaire missed by the researchers and make further adjustments (van Teijlingen and Hundley, 2002), a pilot study was also conducted (Study 6). Qualtrics survey links were sent through emails to a convenience sample (author's network and snowballing). A screening question to ensure that participants satisfied the study requirements was added (if they have engaged negatively with a brand online). The pilot generated 41 usable responses over the period of 1 month.
Following initial evaluation, the main data collection included two different samples accessed through different methods (Churchill, 1979; Gerbing and Anderson, 1988). First, members of social media (Facebook) anti-brand groups were approached as they actively and negatively engage with brands (Wong et al., 2018). In total, 52 anti-brand moderators agreed to share survey links in their groups. The responses comprise the calibration sample. Second, the snowball method was adopted to recruit participants from the authors' contacts on social media (Facebook and LinkedIn). For this validation sample, the screening question Have you interacted negatively online with a brand (e.g. reading, writing or posting negative comments about the brand)? was added to ensure respondents' negative online brand engagement. All respondents were fluent in English (mother tongue or commonly used foreign language) and were asked to complete the questionnaire focusing on a specific brand. The calibration respondents focused on a target brand within the anti-brand community; the validation response considered a brand target of negative interaction within the last 6 months.
Forcing respondents to answer all content questions and avoid missing data (Décieux et al., 2015) resulted, as expected, in high dropout rates (Wright, 2005; Beynon et al., 2010). Out of 1,356 individuals who started answering the questionnaire, 502 responses were retained, with only non-scale development-related demographic information missing responses – as expected. Further, respondents who failed to answer any of the attention check questions were excluded, resulting in a usable sample of 410 cases (Table VII).
Participants' demographics
| Calibration sample (N = 205) – Facebook anti-brand groups | Validation samples (N = 205) – the authors' contacts | |
|---|---|---|
| Gender | ||
| Female | 119 (58%) | 129 (63%) |
| Male | 86 (42%) | 76 (37%) |
| Age | ||
| 18–24 | 45 (21%) | 31 (15%) |
| 25–34 | 73 (36%) | 100 (49%) |
| 35–44 | 36 (18%) | 31 (15%) |
| 45–54 | 23 (11%) | 30 (15%) |
| Over 55 | 28 (14%) | 13 (6%) |
| Education | ||
| High school | 18 (9%) | 29 (14%) |
| Technical training | 7 (3%) | 12 (6%) |
| Professional qualification | 27 (13%) | 22 (11%) |
| Undergraduate degree | 53 (26%) | 80 (39%) |
| Postgraduate degree | 83 (41%) | 57 (28%) |
| Other | 17 (8%) | 5 (2%) |
| Employment | ||
| Student | 46 (22%) | 38 (19%) |
| Self-employed | 17 (9%) | 21 (10%) |
| Working full-time | 84 (41%) | 117 (57%) |
| Working part-time | 20 (10%) | 5 (2%) |
| Out of work | 10 (5%) | 10 (5%) |
| Retired | 19 (9%) | 12 (6%) |
| Others | 5 (2%) | - |
| N/A | 4 (2%) | 1 (<1%) |
| Country of residence | ||
| Canada | 5 (2%) | 1 (<1%) |
| China | 32 (16%) | 157 (77%) |
| India | 10 (5%) | 1 (<1%) |
| UK | 56 (27%) | 3 (1%) |
| US | 32 (16%) | 1(<1%) |
| Others | 61 (30%) | 23 (11%) |
| N/A | 9 (4%) | 19 (9%) |
| Respondents selected brand | ||
| Nestlé | 25 (12%) | - |
| Apple | 16 (8%) | - |
| HM | 8 (4%) | 18 (9%) |
| Nike | 6 (3%) | 28 (14%) |
| Adidas | - | 4 (2%) |
| Zara | - | 3 (1%) |
| Others | 150 (73%) | 152 (74%) |
Source: Authors' own work
Data collection deployed several means of addressing the common method variance (Tehseen et al., 2017). In terms of procedural remedies, (1) data were collected from different samples; (2) the anonymity of the respondents was protected to reduce evaluation apprehension; (3) the questionnaire pre-testing supported clear instructions and simple, specific and concise questions; and (4) the items capturing constructs were mixed and the order of variables measurement counterbalanced to neutralise method bias related to items' embeddedness. Further statistical remedies are presented in Section 3.4.
In line with guidleines (Churchill, 1979; Gerbing and Anderson, 1988), the calibration sample (205 responses) was used to examine the patterns of data in exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). The validation sample (205 responses) served as validation material for CFA (Study 8) and for the testing the scale's nomological network and discriminate validity (Study 9). Considering 32 items that form the scale, each sample generates a sufficient item to response ratio (Gorsuch, 1983; Cottrell et al., 2007), specifically 6.4:1 cases per item. The samples meet Bartlett's test of sphericity (p < 0.05) (Sun et al., 2020) and exceed the Kaiser–Meyer–Olkin Measure of Sampling Adequacy recommended minimum of 0.6 (Şahan et al., 2019) with 0.938 for the calibration sample and 0.907 for the validation sample.
Study 7 involved EFA to reveal the structure of the negative online brand engagement scale (Costello and Osborne, 2005). Factors were extracted using Maximum likelihood with eigenvalues greater than 1 (Henson and Roberts, 2006) and promax rotation, as the set of loadings with this method frequently reveals a simple structure (Finch, 2006). The analysis was performed in two rounds. In the EFA first round, five items were excluded due to cross-loadings (T1, T2, T4) and low loadings (T3, NED7). Further, data revealed that only the behavioural dimension can be measured with sub-dimensions (online constructive behaviour, online destructive behaviour), while the items of cognitive (attention, thinking) and affective (diversity of negative feelings, negative emotion demonstration) sub-dimensions loaded on one factor each, rather than the theorised sub-dimensions. Based on those results, the cognitive and affective dimensions were measured without sub-dimension.
Parsimony is an important criterion when developing a scale (Ferreira et al., 2020). The data show that behavioural sub-dimensions reflect two different drives of action: one aiming to solve problems and sustain the relationship, the other intending to harm the brand (Kim and Lim, 2020; Naumann et al., 2017b). Both were retained as two separate dimensions. Consequently, the measure of negative online brand engagement, developed here, has four dimensions, namely affective (12 items), cognitive (5 items), online destructive behaviour (5 items) and online constructive behaviour (5 items). EFA using only the retained items shows that the four factors explain 78 per cent of the overall variance each with an eigenvalue higher than one (Table VIII). The items load at 0.565 or over, onto one dimension with no cross-loadings. The dimensions exhibit good reliability, with Cronbach's α values above 0.916, higher than the advocated cut-off point of 0.70 (Al-Osail et al., 2015; Hair et al., 2013, 2019).
EFA scale development-pattern matrix (second round)
| Measured items | Factor | |||
|---|---|---|---|---|
| Affective (Cronbach's α = 0.966) | ||||
| DNF1: This brand arouses intense negative emotions | 0.964 | −0.074 | −0.020 | 0.053 |
| DNF4: I feel uncomfortable when I think about this brand | 0.899 | −0.019 | −0.075 | 0.012 |
| NED2: My negative feelings about this brand could show on my face | 0.853 | −0.071 | −0.115 | −0.022 |
| DNF5: I can use many negative words to describe my feelings towards the brand | 0.845 | 0.009 | 0.135 | 0.031 |
| NED3: People can tell my negative feelings about the brand from my face, body or voice | 0.842 | −0.113 | −0.059 | −0.019 |
| DNF2: I always feel critical about this brand | 0.838 | 0.131 | −0.006 | −0.041 |
| NED5: This brand can make me upset | 0.817 | −0.061 | −0.056 | 0.053 |
| NED1: I experience my negative emotions about this brand very strongly | 0.804 | 0.075 | 0.003 | 0.074 |
| DNF3: I cannot tolerate this brand | 0.804 | 0.106 | 0.101 | −0.058 |
| NED4: I cannot hide my negative feelings about this brand | 0.780 | 0.025 | −0.087 | −0.101 |
| NED6: People can read my negative feelings about this brand | 0.778 | −0.070 | 0.080 | 0.065 |
| DNF6: I detest this brand | 0.763 | 0.103 | 0.129 | −0.044 |
| Cognitive (Cronbach's α = 0.925) | ||||
| A2: If there is anything damning about the brand, I tend to notice it | 0.012 | 0.964 | 0.020 | −0.006 |
| A3: I become aware of anything negative about the brand | 0.028 | 0.911 | −0.012 | 0.007 |
| A4: I tend to observe anything negative about the brand | 0.101 | 0.910 | −0.037 | 0.000 |
| A1: My mind is attracted by anything critical about the brand | 0.025 | 0.858 | 0.021 | 0.025 |
| T5: I consider the negative issues related to the brand | 0.287 | 0.565 | 0.022 | 0.030 |
| Online destructive behaviour (Cronbach's α = 0.973) | ||||
| DB2: If I have the opportunity, I express online my negative thoughts to hurt or damage the brand | 0.034 | −0.039 | 0.971 | −0.010 |
| DB1: If I have the opportunity, I express online my negative feelings to hurt or damage the brand | −0.015 | 0.003 | 0.949 | −0.030 |
| DB4: If I have the opportunity, I post online negative views to hurt or damage the brand | 0.049 | −0.025 | 0.939 | 0.006 |
| DB3: If I have the opportunity, I share online negative comments I noticed to hurt or damage the brand | −0.044 | 0.045 | 0.904 | 0.008 |
| DB5: If I have the opportunity, I take part in online movements against the brand aiming to hurt or damage the brand | 0.006 | 0.020 | 0.898 | −0.029 |
| Online constructive behaviour (Cronbach's α = 0.916) | ||||
| CB2: If I have the opportunity, I express online my negative thoughts to help or improve the brand | 0.000 | −0.023 | −0.079 | 0.932 |
| CB4: If I have the opportunity, I post online negative views to help or improve the brand | 0.037 | −0.008 | −0.047 | 0.931 |
| CB3: If I have the opportunity, I share online negative comments to help or improve the brand | 0.047 | −0.057 | −0.006 | 0.916 |
| CB1: If I have the opportunity, I express online my negative feelings to help or improve the brand | −0.055 | 0.074 | −0.052 | 0.873 |
| CB5: If I have the opportunity, I take part in online movements against the brand to help or improve the brand | −0.021 | 0.082 | 0.273 | 0.656 |
Source: Authors' own work
CFA verified the dimensionality of the negative online brand engagement scale (Jackson et al., 2009; Ou et al., 2016). Using both calibration and validation samples, Study 8 estimated the regression coefficients between the items and the latent constructs. Two different methods of estimation were used: covariance-based (CB) and composite-based partial least squares (PLS) structural equation modelling (SEM) (Astrachan et al., 2014; Dash and Paul, 2021). Considering item structure and loading, both methods produced the same results. Considering that CB-SEM is prefered for factor models (Rigdon et al., 2017), the manuscript reports CB results and the details of PLS SEM (SmartPLS) can be found in the supplementary material. The initial CFA on the calibration sample exhibited poor model fit. Redundant or irrelevant items were deleted through model re-specifications with the modification indices. A total of 10 items were deleted and the reduced 17-item scale exhibited a good fit with both calibration and validation samples (with CMIN = 158.112, DF = 113, CMIN/DF = 1.399, CFI = 0.987, NFI = 0.957, TLI = 0.985 and RMSEA = 0.044 and CMIN = 211.949, DF = 111, CMIN/DF = 1.909, CFI = 0.968, NFI = 0.936, TLI = 0.961 and RMSEA = 0.067, respectively). All the standardised regression weights were above the acceptable threshold of 0.5 (Hair et al., 2006) (Table IX).
CFA: Negative online brand engagement – covariance-based SEM
| Items | Calibration sample | Validation sample | ||
|---|---|---|---|---|
| Estimate | t-value | Estimate | t-value | |
| Affective | Alpha = 0.93, AVE = 0.73 CR = 0.93 | Alpha = 0.94, AVE = 0.73 CR = 0.93 | ||
| This brand can make me upset. (NED5) | 0.756 | 12.51 | 0.877 | 15.64 |
| I experience negative emotions about this brand very strongly. (NED1) | 0.889 | 16.04 | 0.930 | 17.24 |
| I detest this brand. (DNF6) | 0.899 | 16.37 | 0.807 | 13.68 |
| I can use many negative words to describe my feelings towards the brand. (DNF5) | 0.889 | 16.05 | 0.872 | 15.50 |
| This brand arouses intense negative emotions. (DNF1) | 0.822 | 14.16 | 0.773 | 12.89 |
| Cognitive | Alpha = 0.93, AVE = 0.73 CR = 0.93 | Alpha = 0.89, AVE = 0.62 CR = 0.89 | ||
| My mind is attracted by anything critical about the brand. (A1) | 0.832 | 14.51 | 0.671 | 10.39 |
| If there is anything damning about the brand, I tend to notice it. (A2) | 0.900 | 16.48 | 0.820 | 13.85 |
| I become aware of anything negative about the brand. (A3) | 0.922 | 17.18 | 0.870 | 15.16 |
| I tend to observe anything negative about the brand. (A4) | 0.934 | 17.57 | 0.887 | 15.62 |
| I consider the negative issues related to the brand. (T5) | 0.657 | 10.41 | 0.651 | 10.05 |
| Online constructive behaviour | Alpha = 0.93, AVE = 0.82 CR = 0.93 | Alpha = 0.95, AVE = 0.86 CR = 0.95 | ||
| If I have the opportunity, I post online negative views to help or improve the brand. (CB4) | 0.918 | 16.80 | 0.883 | 15.97 |
| If I have the opportunity, I share online negative comments to help or improve the brand. (CB3) | 0.924 | 16.99 | 0.986 | 19.33 |
| If I have the opportunity, I express online my negative thoughts to help or improve the brand. (CB2) | 0.868 | 15.38 | 0.907 | 16.67 |
| Online destructive behaviour | Alpha = 0.97, AVE = 0.88 CR = 0.97 | Alpha = 0.95, AVE = 0.84 CR = 0.96 | ||
| If I have the opportunity, I take part in online movements against the brand aiming to hurt or damage the brand. (DB5) | 0.930 | 17.55 | 0.905 | 16.67 |
| If I have the opportunity, I post online negative views to hurt or damage the brand. (DB4) | 0.970 | 18.96 | 0.946 | 18.03 |
| If I have the opportunity, I share online negative comments I noticed hurt or damage the brand. (DB3) | 0.907 | 16.79 | 0.935 | 17.63 |
| If I have the opportunity, I express online my negative thoughts to hurt or damage the brand. (DB2) | 0.951 | 18.27 | 0.879 | 15.86 |
Source: Authors' own work
Further tests of the reliability and validity of the calibration and validation samples (Table X) demonstrate that the negative online brand engagement dimensions attain good composite reliability (CR) exceeding the recommended level of 0.7 (Bacon et al., 1995; Hair et al., 2006). Convergent validity with the average variance extracted (AVE) ranges from 0.727–0.883 for the calibration sample and 0.618–0.858 for the validation sample, exceeding the minimum acceptable value of 0.5 (Fornell and Larcker, 1981). The square root of AVE for each scale dimension is higher than any of the associated correlations, evidencing discriminant validity (Voorhees et al., 2016).
Negative online brand engagement CFA model – covariance-based SEM
| CR | AVE | Affective | Cognitive | OCB | ODB | ||
|---|---|---|---|---|---|---|---|
| Calibration sample | Affective | 0.930 | 0.727 | 1 | |||
| Cognitive | 0.931 | 0.731 | 0.614*** | 1 | |||
| Online constructive behaviour (OCB) | 0.930 | 0.817 | 0.087 | 0.158* | 1 | ||
| Online destructive behaviour (ODB) | 0.968 | 0.883 | 0.646*** | 0.589*** | 0.020 | 1 | |
| The square root of the AVE | - | - | 0.853 | 0.855 | 0.904 | 0.940 | |
| Validation sample | Affective | 0.930 | 0.729 | 1 | |||
| Cognitive | 0.888 | 0.618 | 0.394*** | 1 | |||
| Online constructive behaviour (OCB) | 0.948 | 0.858 | 0.182* | 0.354*** | 1 | ||
| Online destructive behaviour (ODB) | 0.955 | 0.840 | 0.215** | 0.342*** | 0.237** | 1 | |
| The square root of the AVE | - | - | 0.854 | 0.786 | 0.926 | 0.917 |
Notes: *p < 0.050; **p < 0.010; ***p < 0.001
Source: Authors' own work
No multicollinearity issues are observed between the scale's dimensions with the variance inflation factors (VIF) (O'Brien, 2007) values below 2.0 (Table XI).
Variance inflation factors – covariance-based SEM
| VIF | ||
|---|---|---|
| Calibration sample | Validation sample | |
| DV: Affective | ||
| Cognitive | 1.588 | 1.237 |
| Online constructive behaviour | 1.033 | 1.165 |
| Online destructive behaviour | 1.553 | 1.159 |
| DV: Cognitive | ||
| Affective | 1.567 | 1.059 |
| Online constructive behaviour | 1.008 | 1.087 |
| Online destructive behaviour | 1.558 | 1.095 |
| DV: Online constructive behaviour | ||
| Affective | 1.802 | 1.160 |
| Cognitive | 1.783 | 1.265 |
| Online destructive behaviour | 1.797 | 1.141 |
| DV: Online destructive behaviour | ||
| Affective | 1.543 | 1.156 |
| Cognitive | 1.568 | 1.275 |
| Online constructive behaviour | 1.023 | 1.143 |
Source: Authors' own work
3.4 Step 4: Nomological network and discriminant validity
Testing the scale's relationship with other constructs evidences nomological validity (Study 9). To this end, the study examined the relationship of negative online brand engagement with two concepts: brand disloyalty and happiness. Brand disloyalty (a brand-related construct) was chosen because negatively engaged consumers may reduce, or deliberately avoid, purchasing brand-related products or services (Naumann et al., 2017b; Heinonen, 2018). Happiness (a consumer-related construct) was used because existing literature suggests that consumers can gain emotional benefits through engaging in brand-related activities (van Doorn et al., 2010; Marbach et al., 2016).
To capture brand disloyalty, the three items brand loyalty scale developed by Lin et al. (2019) was adapted (reversed), with the term “never” added in each item to fit the research context and construct definition. To measure consumer happiness, three items from Li and Atkinson (2020) were adapted to fit the research context, with the term “book” replaced by “my negative engagement with this brand”. All items were captured on a seven-point Likert scale. Both constructs demonstrated high reliability in previous studies. The SEM model included 23 items in the analysis, 17 items capturing online negative brand experience and 3 items for each outcome. Item to case ratio for a sample of 205 responses is acceptable at 8.9:1 (Gorsuch, 1983; Cottrell et al., 2007).
Three statistical tests for common method variance assure the lack of bias (Tehseen et al., 2017). First Harman's single-factor test produced a factor explaining 42 per cent and 34 per cent of the total variance for the calibration and validation samples respectively, amounting to less than the 50 per cent threshold (Chang et al., 2010). A marker variable on healthy and balanced diet scale (Żakowska-Biemans et al., 2019) did produce significant change in the model fit indices for both calibration and validation samples (ΔRMSEA = 0.010/0.002, ΔCFI = 0.031/0.030, ΔTLI = 0.036/0.035). Finally, SmartPLS produced an insignificant p-value in the path coefficients for the marker variable in both calibration and validation samples. Contrasting models versus model without the marker showed insignificant change (<0.05). Therefore, common method bias does not seem to be an issue in this study (Chang et al., 2010; Chin et al., 2013).
To test the relationships, the validation sample was used. The SEM model provides support for both hypothesisd negative online brand engagement outcomes, with brand disloyalty (β = 0.201; p < 0.001) and consumer happiness (β = 0.256; p < 0.001). The model demonstrated good fit (with CMIN = 428.269, DF = 221, CMIN/DF = 1.938, CFI = 0.953, NFI = 0.909, TLI = 0.946 and RMSEA = 0.068). All factor-loading estimates were statistically significant and ranged from 0.651 to 0.986 (p < 0.001). The t-values ranged from 10.14 to 19.35. The Cronbach's α values for each scale varied from 0.859 to 0.961 and CRs ranged from 0.864 to 0.961, indicating the internal consistency of the scales. The AVE values ranged from 0.616 to 0.893 and the square root of the AVE is higher than any of the associated correlations demonstrating discriminate validity. The results support the nomological validity of the negative online brand engagement scale and indicate that the new scale is reliable and valid.
4. Discussion and theoretical contribution
This paper aimed to enhance the understanding of negative online brand engagement by addressing its conceptualisation and operationalisation. The study adopted a robust approach to measure development relying on theoretical insights from consumer engagement literature and empirical data from nine studies. The development process drew on experts in the micro-area (academic panel) and in marketing (pilot test) as well as consumers engaging negatively in many online contexts including online commerce (Amazon complainers), social media (anti-brand communities' moderators and members) and in a broader context (researchers' network and snowballing active on social media). Given consumers' extensive engagement in social media platforms (Schultz and Peltier, 2013; Barger et al., 2016), also evidenced for negative engagement (Liao et al., 2023; Lievonen et al., 2022), the scale development secured high participation in the process from social media users.
The findings offer several contributions to the existing knowledge. The first contribution relates to the nature and conception of negative online brand engagement. This paper advances the understanding of negativity by offering a four-dimensional notion of the concept which embraces cognitions, affections, online constructive behaviours and online destructive behaviours. Contrasted with past studies that simply considered the behavioural dimension (Azer and Alexander, 2020a, 2020b; Obilo et al., 2021), this paper enhances precision and integrates diverse approaches. The new conceptual definition focuses specifically on negative online brand engagement, paving pathways for comparative studies of customer/consumer engagement with various objects (Hollebeek et al., 2022, 2023). The conception stresses that both objects and the context of engagement are relevant (Hollebeek et al., 2023). The attention to online context as a moderator of behaviours advances past definitions (Hollebeek and Chen, 2014), providing more detailed explanations of negative online brand engagement.
The second contribution concerns the operationalisation of negative online brand engagement. Responding to calls for increased rigour in scale development for engagement constructs (Ferreira et al., 2020; Hollebeek et al., 2023), this paper makes important headway in building on qualitative insights to develop a valid and reliable scale for the construct. This endeavour offers a major contribution to the existing literature on negative engagement measurement (e.g. Naumann et al., 2020) which, to date, remains limited, in spite of some progress (Obilo et al., 2021). The empirical validation advances, thus far, partial and contested understanding of dimensionality, with some scholars focusing on the behavioural dimension (e.g. Dolan et al., 2016; Obilo et al., 2021) and others supporting three dimensions (e.g. Bowden et al., 2017; Villamediana-Pedrosa et al., 2020). This newly developed four-dimensional scale provides a potential explanation of the exact meaning and applications of negative online brand engagement and differentiates it from the positive side.
The new instrument captures the varying nature of negativity in cognitive and affective engagement, thereby advancing Obilo et al.'s (2021) focus on behaviour. The developed measures of the cognitive dimension show consumers' negatively valenced attention and thinking about certain brands, reflecting a dynamic process. The affective dimension captures a range of brand-related negative feelings which have not been identified in previous consumer engagement literature.
Important theoretical advancement concerns the behavioural dimension. Whereas past scholarship tended to view consumers' destructive behaviours as an aspect of negative consumer engagement (Bowden et al., 2017; Nangpiire et al., 2022; Naumann et al., 2020; Zhang et al., 2018), the treatment of constructive behaviours was inconsistent. Some authors identified constructive behaviour as a sub-dimension of negative behavioural engagement (Naumann et al., 2017a, 2017b), while others assigned them to positive engagement behaviours (Azer and Alexander, 2018). Negative engagement behaviour on social media takes various forms and might even aim to brand experience enhancement (Lievonen et al., 2022). The current research confirms that consumer sharing of negativity online has both destructive and constructive aims and both are related to their negative engagement. In providing a valid and reliable measurement, this paper highlights the unique traits of the behavioural dimension and enhances recent research findings by detailing different types of negative engagement behaviours (Obilo et al., 2021). This observation reconciles previously conflicting findings (Hollebeek et al., 2023).
An important contribution concerns the portability of the scale. Although the new measure reflects both the idiosyncrasies of the brand as an engagement object and the online environment as an engagement context, the scale seems to be adaptable to different contexts and objects, enhancing existing operationalisations (see Obilo et al., 2021). Specifically, all four dimensions can incorporate negative feelings, thinking and behaviours for other engagement objects, akin to some positive engagement scales (Dessart et al., 2016). The validated items seem to be adaptable to other contexts, aligning with calls to treat the online environment as a distinct context of consumer brand interaction (Dwivedi et al., 2023) and paying attention to engagement as context specific. Therefore, the specific uniqueness of the developed measure lies in the negative valence of the engagement spectrum, which remains a strong contribution of this work.
The nomological validity tests reveal a novel relationship between negativity and brand disloyalty. Past studies focusing on brand disloyalty tended to be qualitative (Naumann et al., 2017b). To the researcher's best knowledge, this paper is the first to offer empirical support for the link between these two constructs. The findings have important implications because the inclusion of loyalty moves engagement closer to purchasing behaviour (Bowden et al., 2015; Hollebeek et al., 2023). Suprisingly, this study also uncovers the positive effect of negative online brand engagement on happiness. The findings suggest that consumers feel pleased after engaging negatively with the brand online. The relationship offers an additional explanation of how consumers' negative brand-related cognitions, feelings and behaviours positively affect consumers' average level of satisfaction but have significant negative effects on the brand value and firm performance.
5. Managerial contribution
This research offers several implications for marketing practice. Given that consumers engage with brands negatively online, this work offers a valid and reliable measure of the phenomenon that practising managers can use to estimate the intensity of this primarily unwanted engagement for companies. The longitudinal measure of negative online brand engagement can provide managers with temporal trends of how the sentiments towards their brand change over time. For example, managers could use the 17-item scale combined with sentiment analysis tools as a useful applicable tool to automatically track negative comments online across a wide range of sites and platforms.
The scale can assist managers in examining the marketing tactics' effects on consumer behaviour, and help them grasp consumers' negative brand engagement in social media in particular, but also widely online. Understanding the nature of negative online brand engagement and its components, affective, cognitive and constructive and distractive behaviour, can be advantageous for brand managers who need to appreciate that unpleasant thoughts and feelings can be associated with behaviours of different character, all seen on the surface as similar. Not all consumers intent to harm brands when engaging negatively online, and this scale enables managers to evaluate the nature of the negative behaviours and act appropriately. Examining what could drive negative online constructive behavioural engagement can be highly beneficial, since it could help the brands improve even when the sentiment seems overall negative. The clear identification of the profile of customers and the reasons for developing constructive or distractive brand online negative brand behaviour is needed. Appreciating the triggers leading to negative online brand engagement can help brands to develop appropriate response tactics and a possible reduction of the intended negative online destructive behaviours.
6. Limitations and directions for future research
Despite its contributions, the paper has several limitations concerning sampling and the generalisability of results. Several suggestions are made to advance research in this emerging domain. The first limitation concerns sampling in the quantitative surveys. To gain rich insights into negative online brand engagement, the selection of participants followed non-random principles (i.e. online anti-brand groups and the authors' contacts). Future studies may consider using larger samples or the application of random sampling.
Considering negative engagement in other contexts and with other objects would extend the findings and enhance their generalisability. Future studies may examine offline negative brand engagement and investigate the differences between online and offline engagement. Given the emergence of new online environments, such as the metaverse, examining negative brand engagement in different online contexts will also be of interest. Other engagement objects should be considered and possible candidates include, for example, brand communities and brand community members. Given that qualitative findings highlight the interplay of negative brand engagement and brand community engagement (Bowden et al., 2017), researchers should seek to further investigate negative relationships between different engagement objects.
Given that the scale introduces two different behavioural dimensions, constructive and distractive behaviour, researchers should also further examine the nature of these behaviours and possible antecedents and mediators that can lead to these. For example, it will be interesting to see if the relational history of consumers with the brand may be influencing their intention to engage with these two, different in nature, behaviours.
The authors would like to thank (a) all the researchers who participated in the expert panel or the pre-test of this work for their support in the scale development and (b) the guest editors and the anonymous reviewers for their support in the development of the final version of this manuscript.
The supplementary material are available online for this article.
