Who doesn’t like sport? A taxonomy of non-fans

ABSTRACT

The motives and behaviour of sports fans are heavily researched. Past work has distinguished “fans” and “supporters” on engagement with teams or athletes and identified “non-fans” who have little interest in sport; the latter are rarely investigated further. Sport’s ubiquity, both socially and in media, means that, unusually, disinterested people are often interacting with sport. A better understanding of non-fans could assist strategies to grow sports markets and encourage engagement. This paper describes a study, using both theory-driven and machine learning approaches, of types of self-identified non-fans of a professional sport. A nationally representative sample of 3,496 adults enabled investigation of non-fandom. Five segments of non-fans are identified, differing in terms of consumption of and passion for professional sport. There is a clear hierarchy of likelihood to consume, driven by social contacts, experiences and access to the product, and impeded by satisfying alternatives. To enable easier practical application of this work, a simplified (four question) segmentation process is also presented. This simplified process maintains a high degree of classification accuracy.

KEYWORDS:

• Non-fans
• Segmentation
• Fans
• Latent Class Analysis

Introduction

Whilst most sports¹ consumer research focuses on highly identified, active “fans” (McDonald et al., 2022), it has long been recognised that a continuum of engagement levels exists, from none to heavy. Fans and supporters are commonly distinguished, the latter being less connected and involved with a particular team (Giulianotti, 2002). Moreover, there are sport “consumers” who have no real personal connection to sport yet consume it purely as a means of social connection (Katz et al., 2020).

There is evidence that some people actively avoid sport as a product category (McCullough et al., 2021; Wann et al., 2003), and others who do not consume sport due to a lack of awareness (D. C. Funk & James, 2001). Therefore, among any group of potential sports consumers, committed “fans” could be the minority, with the majority less engaged to varying degrees through to not being engaged at all, for myriad reasons. FIFA (2018) boasted that half the world’s population watched the 2018 men’s World Cup, meaning half did not. Wann et al. (2003) classified 46% of the sports watchers as “casual” (occasional games), with another 9% only watching championship (playoff) games and 13% as non-sports fans at all. The conclusion is that we have sports consumers who range from hardcore fans to very light users and large groups of non-consumers who avoid sport completely.

Research into those who do not identify as sports fans or engage only lightly with sports is scarce. Given the breadth of consumer engagement from none to committed fans, we sought to identify segments across the less engaged/unengaged end of the spectrum, to improve our understanding of, and ability to manage, what might broadly be call “non-fans”.

There are compelling reasons to understand non-fans. The first of those is to respond to customer churn. In order to replace lost customers, sports organizations must acquire new customers from competitors or current non-users (Sharp, 2010). Thus, investigating obstacles to consuming sport and reasons for consuming it lightly or not at all, is beneficial. Given the recent decline in sports fandom, particularly among the young (D. Funk et al., 2016), insights into non-fans are critical to reversing this trend.

Second, understanding non-fans reduces wasted marketing efforts and identifies opportunities for product development (James et al., 2019). As sports organizations seek to grow market share by introducing new product variants (Kunkel et al., 2014), such as Twenty20 Cricket or fantasy sports, information on current non-fans can enable optimal design of products to reach new fans (e.g., Hyde & Pritchard, 2009).

Whilst researchers and writers have long assumed that sport consumers are unique (e.g., “the irrational passion of fans” (Smith & Stewart, 2010, p. 3), some research suggests otherwise (Doyle et al., 2013; Fujak et al., 2018). Studying non-users of your product is vital to understanding market dynamics and growth potential (Sharp, 2010). The purpose of our research is therefore to improve our understanding of non-fans of professional sport. Consequently, we identify and profile types of non-fans to help managers to retain and acquire customers.

Literature review

Defining fans and non-fans

Sports random is an investment in time, effort, and money to know and follow a sports team in anticipation of strong emotional experience and enhanced socialization (McPherson, 1976). However, Wann and James (2019) comprehensive review defined sports fans simply as “individuals that are interested in and follow a sport, team, and/or athlete” (p. 2). They distinguished fans and “spectators” – defined as active witnesses of a sporting event in person or through broadcast means – by degree of interest in the team/athlete, although they acknowledge the terms are often interchanged, suggesting a nuanced perspective.

Numerous attempts to measure and quantify the degree of fandom (avidity) have been reviewed and critiqued (D. Funk et al., 2016; Wann & James, 2019). They typically take one of the three approaches. The first creates continuums or categories of fandom, including Team Identification scales (Wann & Branscombe, 1993), Fan Loyalty (D. C. Funk & James, 2001), Sports-Consumer-Team Relationship Quality (Kim et al., 2011), Psychological Commitment to Team (Mahony et al., 2000), Team Attachment (Robinson & Trail, 2005), Brand Evangelism (Dwyer et al., 2015) or Spectator-Based Brand Equity (Ross et al., 2008). The second approach uses behavioural characteristics (e.g., live attendance or broadcast viewing) to quantify fandom interest and/or intensity (Karg et al., 2019). The third involves segmentation studies, which use variable combinations to define fan groups – typically ranging in avidity (Stewart et al., 2003). In a recent summary, D. Funk et al. (2016) noted that multi-dimensional models were the dominant approach and that segmentation had been applied to many sporting contexts. This work, however, is inconsistent with the clustering variables employed and involves little replication or extension of past work, favouring new categorisation approaches over testing established segmentation approaches in new contexts (e.g., DeSarbo et al., 2017).

Regardless of the classification approach, the conclusions are similar: some sports consumers are highly engaged and interested, others less so. Little of this work goes on to examine those lightest consumers, such as people unaware of sports options, those who tried and rejected sport or those transitioning to or from regular consumption. James et al. (2019) discussed a related issue of many team identification studies incorrectly classifying people without team identification as having “low” identification, simply because some degree of fandom is assumed. This confusion around how to treat those with lower levels of fandom highlights the need for sports management research on non-fans.

Typically, non-fans are defined negatively – as having “no interest (e.g., involvement), no attachment (e.g., identification) and no emotional connection (e.g., attitude) toward the sport product” (Goldsmith & Walker, 2015, p. 232). These terms neglect the self-definitional nuances of non-fandom. Accordingly, we define non-fans as individuals who do not consider themselves fans, regardless of behaviour. We include people who consume sport but do not identify as fans. This definition follows seminal work (e.g., McPherson, 1976) in which engagement beyond consumption is inherent to fandom (Fuschillo, 2020). Moreover, because behavioural measures (e.g., attendance) may not capture important attitudinal aspects of fandom, researchers recommend applying a self-identification measure, such as James et al. (2019) SSIS-R scale, before undertaking further work.

Non-fans are not necessarily non-users

Examination of non-users of various products is commonplace (Sheldon, 2012), with public transport (e.g., Chen et al., 2016), and financial technologies (e.g., Laukkanen, 2016) heavily researched. This work is often grounded in innovation diffusion and adoption (e.g., Rogers, 2010) with early adopters compared to others. As diffusion research progressed, it found that non-adopters varied: some had postponed adoption until market conditions changed, some had tried and rejected, and some had rejected without trying (Szmigin & Foxall, 1998).

These research traditions from other contexts can be applied to conceptualize non-fans of sport similarly. Non-fans, as a broad category, are comprised of those who are unaware, aware but without formed consumption intentions, and aware but have rejected the idea of consuming sport (with varying intensity). However, here the sports market deviates from other fields. Two issues complicate sport's non-use: the ubiquity of sport and the emotional attachment that underpins fandom.

Modern professional sport’s ubiquity removes a major barrier to adoption (or use) – low awareness of/access to the product. Further, the strong emotional connection sport engenders means many fans are evangelical about their teams, more so than in most product categories (Asada & Ko, 2016). Beyond recommendation, fans encourage consumption of their sports team (e.g., hosting a BBQ around a game and inviting non-fans). So, unlike most non-user research, our study of sports fans includes people who may (or not) have consumed – but do not consider themselves fans – of professional sport.

Factors influencing non-fandom

Myriad reasons explain why someone might not be a fan. We identify three categories from past work: lack of awareness or sufficient quality of awareness, constraints that prevent actions being undertaken, and poor experiences or outcomes from engagement. The likelihood that variables usually employed in fan research (e.g., identification, loyalty) accurately assess these categories is minimal (James et al., 2019). While utilizing traditional fan measures (attendance, motives, passion, fandom) to assist with classification and characterisation of segments, we draw from the literature on non-users in innovation adoption, consumer decision-making, and marketing. We also acknowledge the potential influence of a “sport brand ecosystem” that includes federations, leagues, teams, athletes, and managers (Kunkel & Biscaia, 2020). This work underpins exploration of engagement with a product category and the variants people might choose.

Our examination of non-fans in sport includes those who are not fans of sport (the category) overall, and those who are not fans of a specific sports league (brands) but may like others. This provides insights into consumer behaviour at the sport vs. other entertainment levels, and into choices between sporting products and related brands. Given our aim of examining non-fans and producing insights for managers, and most managers operating at league or team level, this focus seems appropriate. Four main theoretical perspectives help shape our study of non-fans.

Hierarchical consumer behaviour models

The Hierarchy of Effects Model (HOEM) suggests consumers fluctuate from use to non-use at key junctures in a process that includes cognition (awareness and knowledge), affect (feelings) and conation (behavioural intentions) (Barry, 1987). Lack of awareness is common in all non-use situations. Sport’s ubiquity means consumption of some sports is unavoidable in many markets, but the profile and reach of professional sports (brands) within that market vary widely (Doyle et al., 2013). At an individual sports level, therefore, a segment of people will be unaware; even the world’s best-known brand, Coca-Cola, lacks 100% market awareness (Finkle, 2018).

Many consumers waver. Sharp (2010) describes situations involving 50% of the people moving from user to non-user status in short periods. That is, in any given period (say 4 weeks) some heavy buyers will not buy and some non-buyers will. For example, the MLB season is long, and a baseball fan may take an overseas holiday mid-season, going from perfect attendance to none in consecutive months. Applying this view of consumer decision-making as a series of feedback loops to sport fandom suggests potential fans are at different stages of evaluation of sport and could be moving towards becoming a fan or moving away.

We could expect non-fan segments to differ in terms of awareness (including number of brand associations and overall brand salience), experience (past use via several channels), response to experience (pleasant, unpleasant), and intentions to use again. In sports research, HOEM models are used primarily in sponsorship impact studies (e.g., Alexandris et al., 2012), although Tsiotsou (2013) incorporated them into research on fan loyalty and D. C. Funk and James (2001) built on the HOEM in their PCM framework. By viewing the decision to become a fan or not as a stage-based process rather than a dichotomous one-off decision, we expect improved understanding of the nuances of non-fans.

Consideration sets

Hierarchical consumer behaviour models align well with the “consideration set” notion – that product usage varies because brands have variable mental availability (recall). For example, consumers facing many alternatives in most product categories are adept at narrowing their attention to a few options (He et al., 2016). As such, brands fall into three categories: the Evoked Set (freely recalled), the Inert Set (consumers know but fail to recall without prompting), and the Inept Set (known but discounted) (Narayana & Markin, 1975). The sports market is crowded (Fujak et al., 2018), so narrowing of products and related brands under consideration is inevitable.

As the number of alternatives in any decision situation increases, the attention paid to each decreases; typically, consumers’ subsets include 3–6 brands (Hauser & Wernerfelt, 1990). Further, the process by which consumers reduce the alternatives being considered is often based on past experiences or reinforcement of existing beliefs (Lleras et al., 2017). Reflecting on the role of brand recall adds granularity to explanations of sport's non-use. Failure to recall impedes usage, whereas use increases the likelihood of consideration in the future. Sharp (2010) contends that use drives favourable brand associations (i.e., we like brands we use more than those we do not), setting up a reinforcement cycle and barriers to new brands. The more we experience a particular sports team, the more favourable associations form (Daniels et al., 2019).

Leisure constraints

Beyond these standard causes of non-use and preference formation are particular aspects of the sport experience, which can prevent someone developing fandom. Leisure constraints (see Choi et al., 2019) “impede or inhibit an individual from attending a sporting event” (Kim & Trail, 2010, p. 191). Early work on attendance identified internal constraints (e.g., ignorance) and external constraints (costs, location) linked to consumer motives. However, constraints are often overstated, because context means some constraints matter less (e.g., media consumption; Larkin et al., 2015), and adequately motivated consumers can overcome them (Blank et al., 2014).

The fan development literature also discusses intrapersonal factors that drive interest and involvement with sport. For example, the shared joy of success (eustress) – or “basking in reflected glory” – was a clear motive to consume sports (Cialdini et al., 1976). D. C. Funk et al. (2009) summarized reasons why individuals become interested in sport, suggesting five core motives: socialization, performance, excitement, esteem, and diversion. Sport also signifies belonging to a community or region (Delia & James, 2018).

Among the constraints, social connections (or lack thereof) seem to be significant drivers of non-fandom (Katz et al., 2020). In sport, socialisation refers to people beginning to consume sport through social network influence and professional athletes and teams enabling people to be part of an “in-group” (Lock & Heere, 2017). Sport is frequently consumed socially and preferences communicated. Socialising agents can create awareness and engage parents, coaches, friends, local communities and media, and their absence could lead to non-fandom.

Customer disengagement

A final factor emerges from marketing studies of the negative dimensions of customer engagement (Hollebeek & Chen, 2014). Past customer disengagement research has identified that some customers disengage over time, and those disengaged customers display a range of reactions from passive avoidance through to active negative behaviours like boycotts and protests (Juric et al., 2015). These negative behaviours can impact other customers and potential customers (Azer & Alexander, 2020), increasing their negativity (Kähr et al., 2016).

The direct research on sports fan disengagement, and how those disengaged with sports express themselves, has been very limited to date (Martin & Goldman 2016). As a highly co-created service, disengagement with sport could be expected to occur less often (Naumann, Bowden & Gabbott, 2020), and the high level of identification many fans feel means they are less likely to disengage even when the team's performance is unsatisfactory (Norris, Wann & Zapalac, 2015). However, a large body of work exists examining churn among sports season ticket holders (e.g., Karg et al. 2021) showing that factors like changes to family structure, work arrangements and team performance can all lead to reduced engagement. Similarly, lockouts and COVID-19 related disruptions to seasons have been found to negatively impact fan engagement (Reade & Singleton, 2021). Research conducted into sport-based social media has identified both fan disengagement being expressed and negative-valenced influencing behaviours like verbal aggression and negative word of mouth (Filo et al., 2015). Given the lack of direct research, at this stage, all we can say is that it is likely that some sports fans disengage and that we should expect variation in the activeness and intensity of the responses of these disengaged fans.

Method

Context

Examination of non-fans could be conducted at team, league, or sport industry level. For this initial examination of non-fans, a national sport league perspective seemed a beneficial starting point, as did Australia as a context. Australia has a reputation as a sport-obsessed country (Fujak et al., 2020); 80% of Australians agree that “sport is a significant part of Australian culture” (McCrindle, 2019). Given this reputation and comprehensive sports coverage (Fujak et al., 2020), the decision to reject sport is probably a conscious one for Australians. Australia has four major professional football codes plus professional netball, many formats of cricket, and basketball. Horse racing, tennis, motor sports, and combat sports attract large audiences. Free-to-air, print and online sports coverage remains strong, with laws protecting free access to major events (Rowe & Gilmour, 2009).

Examining non-fans required both category and brand foci. We chose sport as the category level and the Australian Football League (AFL) as the brand level. A team level focal point was considered, but some consumers like a league without a specific team (Kunkel et al., 2013), especially those new to the sport. The AFL is Australia’s largest professional sports code but operates in a highly competitive market with eight other professional codes. The AFL context is covered comprehensively in Stewart et al. (2005), but briefly, long-standing regional preferences for football codes exist, the AFL is the world’s fifth highest attended league, and its 18 professional teams are viable with only 26 million Australians.

Measures

We began with an a priori segmentation approach using measures based on the theories described above. Before classifying consumers into fans and non-fans, we collected a sports league awareness measure. Non-fans unaware of the AFL were not questioned further about the AFL, but general demographics and behaviors were collected. Then, we measured self-identified fandom at the category (sport) and brand (AFL) level, following James and Ridinger (2002), but using a 7-point scale: labelled “1 = casual observer” and “7 = hardcore fan” (Kunkel et al., 2021). Additionally, “0” = “not a fan at all” captured definite non-fans, distinct from low-level fans (James et al., 2019). The use of such single-item “avidity” scales occurs in commercial research for major sporting leagues (K. Wakefield, 2016) and in academic work (James et al., 2019; Kunkel et al., 2021).

This self-identification measure was supplemented by measures of behaviour (again, at category and brand level), both live and broadcast consumption (McDonald et al., 2014). Attitudes towards the AFL were measured on a single-item passion scale (K. Wakefield, 2016) and used as a consistency check against the self-identification scale. Measures of internal and external constraints to AFL consumption were adapted from Larkin et al. (2015) and modified for relevance to non-fans. To capture possible customer disengagement, the brand association was measured (after Daniels et al., 2019) by asking respondents “When you think of AFL, what words come to mind?” The first five words were used to examine the most salient associations, as listed in the segment portraits (Results).

To assess whether respondents had fixed or changed avidity, future intentions were collected using purchase probability measures (McDonald et al., 2014), for broadcast and live viewing of the AFL men’s and AFL women’s (AFLW) competitions. AFLW was established in 2016; its season runs over the summer, leading into the AFL men’s competition. The rationale for starting AFLW included broadening audiences and capturing new fans so, whilst technically in the same league, consumers could be fans of one and not the other.

These measures allowed distinguishing non-users (no consumption) and non-fans (possible consumption, but without positive attitudes or sense of identity). To improve our understanding of the segments and to profile them effectively, we assessed demographics, overall engagement with sport, AFL-related behaviours, and perceptions of AFL, including social, economic and access constraints.

Participants

Two commercial survey panel providers gave access to 3,496 respondents, population-matched in key demographics (age, gender, birthplace) and geographic location (state and capital city versus other) provided by the Australian Bureau of Statistics. Respondents were 50.9% female; 23% were born outside Australia (mostly New Zealand, China and the United Kingdom). An online survey was used to collect data.

In our sample, 38% were identified as both sports fans and fans of the AFL specifically, consistent with previous reports (Fujak et al., 2018), with 2,197 respondents self-described as non-fans of the AFL. This measure was checked for consistency by comparing respondents’ levels of consumption and attitudes towards AFL.

Analysis

Our approach to segmenting augmented typical sports consumer research practice in three ways. First, we used an a priori method (based on previous segmentation research, Funk et al., 2016), combined with a data-analytical approach to produce the final cohorts. Second, we employed latent class analysis (LCA), instead of the k-means clustering used in most sports research, based on several key advantages. LCA makes no assumptions about normality or linearity, and requires no standardisation of continuous variables before analysis (Magidson & Vermunt, 2002). LCA clusters are probabilistic rather than ad hoc, generating a higher accuracy of allocation of individuals to clusters (Magidson & Vermunt, 2002; Masyn, 2013). LCA provides multiple statistical methods for identifying the “best” solution and number of clusters (Nylund et al., 2007). Finally, LCA allows direct inclusion of covariates (predictors/causal drivers of cluster membership) in the cluster model, giving a more robust test for covariate effects by accounting for errors in classifying individuals into clusters (Vermunt, 2010).

Our third advance on traditional segmentation, following recent criticism of it (Ernst & Dolnicar, 2018), was to employ LCA techniques to reduce the possibility of segmenting based on idiosyncrasies. LCA estimates solutions using multiple starting parameter values and bootstraps the solution to ensure a robust and replicable solution (Magidson & Vermunt, 2002) – a major advantage over k-means clustering. We estimated each possible solution (including numbers of clusters and profiles) used 100 starting values or model parameters, retaining the best solution from each iteration set (Masyn, 2013) to choose the solution that represents the “global maximum” rather than a “local maximum” (McCutcheon, 2002).

Hence, LCA was used to produce data-driven segments (posterior segmentation). A priori and post hoc analysis were highly consistent; they are presented in the following section, with more details of our approach.

Results

Clusters of non-fans

We performed LCA using the LatentGold version 5.1 (Vermunt & Magidson, 2013). Our indicator variables describe attitudes and behaviour related to the AFL and the other sports respondents followed. With the number of clusters being unknown prior to analysis, we estimated solutions involving 1 to 10 clusters. We selected the number of clusters based on the Bayesian information criterion (BIC), with the lowest value representing the best relative fit while accounting for model complexity (Masyn, 2013). As Table 1 shows, the BIC is the lowest for the 5-cluster solution.

Table 1. Model fit statistics.

Number of Clusters	LL	BIC (LL)
1 Cluster	−15115.9	30539.4
2 Clusters	−13640.7	27704.4
3 Clusters	−13514.2	27566.6
4 Clusters	−13440.4	27534.5
5 Clusters	−13360.7	27490.4
6 Clusters	−13311.8	27508.0
7 Clusters	−13272.8	27545.3
8 Clusters	−13244.5	27604.1
9 Clusters	−13232.6	27695.7
10 Clusters	−13212.5	27770.8

We inspected the profile of the cluster to ensure classes were separate, interpretable, and no small clusters were generated purely to fit the data (Masyn, 2013). The 5-cluster solution meets these criteria: the smallest represents over 10% of the sample, and the cluster profiles are quite separate (Table 2).

Table 2. Cluster profiles.

	Haters	Dabblers	Rival Coders	Connected	Flaneurs
Cluster Size	21.1%	29.2%	17.6%	21.3%	10.7%
Cluster Profile
Awareness of AFL	79.4%	100.0%	95.4%	99.5%	99.9%
Follow Professional Sports	0.1%	0.0%	100.0%	100.0%	100.0%
Passion for AFL (1–5)	1.1	1.8	1.3	2.6	2.5
Level of AFL Fandom (0–7)	0.0	1.6	0.5	2.8	2.9
Watched AFL on TV or Online	1.6%	72.6%	45.5%	81.4%	93.4%
Attended an AFL game	3.6%	41.5%	16.8%	46.0%	55.6%
Other Sports Followed
Rugby League (NRL)	0.0%	0.0%	59.4%	28.8%	87.3%
Rugby Union (ARU)	0.0%	0.0%	19.4%	12.2%	59.6%
Australian Soccer	0.0%	0.0%	5.9%	28.4%	54.2%
Australian Cricket	0.0%	0.0%	37.6%	20.0%	85.5%
Basketball	0.0%	0.0%	2.5%	16.0%	21.8%
Netball	0.0%	0.0%	9.9%	9.9%	22.8%
Tennis	0.0%	0.0%	18.1%	51.3%	52.9%
V8 Supercars	0.0%	0.0%	31.1%	16.2%	31.8%

Having determined the optimal number of clusters, we used LatentGold’s Step 3 module to test for the impact of covariates – factors that may influence cluster membership (Vermunt, 2010). The impact of covariates was tested overall through the Wald statistic and in relation to each cluster (Table 3). A significant positive coefficient for a covariate represents that covariate’s positive impact on membership of that cluster and vice versa.

Table 3. Covariate results for clusters.

	1. Haters	2. Dabblers	3. Rival Coders	4. Connected	5. Flaneurs	Wald
	Coeff.	Coeff.	Coeff.	Coeff.	Coeff.	Coeff.
State						123.00***
ACT	−0.57	−0.50	−0.51	0.27	1.31*
New South Wales	−0.68**	−0.54**	−0.24	0.26	1.21
Northern Territory	1.12	−0.20	1.16	−1.99*	−0.09
Queensland	−0.61*	−0.52**	0.30	−0.39	1.21
South Australia	0.40	0.11	0.07	0.22	−0.80
Tasmania	0.42	0.63	0.33	0.31	−1.69
Victoria	−0.33	0.73**	−0.98*	1.02***	−0.43
Western Australia	0.25	0.29	−0.12	0.30	−0.72
Gender (Female)	0.66***	0.19	0.02	−0.15	−0.71***	124.8***
Age						46.15***
18–24	−0.20	0.14	−0.56	0.77***	−0.15
25–29	0.27	−0.19	−0.17	0.19	−0.09
30–39	0.07	−0.11	0.05	0.30*	−0.30
40–49	−0.13	0.05	0.15	−0.24	0.16
50–59	−0.22	0.25*	0.20	−0.35*	0.11
60+	0.21	−0.13	0.33	−0.67**	0.26
Country of Birth (Australia)	−0.11	−0.06	0.23*	−0.28**	0.21	14.93**
Parents follow/ed (Yes)	−0.32*	0.26**	−0.33*	0.31**	0.09	13.92**
Partner follows (Yes)	−1.01*	0.26	0.00	0.38*	0.38*	13.51
Children follow						11.98
Yes	−1.04	0.24	−0.31	0.51*	0.60*
No	0.47	−0.10	0.23	−0.30*	−0.31
No children	0.56	−0.14	0.08	−0.22	−0.29
% of friends that are fans						102.08***
0%	2.12***	−0.32	0.84***	−0.25	−2.38***
1–25%	0.28	−0.13	−0.08	−0.17	0.10
26–50%	−1.05*	0.19	−0.44*	0.41*	0.89**
51%-100%	−1.35*	0.26	−0.31	0.01	1.39***
Hours per week for leisure	0.01	−0.01	0.00	−0.01	0.00	1.62
Pay TV Connection (Yes)	−0.37***	−0.08	0.04	0.09	0.31***	34.66***
Income						14.93
Up to $50,000	0.05	0.15	0.14	−0.31*	−0.02
$50,001 to $100,000	−0.01	−0.05	−0.07	0.01	0.11
$100,001+	−0.21	0.05	−0.05	0.29**	−0.08
Prefer not to say	0.17	−0.15	−0.02	0.01	−0.02

*p < 0.05, **p < 0.01, ***p < 0.001.

Cluster validation – outcome comparison

We validated our clusters by comparing outcome variables we expected to differ: likelihood to attend or watch AFL or AFLW in the next 12 months. The results (Table 4) can be interpreted similarly to the covariate analysis.

Table 4. Cluster behavioural likelihoods.

	1. Haters	2. Dabblers	3. Rival Coders	4. Connected	5. Flaneurs	Wald
	Coeff.	Coeff.	Coeff.	Coeff.	Coeff.	Coeff.
Likelihood to attend AFL	−2.09***	0.04	−2.28***	2.13***	2.20***	877.25***
Likelihood to watch AFL	−3.70***	0.28*	−2.93***	2.44***	3.91***	1719.89***
Likelihood to attend AFLW	−1.70***	0.01	−1.89***	2.07***	1.50***	602.23***
Likelihood to watch AFLW	−2.48***	0.13	−2.15***	2.11***	2.40***	856.29***

*p < 0.05, **p < 0.01, ***p < 0.001.

Clusters differ in every outcome variable (Wald statistics are significant). Clusters 4 and 5 are positively related with each outcome, representing higher likelihoods to attend or watch in the next 12 months. In clusters 1 and 3, watching and attending are strongly negative, signifying a lack of interest in sport generally and AFL specifically, respectively. Finally, a positive relationship for cluster 2 signifies likelihood to watch AFL but not attend, and not to watch or attend AFLW.

These differences emerge from engagement in AFL-related activities (Table 5) included in our survey to capture more passive consumption than attendance or viewership. For example, tipping contests are common in Australian workplaces, and people with little interest and/or knowledge participate due to social norms. Similarly, purchasing sporting merchandise as a gift does not necessarily indicate high involvement. Cluster 1 has almost no involvement in any activities – not easy, given AFL’s ubiquity in Australia. Similarly, few individuals in cluster 3 participated, despite being sports fans.

Table 5. Segment participation in AFL-related activities (% participating).

	1. Haters	2. Dabblers	3. Rival Coders	4. Connected	5. Flaneurs	All
Played in an AFL Tipping competition	0.9	14.3	3.0	13.9	25.2	10.6
Purchased AFL merchandise for yourself or as a gift	1.6	15.2	4.6	16.6	19.0	11.3
Placed a bet on AFL	0.0	5.2	1.4	13.4	16.4	6.5
Follow AFL or related pages on Social Media	0.0	4.7	0.5	10.7	12.4	5.2
Use AFL or related apps on your mobile device	0.0	3.3	0.5	5.5	7.5	3.1
Played in an AFL Fantasy Football competition	0.0	1.4	0.8	4.7	4.9	2.1
Played AFL (e.g., having a kick of the footy, playing for a local club, playing Auskick)	1.2	9.7	3.5	13.4	16.4	8.4
None of these	96.2	65.2	88.9	46.3	47.8	69.6

Table 6 summarises cluster responses to questions testing respondents’ perceptions of AFL’s affordability, ease and safety of attendance, and the relevance of sport to them, scored on a 1 [strongly disagree] to 7 [strongly agree] scale. Clusters vary widely. Cluster 1 has negative responses (scores below 4) on all items and is the lowest scoring cluster on each. Cluster 3 is largely apathetic, tending towards negativity, while other clusters are largely positive.

Table 6. Segment attitudes to AFL (mean 1 “strongly disagree” − 7 “strongly agree”).

	1. Haters	2. Dabblers	3. Rival Coders	4. Connected	5. Flaneurs	All
AFL matches are played at convenient times for me to attend.	2.8	4.0	3.2	4.4	4.7	3.9
Crowd behaviour at AFL games is good.	2.9	4.2	3.7	4.5	4.9	4.1
I would feel safe attending an AFL game.	3.5	4.7	4.4	4.9	5.4	4.6
I would prefer to watch AFL on TV rather than live.	2.5	4.2	3.5	4.7	4.8	4.1
AFL games are affordable to attend.	2.5	3.4	3.0	4.2	4.3	3.6
AFL games are easy to attend if you want to.	3.4	4.4	3.8	4.6	4.8	4.3
My family/social group are interested in going to AFL	1.6	3.7	2.2	4.1	4.2	3.3
Tickets to AFL games are easily purchased in advance.	3.3	4.8	4.4	4.9	5.3	4.7
AFL is a sport that is relevant to me.	1.4	2.6	1.7	3.5	3.9	2.6
AFL players are good role models.	2.5	3.4	2.8	4.0	4.1	3.4
AFL is a safe sport to play.	2.8	3.8	3.5	4.2	4.6	3.8

In addition, respondents were asked to rate marketing incentives (e.g., free tickets, food vouchers) based on their influence on attendance behaviour, and about their other leisure activities. These data were used to complete the segment portraits presented below.

Segment portraits

Combining profiles (Table 2) with significant covariates (Table 3) provides a detailed cluster description. Data presented in Tables 4–6 also provided insight into how clusters that behave similarly think differently.

Cluster 1 includes 21% of the respondents with the lowest AFL awareness (Table 2), lowest passion for AFL and lowest self-perceived level of AFL fandom among clusters, with tiny percentages ever watching or attending a game. This lack of interest extends to other sports, and clustered individuals have few connections to AFL fans. Cluster 1 can be described (pejoratively, and in contemporary language) as sports “Haters”. The group’s most common brand associations for AFL were “boring”, “overpaid”, “stupid/dumb”, “rough”, “scandal” and “alcohol”, indicating negative customer engagement. This characterization is supported by negative attitudinal responses (Table 6) and behavioural intention outcomes, with this cluster being the least likely to attend or watch AFL or AFLW in the next year.

Haters do not like sport, and although most are aware of AFL, choose not to consume it. A major barrier to consuming AFL is lack of interest; they do not respond favourably to the presented marketing activities (e.g., free tickets, competitions). Their leisure time is spent on cultural pursuits and activities, such as gardening or reading. They have concluded their decision-making with an informed rejection of the AFL.

Cluster 2 represents the largest segment (29%); they are aware of the AFL, but none follow professional sports. Consistent with this profile, the average self-rated passion for AFL and the level of AFL fandom (Table 2) are low. Nonetheless, most have watched AFL on TV or online, and many have attended a live game. Their associations with the AFL are largely basic descriptors like “Australian” and “fast”’. Due to their surface-level engagement and sporadic consumption, we label cluster 2 “Dabblers”.

Dabblers do not see themselves as sports fans but have tried AFL; 70% are women. Typically, Australian born, about a third have parents who followed AFL. However, only about a third have ever engaged in AFL-related activities, and only around a quarter expect to consume AFL in the future – due more to circumstances (e.g., where they live/socialise) than interest. Realistically, this group has barely left the cognitive phase – aware of AFL but do lack strong attitudes and intentions to consume. They have considerable leisure time, but other hobbies (gardening and walking were commonly named), and a lack of interest stop them consuming more AFL. They are price-sensitive and respond well to free tickets.

Cluster 3 (18%) has high awareness of AFL (95%), but the passion and fandom of this cluster is second-lowest (after Haters) (Table 2). Additionally, less than half have watched an AFL game, and few have attended. However, the overall following of professional sports in this cluster is 100%; this cluster follows rugby league, cricket or V8 supercars (Table 2). This profile, particularly of interest in rugby league, suggests the designation “Rival Coders”. Like Haters, they are unlikely to have friends that follow the AFL.

Rival Coders are “informed rejecters” of AFL, placing it in their inept set. Common brand associations for AFL include “team”, “winter” and “confusing”. They have few connections to the AFL that might influence their future behaviour, and thus favour socially and geographically embedded alternatives. Most have never engaged in AFL activities and have negative attitudes towards AFL. Only a fifth were interested in free AFL tickets.

In Cluster 4 (21.3%), all members self-rate as followers of professional sport. Over half follow professional tennis, with rugby league and Australian soccer strongly supported (Table 2). The average passion for AFL and fandom level (2.8) are equal highest (with Cluster 5), although moderate in absolute terms. Most of this cluster have watched AFL, while slightly under half have attended a game. Cluster 4 includes younger respondents with partners and/or children who follow the AFL; we label them “Connected”.

The Connected have consumed at a reasonably high level due to their strong connections with AFL fans but did not consider themselves fans. They are positive towards AFL, with top brand associations being “exciting”, “fast”, and “fun”. AFL remains in their consideration set of entertainment options, due to their desire to socialise; given this motivation for consumption, it is unclear whether this group has reached the affective stage for AFL. They have the least leisure time but respond very well to marketing actions, particularly ones that reduce cost. They are an attractive segment to managers, being young and having disposable income.

Finally, Cluster 5 (11%) were consuming so much sport that the AFL did not fit into their top choices. Like Cluster 2, Cluster 5 has higher AFL awareness, passion and fandom than other clusters (Table 2). Nearly all of them have watched AFL and over half have attended. A key distinguishing feature are majorities following other sports – rugby league and cricket heavily, plus rugby union, Australian soccer and tennis. Based on the variety of sports followed, we borrow from Giulianotti (1999) and call them “Flaneurs”. They have some interest in the AFL, but not as a top sport.

Flaneurs have tried AFL, but prefer other activities or sports. They have moved into the cognitive stage of the HOEM, but still lack self-identification as a fan. Existing habits prevent increased use. Unlike Rival Coders, their attitudes to AFL are generally positive, with associations led by “exciting” “skilful” and “fun”, and they expect to consume AFL in the future. They are occasional users of the AFL, which remains in their consideration set but is not their preferred brand.

Discussion

We found that 38% of the population self-identified as fans of the AFL, leaving 62% who did not. Five distinct segments of non-fans were found, with significant variation in propensity to consume in the future (and possibly become fans) and factors affecting avidity.

Theoretical implications

This study is the first to profile distinct cohorts within self-identified non-fans. However, some theoretically likely segments were absent. For example, most consumer research and non-user studies find large numbers of people unaware of the product or innovation long after launch (e.g., after 13 years, 22% of Canadians had not heard of Toyota Prius – Long et al., 2019). That is not the case here, where awareness was almost universal – a testament to AFL’s broad coverage in Australia. However, minor leagues (e.g., lower-tier soccer in Europe) and less prominent sports (e.g., field hockey), could expect large numbers of non-fans in the unaware and aware-but-unable-to-consume segments.

Most non-fans were aware of the AFL and had consumed it in some form (Table 5) but did not define themselves as fans. Like past consumer innovation researchers, we found distinct cohorts amongst non-fans: those who had rejected ever-consuming AFL, those who had consumed and rejected, and those who intended to consume but had not yet done so. With insights from the HOEM and sports consumer models (e.g., PCM), a stage-based process of fandom development emerged, where awareness initially prevents adoption, experience influences desire to consume again, and constraints such as competing interests and minimal social connections dampen ability to action desire for engagement. The non-linearity of the HOEM in a sport consumption context can be inferred from analysis of our three segments not typical of non-user or adoption research.

First, the Connected group’s consumption decisions are not driven by desire to adopt the sport. They consume AFL as a conduit, mainly to socialising with friends or family. While social and reference groups are sources of risk when deciding to adopt a new product (Kleijnen et al., 2009), the influence of family and friends on introducing sports reported (Katz et al., 2019), engaging in sport enables this segment to achieve a non-sport outcome of social interaction.

Similarly, Dabblers exist because of the ubiquity of sport. This segment reported inability to avoid adoption due to widespread (free) broadcast, media coverage and a large existing user base. This resembles “forced adoption” (Feng et al., 2019); the experience can cause long-lasting harm to customer satisfaction and intention to reuse. However, Dabblers held largely positive attitudes, suggesting they were not unhappy with the product, even if consumption was unintentional.

Haters reject the product category entirely. Whilst ardent rejectors of certain brands and even sub-categories of products (such as credit cards – Szmigin & Foxall, 1998) exist, a group rejecting an entire category outright is unusual. Haters’ strong negative attitudes, brand associations, and defiant anti-consumption stance differentiate them from rejectors of other categories, such as fashion (Jahanmir & Lages, 2016). In addition, Haters have low-fandom family and friends, possibly forming an anti-brand community (Hollebeek & Chen, 2014). Their strong, negative passion suggests more than a rejection based on assessment of functional benefits (Hollenbeck & Zinkhan, 2010). Unlike some reticent consumers, their conversion to sport is improbable (Gilly et al., 2012).

The Hater segment highlights that negative sentiment can exist toward sport in response to its wide distribution and public discussion, questionable benefits of mega-events, government support and funding, and high-profile athletes’ behaviour (Giulianotti et al., 2015). Given the effort required to avoid sport, it seems feasible that negative feelings will intensify, similar to the effect in contexts like self-service, where consumers lack the ability to adopt or not (Feng et al., 2019). Given the size of this segment, even in this pro-sports context, the genesis of the Hater segment seems worthy of further study.

Managerial implications

There are clear implications of knowing that some people who consume your sport are not fans. First, some segments attend or watch but are not directly engaged with the team. These “customers” are therefore out of the direct reach of the team’s marketing efforts, and building a relationship with them requires assistance from other fans in their network. Second, many are consuming with comparative reference to sports they prefer or activities they would rather be doing. The answer to “why should I consumer your sport” therefore is best framed in terms of that comparative framework (e.g., AFL cleverly positioned against competing sports in 1994 (www.youtube.com/watch?v=w2TO35rZjO4). Third, opportunities to improve engagement with non-fans exist, but are likely to be peripheral to the team (e.g., education, community programs).

In examining segments’ relative propensity to consume and think positively about a sport, three things stand out. Firstly, consumption itself is not a precursor to fandom. Viewership, and particularly attendance, foster positive attitudes (D. C. Funk et al., 2009), but infrequent and semi-regular attendance or viewership may not promote identification as a fan. This challenges the adopt/postpone/reject classification of most consumer adoption work, typically based on single consumption decisions (Rogers, 2010).

Social connections are important in determining the type of sport consumed (Wann et al., 2008), particularly across genders (James & Ridinger, 2002). More social connections with existing AFL fans encourage people to consume and have neutral or positive attitudes towards that sport; thus, teams supporting fan referral programs makes sense. In contrast, those without social contacts can more easily isolate themselves; the absence of AFL fans in their network removes a major motivator of attitude change (East et al., 2008). This leaves marketing communications as the main influencer, which these groups (e.g., Haters and Rival Coders) are likely to avoid or miss unless frequency and reach exist (Sharp, 2010).

Other segments consume and have social network connections to the sport yet are not committed attitudinally; they are not postponing adoption in the classic sense, nor rejecting it. Dick and Basu (1994) theorised that a segment of consumers could behaviourally engage but be psychologically uncommitted (i.e., spuriously loyal), but in practice, this is rare (e.g., Ngobo, 2017). Within team sports, spurious loyalty exists when a segment desires to connect socially (Connected) or likes the broad product category (Flaneurs), so may be more common. Sport’s importance as a conduit for social interactions highlights the criticality of designing and delivering venue and virtual experiences and services (Funk, 2017; K. L. Wakefield & Sloan, 1995) and players being role models.

Finally, we gained insight for each cluster’s propensity of becoming a fan. Haters cannot be converted, but negative attitudes could be softened through community engagement. Rival Coders are satisfied with another sports brand that commands most of their attention and spending, reinforcing their habits (Kleijnen et al., 2009) but will tune into marquee games in other sports because of their love for sport. Understanding what they like about their preferred sport and highlighting similarities with other sports could broaden their consumption, but lack of social connections remains the ultimate barrier.

Dabblers consume but are very low on fandom measures. Given this segment represents almost 30% of the non-fan market examined, a beneficial strategy may be to encourage their existing consumption level rather than escalate them. Campaigns could remind them of the sport as an entertainment option and ease consumption (e.g., easy ticket buying, game entry). Such efforts have successfully maintained “light users” in other categories (Sharp, 2010).

The Connected segment is attractive for conversion to fandom. Their behaviours are strong, so the challenge is building identification. They must transition from perceiving sport as something they attend for others’ pleasure to it giving themselves pleasure. It may be possible to educate this group to consider themselves fans by redefining “fan”; a simple campaign to acknowledge attendees or watchers of a game as “fans” who add value may shift attitudes and build identification. The value of these consumers seeing themselves as fans, even without behaviour change, is unknown.

Simplified classification of non-fans

Strategies for non-fans rely on identifying non-users easily. A criticism of recent complex segmentation approaches is their difficulty of application; hence, we present a simplified version of our non-fan segmentation process. Whilst sacrificing some accuracy, the four-question process can be easily used and analysed in most contexts (Figure 1). Based on their profiles, our clusters differ most in four variables:

Figure 1. Simplified decision tree for segmenting non-fans.

• Follow any professional sports (Yes/No)

• Level of passion for AFL (Rating/5)

• Ever watched an AFL game (Yes/No)

• Number of other sports followed (Count)

We tested multiple iterations of classification rules based on segment profiles and compared performance against segment membership for everyone based on the full cluster model. Table 7 displays the best performing classification rules and the proportion of individuals correctly identified per cluster. Blank cells mean a variable is irrelevant to a particular cluster.

Table 7. Simplified Classification process for non-fans.

	Follow any professional sports	Level of passion for AFL (/5)	Ever watched an AFL game	Number of other sports followed	% correctly identified
Cluster 1: Haters	No		No		81.7
Cluster 2. Dabbler	No		Yes		99.1
Cluster 3: Rival Coders	Yes	≤1.9			74.6
Cluster 4: Connected	Yes	>1.9		≤3	78.5
Cluster 5: Flaneurs	Yes	>1.9		>3	81.1

The correct cluster was identified for 82.3% of the individuals overall, and for at least 74% per cluster. This scheme represents a simple yet relatively accurate way to classify individuals into segments without collecting full survey data or applying LCA. A four-item scale is easily embedded in existing data collection via online ticket buying, commercial surveys and competitions/event registrations.

Conclusion

We defined non-fans as individuals who do not consider themselves fans, regardless of behaviour. We uncovered five unique segments within AFL non-fans that could be profiled using identifiable characteristics and propensity to become fans. The segment profiles suggest that increasing engagement with a specific sport is not continuous and episodic. Awareness is a precursor, but consumption (live or broadcast) can occur before positive affect or liking forms. While behaviours and attitudes usually correlate (Sharp, 2010), non-fan consumption over the years may never correspond to establishing identity with a sport or team. Consumption among non-fans is not driven by goal-directed behaviour but is involuntary and learned through wide distribution or a desire to socialise with others who are fans. For those inclined towards sports fandom, alternative sports or too many sports to follow can limit engagement. We also observed a segment of non-fans so staunch in their rejection of sport that they could form an anti-brand community. Typically, these anti-brand communities in a sport context have been formed by rival club fans (Popp et al., 2016), but in other fields like technology, they can be formed by those opposed to the category as a whole like our “Haters” (Brandão & Popoli, 2022).

We acknowledge that this study was limited to focusing on one sport. To generalise these segments, application to other major professional sports leagues and contexts would be beneficial. In addition, research on the relationship between segments and conditions for transition can provide insight into fan development. Awareness is normally the starting point for such models, with behaviour ranging from none to unplanned to coerced due to minimal and inconsequential attitude formation (e.g., PCM). Our findings suggest that within awareness, multiple segments and a progression exist, with potent attitudes forming and directing behaviour as a function of individual and sociological processes unrelated to the focal sport or team. While fan definitions typically refer to a strong positive emotional connection to an athlete or team (Wann & James, 2019) researchers should define and investigate non-fan psychology and how a connection fluctuates and/or weakens. Future research on the drivers and barriers to non-fan engagement could integrate work on social networks (e.g., Katz et al., 2018), content distribution and sharing consumers across sports brands (e.g., Fujak et al., 2018).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by the Australian Research Council [LP10010022].

Notes

¹ Throughout this paper, “sport” refers to the consumption of professional sport rather than participation in sport or exercise. Whilst there is a link between playing and watching (Wann et al., 1999), participation is a distinct activity requiring specific investigation (e.g., Baker et al., 2018).