The Impact of Status Threats on Sport Fans ...

1 downloads 0 Views 2MB Size Report
supremacy during Apartheid, and later (at the 1995 Rugby World Cup) helped ... football fans evaded assimilation into American sports by establishing local.
The Impact of Status Threats on Sport Fans’ Commitment and Collective Self-Esteem: A Social Identity Perspective

Gareth Hall

A submission presented in partial fulfilment of the requirments of the University of Glamorgan for the degree of Doctor of Philosophy

March 2009

i

Declaration

This work has not previously been accepted in substance for any degree and is not being concurrently submitted in candidature for any degree.

Signed…………………………………………… (Candidate)

Date…………………………………………….

STATEMENT 1 This thesis is the result of my own investigations, except where otherwise stated. Other sources are acknowledged by footnotes giving explicit references.

Signed…………………………………………… (Candidate)

Date…………………………………………….

STATEMENT 2 I hereby give consent for my thesis, if accepted, to be available for photocopying and for inter-library loan, and for the title and summary to be made available to outside organisations.

Signed…………………………………………… (Candidate)

Date…………………………………………….

ii

Certificate of Research

This is to certify that, except where specific reference is made, the work described in this thesis is the result of the candidate’s research. Neither this thesis, nor any part of it, has been presented, or is currently submitted, in candidature for any degree at any other University.

Signed

……………………………………… Candidate

Date

………………………………….......

Signed

……………………………………… Director of Studies

Date

………………………………………

iii

Acknowledgements I would like to thank the University of Glamorgan, and the Division of Psychology for giving me the opportunity to develop my career in Psychology, firstly through employment, and secondly, by allowing me to embark on a Ph.D. programme along side my employment. I am also grateful to all staff at Glamorgan who made every effort to accommodate my needs, and to the participants who without their consent the research would not have been possible. I would also like to thank my colleagues Dr Gareth Davies, Dr David Shearer, Dr Peter Mayer, and Ross Hall whom I worked with on the ESRC funded project, and contributed to aspects of my research in Chapter Three. I am also grateful to my supervisors, Professor Leo Hendry for his support and I am especially grateful to Dr Rachel Taylor who took over as a supervisor late in my research; I will forever appreciate her support, patience, and excellent guidance. Thanks must be attributed to Dr Rob Thomson who began this project as a supervisor and contributed towards the initial direction of the thesis. I would also like to thank Professor Nick Wheeler from the Department of International Politics at Aberystwyth Univeristy for his advice and support during my transition to Aberystwyth. A special thanks to the Department of Psychology at Aberystwyth for their support while I finished writing, and I am especially grateful to Dr Rachel Rahman for her advice and support. Above all, I would like to thank my family for their constant support, without their continuous encouragement throughout, this Ph.D. would not have been possible.

iv

Abstract Working within a social identity framework (Tajfel & Turner, 1979; Turner, 1999), this thesis aimed to advance the understanding of the construct commitment and collective-self esteem, and non-performance related social identity threats in sport fans. Chapter three describes the adaptation and psychometric validation of the collective self-esteem scale (Luhtanen & Crocker, 1992), and then outlines the development and validation of the sport fan commitment scale. Exploratory factor (principal components analysis) and confirmatory factor analysis showed that the construct validity of the collective self-esteem scale was was upheld but was best represented by a four factor correlated model and not a four factor hierarchical model as previously thought (cf. Luhtanen & Crocker, 1992). Mixed methodology using thematic analysis and both exploratory factor (using PCA) and confirmatory factor analysis revealed that the construct commitment could be reliably assessed by a 12 item, three factor correlated model, comprising affective loyalty, involvement, and BIRGing. Chapter Four utilises the new scales to examine how non-performance related threats to sport fans’ social identity. The results of this chapter suggest that fans derived a more positive social identity by being seen to belong to a better in-group (one that is high in commitment and loyalty) than when their social identity was defined by their team’s performances. Furthermore, threats to fans’ social identity led to stronger loyalty to their in-group than to their team (see studies eight and nine). The implications of the results are discussed in Chapter Five, which highlights how the new scales have advanced theory and understanding about the multidimensionality of sport fans’ commitment. Integration of the commitment scale with more general models of social identity (e.g. Cameron, 2003) in order to assess

v

its utility with other social groups is one direction identified for future research. Finally, the finding that team performance was not as important for fans’ social identity as initially thought, could be expanded through other novel methodologies (e.g. pre-game contexts and mixed method approaches). These directions for research might allow a larger range of fan behaviours to be examined, as well as enabling sport clubs to understand their consumers’ commitment further

vi

Contents Declaration

ii

Certificate of Research

iii

Ackowledgements

iv

Abstract

v

Contents

viii

List of Figures

xiii

List of Tables

xiv

List of Appendices

xv

Chapter One: General Introduction

1

1.1

Introduction

1

1.2

Groups and the social identity approach

3

1.3

The social identity perspective

8

1.4

Sport fans and the social identity approach

11

1.5

Social identity and sport fans’ self-esteem and commitment

16

1.6

Structure of the thesis

20

1.7

Conclusions

21

Chapter Two: Literature Review

23

2.1

Introduction

23

2.2

The Significance of sport and sport fandom

25

2.3

Social identity and sport fandom

27

2.3.1

The social identity perspective

29

2.3.2

Social identity and self-esteem

36

Image management in sport fans

40

Individual mobility

41

2.4 2.4.1

2.4.1.1 Basking in reflected glory

43

2.4.1.2 Cutting off reflected failure

45 vii

2.4.2

Social creativity

49

2.4.2.1 Attribution Bias

50

2.4.2.2 Changing the comparative dimensions

54

2.4.3

Social competition

56

2.4.4

Measuring self-esteem

61

2.5

Non-performance status threats and sport fans

65

2.6

Commitment and sport fans

75

2.6.1

Conceptualising commitment

77

2.6.2

Measuring sport fans’ commitment

84

Summary

89

2.7

Chapter Three: Adaptation and Development of two Multifactor models 3.1

Introduction

92 92

Section A:

95

Adapting the collective self-esteem scale 3.2

Introduction

95

Study one: Adapting the collective self-esteem scale

99

3.2.1

Method

99

3.2.2

Results

100

3.2.3

Discussion

111

Section B: Developing a commitment scale for sport fans

119

Introduction

119

Study two: Conceptualising commitment

122

3.3.1

Method

122

3.3.2

Results

124

3.3

viii

3.3.3

Discussion

125

3.3.4

Conclusion

140

Study three: Developing and examining the sport fan

141

commitment scale 3.4.1

Method

141

3.4.2

Results

147

3.4.3

Discussion

150

Study four: Confirmatory factor analysis of the sport fan

152

commitment scale 3.5.1

Method

152

3.5.2

Results

153

3.5.3

Discussion

162

Study five: Revising the sport fan commitment scale

163

3.6.1

Method

163

3.6.2

Results

164

3.6.3

Discussion

165

Study six: A test of criterion validity

166

3.7.1

Method

166

3.7.2

Results

169

3.7.3

Discussion

171

3.8

Section B discusiion

174

3.9

Chapter Discussion

177

ix

Chapter Four: The effects of in-group and team status

184

threats on sport fans’ self-esteem and commitment 4.1

Introduction

184

Study seven: Manipulating commitment and in-group status

189

4.2.1

Method

189

4.2.2

Results

192

4.2.3

Discussion

194

Study eight: Manipulating team status

199

4.3.1

Method

199

4.3.2

Results

200

4.3.3

Discussion

202

Study nine: Manipulating team performace

207

4.4.1

Method

207

4.4.2

Results

209

4.4.3

Discussion

210

Chapter Discussion

213

4.4

4.5

Chapter Five: General Discussion

218

5.1

Introduction

218

5.2

Overview of research aims

218

5.3

Team performances are not the most valued comparative dimension for sport fans’ social identity, self-esteem, and committment

x

220

5.3.1

Commitment in sport fans differs to the conceptualisation

222

proposed by social identity 5.3.1.1 Introduction

222

5.3.1.2 Overview of findings

223

5.3.1.3 Theoretical implications and methodological considerations

226

5.3.2

The collective self-esteem scale was affected by adapting the scale for specific use with sport fans

230

5.3.2.1 Introduction

230

5.3.2.2 Overview of findings

231

5.3.2.3 Theoretical implications and methodological considerations

232

5.3.3.

Non-performance related threats affect fans’ collective

239

self-esteem and commitment? 5.3.3.1 Introduction

239

5.3.3.2 Overview of findings

240

5.3.3.3 Theoretical implications and methodological considerations

242

5.3.4

Summary of research

248

5.4

Further methodological considerations for future directions

251

5.5

Sumamry of thesis

258

5.5.1

Conclusion

261

References

264

Appendices

299

xi

List of Figures

1

Example of a one factor hierarchical model

106

2

Example of four factor uncorrelated model

106

3

Example of four factor correlated model

107

4

Example of a four factor hierarchical model

107

5

Thematic map of sport fans’ commitment

126

6

Examples of the status manipulations

191

xii

List of Tables

1

Mean scores for the Collective Self-Esteem scale

101

2

Alpha scores for the Collective Self-Esteem scale

102

3

Correlations for the adapted Collective

103

Self-Esteem scale 4

Factor loadings for the adapted Collective

105

Self-Esteem scale 5

Goodness of fit indices for the Collective Self- Esteem

110

scale 6

Factor loadings for the 19 item Sport Fan

149

Commitment scale 7

Factor loadings for the 29 item Sport Fan

154

Commitment scale 8

Mean and alpha scores for the Sport Fan

155

Commitment scale 9

Correlations for the Sport Fan Commitment scale

156

10

Factor loadings for the 16 item Sport Fan

160

Commitment scale 11

Goodness of fit indices for the Sport Fan

161

Commitment scale 12

Criterion validity for the Sport Fan Commitment scale

171

13

In-group mean scores

193

14

Team status mean scores

201

15

Team performance mean scores

209

xiii

List of Appendices

A

Sport Spectator Identification scale

300

(Wann & Branscombe, 1993) B

Psychological commitment to team scale

301

(Mahoney, Madrigal, & Howard, 2000) C

Example of a three factor plus one model of commitment

303

D

Original, reworded, and additional items for the 19 and 29

304

item Sport Fan Commitment scale

E

Example of status manipulations

306

F

Examples of questions used in study nine

310

G

Further fan quotes

311

xiv

Chapter 1 General Introduction

1.1 Introduction The groups we belong to have an important influence on how we define ourselves in relation to others, and our behaviours in groups are suggested to be more extreme and depersonalised than when we act individually (Oldmeadow, Platow, Foddy, & Anderson, 2003). Of the theoretical frameworks and research explaining our group memberships and intergroup relations (e.g. Realistic conflict theory, Levine & Campbell, 1972; Sherif, 1966, and the common in-group identity model, Gaertner, Dovidio, Anastasio, Bachman, and Rust, 1993), the social identity approach, consisting of both social identity theory (SIT; Tajfel, 1978; Tajfel & Turner, 1979; 1984) and self-categorisation theory (SCT; Turner, 1985; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987), has contributed the most comprehensive account of our membership within social groups, attracting a wealth of research in social psychology and related disciplines (Hogg, Abrams, Otten, & Hinkle, 2004). In particular, this theoretical framework places emphasis on interactions between perceptions of our social selves and the social and contextual factors that enhance or diminish the salience of our personal and social identities (Ellemers, Spears, and Doosje, 2002; Hogg & Williams, 2000). Given the wealth of research and scholarly impact the social identity perspective has had on research about membership in social groups, is one reason it is used as the theoretical framework for this thesis. In this introductory chapter, the focus will be on providing an outline of the thesis’ aims and main arguments, and explaining how these extend the literature based on social identity traditions, as well as extending the social psychology 1

literature about sport fandom. Working within a social identity framework, the first aim of this thesis is to examine the construct of commitment in two different but related ways. These are first to examine how commitment may be conceptualised and measured, and second, to examine how different types of social identity threats to the groups’ value affect group members’ commitment. A second aim of the thesis is to examine the problematic measurement of collective self-esteem in terms of the underlying factor structure. These aims are achieved in two research chapters, Chapters Three and Four, and the contributions of the research are discussed in Chapter Five. Furthermore, the aims of the thesis are examined within the novel context of sport fans since they are an appropriate and interesting social group for investigating social identity processes (Donovan, Carlson, & Zimmerman, 2005, Ellemers, Spears, & Doosje, 1997), such as commitment and self-esteem (e.g. Abrams & Hogg, 1988; Ellemers, Kortekaas, & Ouwerkerk, 1999). Sport fans are particularly appropriate for the current thesis given that being highly committed has been identified as an important feature of this social group’s members (Wann, 2006b; Wolfson, Wakelin, & Lewis, 2005), as well as self-esteem playing an important part in determining sport fans’ reactions to their team’s performances (Wann, 2006b). Furthermore, they are often used as an example of a real-world group that display extreme depersonalised behaviours (e.g. Levine, Prosser, Evans, & Reicher, 2005), a key assumption of social identity theory. The following section (1.2) will provide an overview describing the types of groups previously used in a social identity framework. More specifically this will examine how sport fans as an example of a self-selected group may provide a novel and interesting approach to examining social identity. Section 1.3 provides the reader with an overview of the conceptual framework of the social identity tradition

2

employed in this thesis. In section 1.4, attention will be turned to literature examining sport fans. This will focus on the limitations of using team performance as the primary form of status threats to sport fans’ social identity. Section 1.5 outlines the contrasting conceptualisation and measurement of commitment provided by a social identity approach with literature examining commitment in sport fans. Following this, the importance of self-esteem in explaining sport fans’ reactions to status threats dominates the research about sport fans, but at the same time, their self-esteem has been measured inconsistently. Therefore, how selfesteem can be measured more accurately in contexts where commitment is an important feature of sport fans is also discussed in section 1.5. Finally, the last sections (1.6 and 1.7) present the structure of the thesis followed by the principal research questions to be examined.

1.2 Groups and the social identity approach Despite the impact on understanding group processes that the social identity tradition has had, there are still several issues. One broad issue is concerned with the ecological validity of the framework in terms of the type of groups used to examine social identity processes (Brown, 2004; Ellemers, Kortekass, & Ouwerkerk, 1999). This is because much of the research within the social identity tradition has utilised minimal group paradigms (Reynolds, Turner, & Haslam, 2000). However, examining social identity processes with a broad range of social groups would provide a better understanding of intergroup differences in terms of social identity processes. Examining different types of groups is also of importance considering that one indispensable element of social identity theory was to explain intergroup

3

relations and conflict between different social groups (Turner, 1999; Turner & Reynolds, 2001). In traditional minimal group experiments, members are assigned to a high or low status group based on their competence in some type of group value task. Groups created in these situations are unlikely to have much realism, entitativity, or emotional value (Bar-Tal, 2004; Gordijn, Yzerbyt, Wigboldus, & Dumont, 2006; Lalonde, Moghaddam, & Taylor, 2001; Stangor, 2004), and may subsequently have little social identity value for group members (Ellemers et al, 1999b; Lickel, Hamilton, Lewis, Sherman, Wieczorkowska, & Neville Uhles, 2000; Stangor, 2004). Furthermore, artificial groups only have categorisation processes in which to guide them in terms of their group membership. For example, Ellemers et al, (1999b) suggested that this might not be the case for real world groups. Instead, they suggested that categorisation might be the only social cue for individuals to make sense of their reality and guide their behaviour in an experimental situation. In real world groups, however, individuals may acknowledge their social category but may feel impartial, uncommitted to their group, and reluctant to behave in terms of that group’s norms. Rather, individuals may prefer to belong to other groups or categorisations they feel better define them. As Ellemers and her colleagues point out, experimental manipulation of categorisation into natural groups may be one reason for inconsistent findings for social identity processes (e.g. intergroup discrimination). Therefore, group members allocated to a group might not engage in prototypical behaviour as much as members of self-selected groups (e.g. individual and voluntary involvement such as sport fans or religious groups; see Deaux, 1995 for a review). Therefore, as outlined above, there are clear justifications for using a real world group in this thesis to examine social identity processes.

4

Although research about social identity processes has been conducted using real world groups, this research is generally limited to ascribed groups such as gender and ethnicity, or groups in organisational contexts (e.g. Cicero, Pierro, & van Knippenberg, 2007; Pfeifer, Rubble, Bachman, Alvarez, Cameron, & Fuligni, 2007). For example, in organisational psychology literature, work groups are often used as a good example of a self-selected group (e.g. Bergami & Bargozzi, 2002). However, it could be argued that they are not a genuine self-selected group since the motivations for joining work groups are different (e.g. earning a salary) to those of other selfselected groups, such as sport fans (e.g. group affiliation and self-esteem, see Wann, 1995 for a review of motivations). This, in addition to the reasons outlined in the previous

paragraph

(i.e.

categorisation,

psychological

realism,

emotional

attachment), is another reason for a growing interest in favouring voluntary selfselecting groups (e.g. Deaux, 1996; Deaux & Ethier, 1998; Ellemers et al, 1999b; Wann & Hamlet, 1995; Wann & Grieve, 2005). These types of social groups have become of interest to social identity researchers because of the voluntary nature of joining them and the emotional attachments that members assign to them. The amount of emotional attachment is an important component of a social identity in that it is the extent to which group members identify (termed amount of social identification, Brown, Condor, Mathews, Wade, & Williams, 1986; Ellemers et a, 1999b; Ouwerkerk, Ellemers, & De Gilder, 1999) with their group that determines whether they act and think in terms of their group’s norms (also known as referent normative influence) or remain as individuals (Hogg & Williams, 2000). Specifically, social identities are seen to be multifaceted consisting of three components: a cognitive (self-categorisation), evaluative (group self-esteem), and affective (affective commitment) component (Tajfel & Turner, 1979). However,

5

when measuring identification researchers tend to use identification scales in terms of the affective component, and hence, often confuse social identification content and amount of ‘affective’ social identification (Ouwerkerk, Ellemers, & De Gilder, 1999) The amount of social identification an individual has with their social group is therefore often described in social identity literature as ‘affective’ commitment (Ellemers, 1999, Ouwerkerk, Ellemers, & De Gilder, 1999). Sport fans for example, are a prominent real world social group who emphasise their affiliation with fellow fans and their team (Wann, Melnick, Russell, & Pease, 2001). Sport fans also have strong emotional attachments to their team, and for the most part, the team they support is an important part of their self-concept (Shank & Beasley, 1998). In other words, in terms of a social identity perspective fans are likely to have strong affective commitment. In general, strong emotional attachments and high identification to a social group have been found to produce greater self-definition and self-stereotyping in terms of our social identity, as well as escalated and continued commitment from group members (Ellemers, et al, 1999b, Haslam, Ryan, Postmes, Spears, Jetten, & Webley, 2006; Hutchison, Jetten, Christian, & Haycraft, 2006). This is because highly committed group members should experience, according to a social identity approach, a higher degree of depersonalisation (Ellemers et al, 2002; Hogg & Terry, 2000; Turner et al, 1987). The process of depersonalisation allows behaviours (such as escalated commitment and self-stereotyping) to take place because highly committed group members diminish their personal identities in favour of the group’s identity when that social identity is made salient. Like other social identities, sport fans should also diminish their personal identities in favour of the group’s identity, particularly when fans are highly committed and are at a game.

6

These processes apply even if highly committed members perceive their group as having a negative social image and the task is considered a ‘lost cause’ (e.g. DietzUhler, 1996; 1999; Haslam, et al, 2006; Mlicki & Ellemers, 1996). Sport fans for example, regularly display such commitment since many fans continue supporting their team even after a loss (Ellemers, Spears & Doosje, 1997). Although sport fans are clearly an appropriate and interesting social group to examine social identity principles (cf. Ellemers, Spears, & Doosje, 1997; Murrell & Dietz, 1992; Wann & Polk, 2007), the use of sport fans has largely been neglected within social identity literature (Wann & Hamlet, 1995; Wann et al, 2001). Instead, social identity researchers have favoured minimal group paradigms and assigned laboratory groups (Ellemers et al, 1999b), which are low in entitativity and ecological validity. Therefore, there is a clear gap in the literature in terms of studying social identity processes within a relatively novel context. In addition to being an interesting social group to examine, sport fans are a self-selecting group, where despite fans not having any formal rules and regulations for joining or leaving, group members join voluntarily and can develop lifelong attachments with their team that can begin in childhood and strengthened through family traditions (Wann, Melnick, Russell, & Pease, 2001). The recent importance given to self-selected groups over minimal group paradigms in social identity research (Ellemers et al, 1999b) further supports the use of sport fans as the population to study social identity processes (i.e. commitment and collective selfesteem) in this thesis. This therefore extends previous social identity literature by examining these processes in a novel and interesting context, and beyond the traditional minimal group paradigm. However, before presenting previous research

7

about sport fans, it may be beneficial for the reader to have a broader overview of the social identity perspective, and therefore this is the focus of the next section.

1.3 Social identity perspective

A social identity perspective views our social groups as distinct identities from our personal identity and idiographic attributes. This perspective also suggests that we use our groups as psychological self-referential representations to define, evaluate, and change our behaviours in terms of our group memberships (Abrams, Wetherell, Cochrane, Hogg, & Turner, 1990; Hogg & Abrams, 1988; Tajfel & Turner, 1979; Turner & Reynolds, 2001). Most research treats identification a single component that indicates a general connection to a social group (Leach, van Zomeren, Zebel, Vliek, Pennekamp, Doosje, Ouwerkerk, & Spears, 2008). According to a social identity perspective, however, our social identities are multidimensional in nature, consisting of a cognitive, emotional, and an evaluative component (Cameron, 2004; Leach et al, 2008; Obst & White, 2005; Tajfel & Turner, 1979). These components equate to measuring identification in terms of self-categorisation, affective commitment, and group self-esteem respectively, but researchers tend to measure identification in terms of affective commitment (Ouwerkerk, Ellemers, & De Gilder, 1999). More recently, these components have been subsumed into a two dimensional hierarchical model that distinguishes between self-involvement and self-definition, where self-involvement encompasses the emotional and evaluative components (i.e. affective commitment and collective self-esteem) of social identity (Leach et al, 2008). Important to this perspective is that our identities change quickly in response to contextual changes, not only in terms of which identity is most salient, but also in 8

terms of what form of relevance the identity will take (Hogg, Abrams, Otten, and Hinkle, 2004). In addition to gaining self-definition, we also gain much of our selfesteem from our membership in social groups (Tajfel & Turner, 1979; Dietz-Uhler & Murrell, 1999; Stangor, 2004; Utsey & Constantine, 2006). Furthermore, the positive and negative effects of our memberships on our self-esteem has also been well documented in social identity literature (see Abrams and Hogg, 1999; Branscombe & Wann, 1994; Ellemers, Spears, & Doosje, 1997; Foels, 2006; Luhtanen & Crocker, 1992; Turner, 1999; Turner & Reynolds, 2003). In its original conception as social identity theory, it was suggested that people strive for positive self-esteem via a positively valued social identity. This was achieved by making favourable social comparisons with relevant out groups (Tajfel & Turner, 1979; Hogg & Abrams, 1988; Ellemers, Van Knippenberg, De Vries, & Wilke, 1988). Group members do this in part by showing in-group bias, particularly when the status position of the group is considered low in comparison to a relevant higher status group (Branscombe & Wann, 1994; Ellemers, Spears, & Doosje, 1997; Hewstone, Rubin & Willis, 2002; Houston, & Andreopoulou (2003); Van Knippenberg & Ellemers, 1990; 1993). More recently, self-categorisation theory (SCT; Turner et al, 1987) elaborated on and complemented the underdeveloped cognitive aspects of social identity theory (i.e. categorisation processes). For example, social identity theory views our identities as existing on an interpersonal-intergroup continuum and suggests that we shift along this between personal identities (low social identity salience) and contextually relevant social identities (high social identity salience) (Abrams et al, 1990; Haslam, 2004; Hogg, 2000; Turner & Reynolds, 2003). In contrast, according to self-categorisation theory, the shift from personal to social

9

identity is underpinned by the cognitive mechanism of categorisation. Rather than shifting along a continuum, categorisation switches on or off a social identity, which then allows intergroup behaviour to take place (Turner, 1982). Self-categorisation theory, therefore offered a more pertinent cognitive basis and explanation of group behaviour, as well as a more conclusive process of how the transformation from personal to a social identity occurs (Hogg & Terry, 2000; Hornsey & Hogg, 2000; Turner & Reynolds, 2001). The categorisation process underpinning selfcategorisation theory produces group prototype based depersonalisation. This process represents a change in self conceptualisation and perceptions of others using relevant stereotypes that define both the in-group and out-group, while at the same time differentiating the in-group from a comparative out-group (Abrams et al, 1990; Haslam, 2001; Hogg, 2001; Hogg & Terry, 2000; Turner et al, 1987; Oakes, 2002). As such, a social identity perspective suggests we are motivated to identify with groups, first to reduce uncertainty of our self-concept (the uncertainty reduction hypothesis, Hogg, 2000; Hogg, 2001; Hogg & Mullin, 1999), and second to enhance our self-concept (Reid & Hogg, 2005; Tajfel, 1986; Tajfel & Turner, 1979). These motives will have the greatest benefits to our self-esteem and selfconcept (e.g. the self-esteem hypothesis, Abrams & Hogg, 1988; Hogg & Abrams, 1990). Groups therefore strive to be both better than and distinct from other relevant out-groups, competing for favourable differentiation, higher status, a positive social identity and positive self-esteem for their members. The relative status of the group is considered to reflect the outcome of contextually relevant intergroup comparisons (Tajfel & Turner, 1979; Hewstone et al, 2002) using any identity relevant comparative dimension (Ellemers, Spears, and Doosje, 2002; Oakes, 2002). To elaborate on the principles outlined above, the following section will highlight how

10

social identity is applicable to sport fans specifically, and how previous research has examined sport fans’ behaviours. In particular, the role of social identity threats is discussed in terms of the sport team’s performances and the consequences that performances as the main status-defining comparative dimension has on fans’ social identity.

1.4 Sport fans and the social identity approach

In addition to reasons outlined in section 1.2 (i.e. self-selection, strong emotional attachments, depersonalisation) a social identity approach is applicable to sport fans for several other reasons. This is because sport fandom, like other social group memberships, involves intergroup comparisons with rival teams and their fans, has well defined group norms (e.g. wearing apparel, codes of conduct in the terraces), and includes a shared common fate that determines the group’s relevant status by team results (Donovan, Carlson, & Zimmerman, 2005). For example, intergroup

relations

(e.g.

rivalries),

categorisation,

group

status,

and

depersonalisation are highly salient for sport fans at a game. In addition to depersonalisation, few other groups show as much deindividuation (Mann, Newton, Innes, 1982), and this has been reported to accentuate member’s social identity further (Reicher, Spears, & Postmes, 1995; see also Klein, Spears, & Reicher, 2007). Moreover, if fans see themselves and their team as one group, they should also be subject to many of the same social identity processes as another social group (Murrell & Dietz, 1992; Sloan, 1979; Wann & Polk, 2007). For example, after a loss sport fans of a specific team would experience threats to their group’s value. In other words, the positivity of their social identity is threatened. Furthermore, the threat to their group’s status is experienced on a weekly basis at games. The status 11

of their group however, is controlled externally by their team’s performances. Sport fans, unlike other groups (ascribed or assigned) cannot control the outcome or their group’s status when the status-defining dimension is based on team performance. As such, fans cannot influence the positivity of their social identity and therefore their self-concept. In a similar manner to other types of low status group members, low status sport fans (fans of a losing team) have reported negatively affected emotions and self-esteem, increased feelings of personal failure and showed intergroup discrimination (Branscombe & Wann, 1994; Crisp, Heuston, Farr, & Turner, 2007; Hirt, Zillman, Erickson, & Kennedy, 1992). A social identity approach postulates that we prefer to belong to groups that provide us with a positive social identity. The dimensions used for social comparison by group members to evaluate their social identity are of particular importance. This is because the dimensions used provide group members with information about their relative or perceived status. Social identity threats are experienced when a less favourable comparison is made with a relevant out-group on whichever comparative dimension is contextually relevant and salient (Ellemers et al, 2002). Since these dimensions are status-defining, a less favourable comparison results in a lower group status and negative social identity. This will also reflect upon an individual at a personal level resulting in a negative selfconcept. For example, Hirt et al, (1992) and Murrell & Dietz (1992) were some of the first studies to provide evidence about the negative consequences of a losing team performance on fans’ personal feelings, self-esteem, and self-worth. Recently, researchers (e.g. Branscombe, Ellemers, Spears, & Doosje, 1999, 1999; Ellemers et al, 2002; Wann & Grieve, 2005) distinguished between several types of social identity threats to group members. These researchers argued that the

12

dimension that is used for social comparison depends on the social context as well as the salience and characteristics of the group membership (see also; Ellemers, Barreto, & Spears, 1999; Ellemers, van Dyck, Hinkle, & Jacobs, 2000). For sport fans, their social identity is particularly salient at a game and the status threat is traditionally to the group’s value. The fans’ group value is evaluated via the consequence of team performances and has been the primary status-defining dimension in social identity and sport fan literature (e.g. Hirt et al, 1992; Murrell & Dietz, 1992). A win boosts their status and therefore the positivity of fans’ social identity, while a loss results in a negative image and lower status for fans. Furthermore, such outcomes on sport fans’ image can influence whether they decide to remain with, or leave their team; in other words, it can help distinguish between highly and lowly committed fans (Cialdini, Borden, Thorne, Walker, Freeman, and Sloan, 1976). Despite research (e.g. Branscombe, Ellemers, Spears, & Doosje, 1999) suggesting that fans may experience several types of threats, research on sport fans has remained focussed on team performances as the most obvious form of social comparison and social identity threat (e.g. Bernache-Assollant, Lacassagne, & Braddock, 2007; Crisp, Heuston, Farr, & Turner, 2007; Hirt et al, 1992; Murrell & Dietz, 1992). This is problematic since fans cannot control this dimension, and therefore we have limited understanding about how fans might react to other more controllable threats suggested by the social identity literature. Given that research has neglected other types of comparative dimensions that fans may find threatening to the status of their social identity, another gap in the literature has therefore been identified that the current research will explore. More specifically, it is intended to focus on the role of ‘non-performance’ related status threats to the group’s value. In

13

addition, since previous methodology has used team performances as the main social identity threat, previous research is limited to being conducted post game. In contrast to previous research about sport fans, different types of social identity threats are examined pre-game as well as using team performance in non-match day context. Therefore, the current thesis provides an interesting context for examining other status-defining dimensions that fans may use to secure a favourable differentiation over rival fans. Subsequently, investigating other types of social identity threats raises another issue. If other status-defining dimensions are used by fans for social comparison, then do fans value these dimensions as much as team performance for providing a positive social identity? In contrast to prior research, the question of whether fans value status-defining dimensions other than team performances is addressed in this thesis. One way this will be achieved is through assessing fans’ self-esteem when exposed to different types of threats. If fans value a specific dimension for comparison, then when that dimension is manipulated negatively, it is hypothesised that fans will experience a social identity threat and therefore lowered self-esteem (cf. Murrell & Dietz, 1992; Crisp et al, 2007). In particular, it is argued that aligning perceptions and placing emphasis on the perceived in-group’s value in terms of commitment and loyalty, will be another way fans enhance their social identity independent of their team’s success. One reason for this is that while fans are unable to control how well their team performs and therefore their subsequent status, they can control how they are perceived as fans of a specific team. For example, as many fans continue to support their team after a loss, it is unsurprising that some research has shown fans are keen to accentuate their commitment and worth as a die-hard fan (Wann & Branscombe,

14

1990; Wann & Branscombe, 1993; Wolfson, Wakelin & Lewis, 2005). Therefore, it is argued that commitment is seen as a key attribute for being perceived to be a ‘good’ sport fan. As such, for sport fans it may be more important to be seen by others as good supporter (particularly after a loss), rather than to be seen as a supporter of a good team, and therefore forms a key proposal in this thesis. Following this proposal, it could be argued that teams’ performances are a less important status-defining dimension for fans than previously thought. Instead, as an example of a non-performance related threat, fans may use ideological attributes (such as being committed) to evaluate their worth as a good fan in order to achieve favourable comparisons over rival fans. Paradoxically, fans would therefore require their team to lose in order to be able to demonstrate their commitment and worth as a fan of a specific team. Such behaviour would then reflect positively on their social group, and thus result in them belonging to a positive social identity. This notion may be because fans have control over their image but not over their team’s performance. If fans have control over their social group’s image, then fans might feel more responsible for their own social identity. In addition, this type of non-performance related threat might be more internalised and salient than team performance. Fans therefore, may be more compelled to enhance their group’s image than when they experience threats because of their team’s poor performances. Subsequently, fans may be likely to use dimensions that they can manipulate such as their perceived amount of commitment and other fan behaviours in order to enhance their social identity. Another reason that threats from team performances may not be as salient as other comparative dimensions is because fans have been reported as being able to deal with their team’s losses effectively (Murrell & Dietz, 1992). This is because

15

sport fans may use several image management strategies proposed in social identity theory. These include strategies such as individual mobility (e.g. distancing i.e. ‘they lost’), social creativity (e.g. attributional biases i.e. ‘the refereeing was poor’), or social competition (e.g. out-group derogation) (Tajfel & Turner, 1979; but see also Wann, 2006a for a review). It is therefore argued that team performance may not necessarily be the primary source for social comparison as suggested in previous research or be as important for a fans’ social identity. Instead, performance may be an additional source of social identity enhancement, or one of many others that have not yet been identified. Exploring different types of social identity threats in sport fans provides an interesting and novel context in which to provide new perspectives on the social psychological and social identity aspects of sport fandom and self-selected groups. The examination of non-performance related comparative dimensions is very limited, where only one study to date has attempted to address this issue (see Wann & Grieve, 2005).Therefore, this thesis explores a clear gap in the literature about sport fans and contrasts with previous literature that has used team performance as a social identity threat.

1.5 Social identity and sport fans’ self-esteem and commitment

The influence of self-esteem on commitment to groups when group status is threatened has been well documented within social identity literature (e.g. Andreopoulou, & Houston, 2002; Brown, 2000; Houston & Andreopoulou, 2003; Long, Spears, & Manstead, 1994; Luhtanen & Crocker, 1992; Spears, Doosje, & Ellemers, 1997). Similarly, the importance in managing self-esteem and a positive social identity has dominated the sport fan literature, especially in relation to team 16

performances (Bizman & Yinon, 2002; Cialdini, Borden, Thorne, Walker, Freeman, & Sloan, 1976; Dietz & Murrell, 1999; End, Dietz-Uhler, Harrick, Jacquemotte, 2002; Hirt et al, 1992; Murrell & Dietz, 1992; Wann, Walker, Cygan, Kawase, & Ryan, 2005). However, since previous research examining sport fans is limited, there is consequently a lack of suitable measures to assess fans’ commitment and self-esteem. Therefore, a further important aim of this thesis is to develop robust psychometric measures for these constructs within the contexts of sport fandom. As well as addressing the lack of adequate measurement tools, this aspect of the research will contribute to other problems with these constructs in generic social identity literature. Measures of self-esteem for example, have not been administered in the way that social identity intended (e.g. collective self-esteem, see Luhtanen & Crocker, 1992; Rubin & Hewstone, 1998). The importance of distinguishing between personal and collective self-esteem in group members is often cited as one reason for inconsistent findings for the motivational role of self-esteem (more specifically within social identity theory) and the directionality of cause and effect of in-group bias

(Brown, 2004; Capozza & Brown, 2000; De Cremer, 2001;

Hewstone, Rubin, & Willis, 2002; Utsey & Constantine, 2006). However, where research has implemented a collective conceptualisation of self-esteem (most notably the Collective Self-Esteem scale, CSES, Luhtanen & Crocker, 1992), it has not been administered in a consistent manner and may therefore contribute to inconsistent findings such as the relationship between in-group bias and self-esteem (Aberson & Howanski, 2002; Houston & Andreopoulou, 2003). These inconsistencies have included using only one subscale to assess individuals’ overall collective self-esteem, or use several items from the whole scale to assess collective self-esteem. Since sport fans are an appropriate social group to assess social identity

17

processes, this research clearly illustrates the importance of using an adequate measure of self-esteem appropriate to the context and social group. To address the issues outlined in the previous paragraphs, the complete Collective Self-Esteem scale was employed in this thesis. More specifically, the Collective Self-Esteem scale has not been adapted for specific use with sport fans. Where it has been adapted for other specific social groups, its psychometric structure has not been examined beyond tests of reliability (e.g. Crocker, Blaine, & Luhtanen, 1993; Verkuyten, 1997). Despite this, tests of reliability have generally found that the scale remained reliable after adaptation. However, recent reexamination of the scale (e.g. Utsey & Constantine, 2006) using the same advanced procedures (i.e. confirmatory factor analysis) as in its development, have recently questioned its psychometric properties, such as the scales construct validity. However, Utsey and Constantine’s research, did not adapt the scale for their population (based on ethnicity), and hence did not fully address the issue. As other research has pointed out, real world social groups differ qualitatively in characteristics and members’ attachment to artificial and ascribed groups (Deaux, 1996; Deaux & Ethier, 1998; Deaux, Reid, Mizrahi, & Ethier, 1995; Ellemers et al, 1999b; Stangor, 2004). Individuals in specific self-selected groups may not respond to items about self-esteem in the same way as individuals in artificial groups. In other words, it might be that group members from different groups gain self-esteem from their social groups in different ways (e.g. some subscales may be more salient in different groups under different circumstances). Consequently, this would provide a stronger necessity for more contextually specific measures. In section 1.2 for example, it was discussed that artificial groups may be less emotionally meaningful for group members. Therefore, members of ascribed groups might not

18

gain self-esteem from their membership in these groups in the same way as perhaps members of self-selected groups would. Regardless of how members of these groups gain self-esteem, level of identification has been reported to be a key moderator of which group members (e.g. high or low identification) benefit from evaluations about their group (Andreopoulou, & Houston, 2002; Hewstone, Rubin & Willis, 2002; Houston & Andreopoulou, 2003; Rubin & Hewstone, 1998; Wann, 2006b). If group members are not highly identified then they are unlikely to benefit from belonging to a positive social identity. Therefore, in contrast to previous research using artificial or ascribed groups to assess collective self-esteem, this thesis explores the construct of collective self-esteem in a novel way by using a selfselected group. The final issue addressed in this chapter is the limited conceptualisation of commitment in both social identity traditions and sports fan literature. Commitment is assessed in social identity using unidimensional measures (e.g. Brown, Condor, Mathews, Wade, & Williams, 1986; Wann & Branscombe, 1993): that is using measures assessing our affective ties with the group (termed ‘affective commitment’; Bergami & Bagozzi; 2002; Ellemers et al, 1999b; Ellemers et al, 2002; Ouwerkerk, Ellemers, De Gilder, 1999; Tajfel & Turner, 1979). In contrast, evidence in related fields of psychology has suggested that commitment is multidimensional (Allen & Meyer, 1990; DeLobbe & Vandenberghe, 2000; Meyer, Becker, & Vanderberghe, 2004; Meyer, Becker, & van Dick, 2006). As such, our current conceptualisation of fan commitment within a social identity approach is limited. In addition, there is also a paucity of measures in social psychology that assess fans’ commitment. Moreover, generic sport specific measures of fans’ attachments (usually from marketing disciplines; see Funk & James, 2004

19

Mahoney, Madrigal, & Howards, 2000) have also been reported to have less than suitable psychometric properties (Dimmock, Grove, & Eklund, 2005; Kwon & Trail, 2003). Therefore, in the following series of studies it is intended to develop a psychometrically robust model of sport fans’ commitment that is derived from both literature and what fans regard as qualitatively important to their commitment. As fans may potentially use strength of commitment as a comparative dimension to achieve superiority over out-groups, it is important that measures that intend to assess their commitment do so in ways that truly reflect fans’ perceptions of commitment. By adopting this methodology, the current thesis contrasts with the majority of measures that are constructed using literature only to design measurement scales (e.g. Wann & Branscombe, 1993).

1.6 Structure of the thesis

Following this introductory chapter, Chapter Two reviews existing literature about the social psychology of sport fandom and relates this to a social identity perspective. The review first provides a global context sport by considering the impact of sport fandom in society and then examines the role of social identity in understanding sport fans’ attachments with their team. The review then explores how self-esteem and commitment have been previously investigated within this self-selected social group. Specifically, this will include the impact of team performances on fans’ self-esteem where particular attention will be given to the conceptualisation of commitment from a social identity perspective and other related fields of psychology. This will consider the nature of commitment and how it has been conceptualised and measured.

20

Two research chapters then present the series of studies that have been conducted. Chapter Three specifically addresses the development of scales used in later studies. In the first study, the Collective Self-Esteem scale is adapted for use with sport fans (providing additional psychometric scrutiny to this inconsistently used scale). This chapter then outlines four studies detailing the construction, development, and factorial validation of a multidimensional scale of sport fans’ commitments. Chapter Four contains three studies using the two new measures of commitment and collective self-esteem under experimental conditions. The objective of this chapter is to investigate non-performance related comparative dimensions that may threaten fans’ social identity other than team performances. Finally, Chapter Five provides a discussion of the results and analysis of the entire research programme, discussing extensions to the literature and subsequent implications. The chapter closes with a discussion about the limitations of the thesis and potential directions for future research to build on positively.

1.7 Conclusions

Given the overview of the literature in the previous sections, there are several objectives that this research intends to achieve. The first of these is to identify status-defining dimensions that are potential sources of threat to sport fans other than team performances. Secondly, the Collective Self-Esteem scale developed by Luhtanen & Crocker (1992) will be assessed and adapted. Thirdly, a robust measure of commitment for use with sport fans based on their own perspective will be developed. Finally, this research offers further ecological

21

validity to the social identity perspective by using sport fans as a novel example of a real world self-selected group. Based on these objectives, and the literature outlined in the previous sections, the overall research question to be examined is as follows:

Are teams’ performances the most valued comparative dimension for sport fans’ social identity, self-esteem and commitment?

More specifically, this has led to the development of several research questions in helping to answer the overarching research question.

Research Question 1: What does ‘being committed’ mean to sport fans, and how does this differ from the conceptualisation proposed by a social identity perspective?

Research Question 2: How is the global Collective Self-Esteem scale influenced by adaptation for specific use with sport fans?

Research question 3: How do different types of status threats affect fans’ collective self-esteem and commitment?

Research questions one and two were addressed first in six studies in Chapter Three while the remaining research question was addressed in three studies in Chapter Four.

22

Chapter 2 Literature Review

2.1 Introduction

As outlined in Chapter One, this chapter reviews the social psychological research conducted with sport fans and how this relates to social identity principles, such as social identity threats, group commitment, and self-esteem. In the first section (2.2), the impact of sport on our society is briefly reviewed. In section 2.3, the psychological research examining sport fan identification from a social identity perspective is discussed. This section further elaborates on the role of self-esteem within social identity and how supporting a team affects fans’ self-esteem. In the following section (2.4), the role of self-esteem is expanded in greater detail and this section focuses specifically on the image management strategies that fans utilise to protect their self-esteem and social identity. Since self-esteem is an important factor in social identity and sport fan research, the next section (2.5) discusses how past research has conceptualised and measured self-esteem. Because recent research has shown that group members can experience several types of threats to their social identity (Ellemers, Spears, & Doosje, 2002), section 2.6 discusses how status threats other than team performances, may be experienced by fans, and how they also threaten their social identity and self-esteem. Finally, section 2.7 considers the issues currently surrounding the measurement and conceptualisation. Within the social psychology research relating to sport fans, these measures have lacked adequate psychometric properties and are unidimensional in nature despite evidence in other psychological fields (consumer and organisational psychology)

23

suggesting that our commitment can and should be conceptualised as multidimensional (Allen & Meyer, 1990; Delobbe et al, 2000; Meyer et al, 2006, Pritchard, Iwaski & Havitz, 2004). Psychology literature about sports fans has received little research attention (Wann & Pierce, 2003), and seems to be true for over a decade. Wann & Hamlet (1995) found that only 4% of social psychology journals published articles about sport fans. Since then, few attempts have been made in social psychology (with the exception of Daniel Wann’s research, 1990 - present) to decrease the deficit of research examining sports fans within a social identity framework. Despite the lack of research attention, sport fans have been utilised as a prime example for describing psychological processes appearing in the real world, such as social identity (cf. Ellemers, Spears, & Doosje, 1997). Furthermore, these links often give examples of sport fans whose behaviour has become dysfunctional and aggressive through extreme social identity and other group related processes, such as depersonalisation and deindividuation (e.g. Stott & Reicher, 1988). In contrast, there is an increasing interest in sport fans from researchers in other social sciences (Giulianotti, 2005; James & Ridginer, 2002; Quick, 2002; Wann & Pierce, 2003). In sociology for example, the commodification of sport fans and its consequences (such as alienation) on fans has become a substantial topic for debate (see Giulianotti, 2005). In addition, the relationship between globalisation and sport spectating has also been given much attention in sociology (most notably Giulianotti & Robertson, 2007a; 2007b). The majority of research about sport fans in other social sciences such as marketing has examined factors influencing fan attendance. This research often concludes that success is an important factor predicting attendance but that there are also external factors that

24

influence fan attendance, such as player selection, social networks, and proximity of stadia to fans’ home (Pan & Baker, 2005; Sloan, 1979; Wann et al, 2001). In addition to external factors, personal factors have been found to influence fans’ attendance. These have included individuals’ levels of extraversion, a need for arousal, and overlapping with a social identity perspective, the need for selfesteem, as well as fans’ strength of identification with a team (Donovan, Carlson, & Zimmerman, 2000; Robinson & Trail, 2005; Wann & Branscombe; 1993; Wann, Tucker, & Schrader, 1996; Wann, Roberts, & Tindall, 1999; Wann, Schrader, & Wilson, 1999). Therefore, in addition explanations by other social sciences about how sport fandom fits into fans’ lives and identity, psychology can offer a unique perspective about sport fandom, such as why fans value their sport teams and why it is important to their identity? Addressing sport fans from a psychological perspective may provide balance and illuminate findings from other disciplines in a complementary way.

2.2 The significance of sport and sport fandom

Belonging to a supporter group as with many other social groups, can have a large impact on our lives (Bairner, 2001; Crisp, Heuston, Farr, & Turner, 2007; Wann et al, 2001). Sport fandom increases peoples’ community pride and collective identity (Sansone, Crabb & Goldstein, 1991), offers people the chance to associate with a larger group (Bairner, 2001), and unifies and divides communities and nations (Raney, 2006). For example, the sport of rugby has helped provide New Zealand and Wales with a national identity at the turn of the 20th century (Thomas, 2004). In South Africa, rugby was used firstly to express white cultural supremacy during Apartheid, and later (at the 1995 Rugby World Cup) helped 25

provide cultural assimilation after the end of the Apartheid era, competing under the slogan ‘one team, one nation’ (Naughright, 1997). In America, migrant Scottish football fans evaded assimilation into American sports by establishing local football teams (glocalisation) as a means to sustain cultural allegiances and meaning through a process of ‘relativization’ (Giulianotti & Robertson, 2007). The fascination that individuals can have with sport has led some researchers to describe the importance and impact of sport on our lives as pseudoreligious (Higgs, 1995 and Novak, 1976). Fans perform rituals and use symbolism such as wearing team apparel, and such behaviour has been described as akin to religion (Price, 1998). Religious terminology is sometimes used to describe fans’ attachments to their teams such as being a ‘devoted’ fan and having ‘faith’ in your team. The terminology is affectionately known in the media as ‘sportugese’ (Wann, Metcalf, Adcock, Choi, Dallas, & Slaton, 1997). Finally, investigating sport fans within social psychology may become more important in future because sport has a role in our lives that is on par with religion and politics (Higgs, 1995; Novak, 1976; Wann et al, 2001). This can be seen in terms of cultural expression (Giulianotti & Robertson, 2007; Naughright & Black, 1998) as already discussed, and strengthening societies through social affiliations (Bairner, 2001; Joiner, Hollar, & Van Orden, 2006; Melnick & Wann, 2004; Wolfson, 2001). One speculated reason for this might be the declining interest and affiliation with other social practices such as religion, politics, and work groups (Branscombe & Wann, 1991; Haut & Fischer, 2002; Putnam, 1995; Wann et al, 2001; Wann, 2006b). For example, Voas & Crockett (2005) found that religious belief was declining faster than attendance at religious venues in the UK. Therefore, through increased importance in society, sport fandom may become

26

another way to maintain overall psychological well being (Wann, Dimmock, & Grove, 2003; Wann & Polk, 2007). There are however, negative aspects of sport fandom as already mentioned in the previous paragraphs (i.e. aggression). One counter-intuitive finding in sociological research for example has found an increase in interpersonal violence (White, Katz, & Scarborough, 1992) after a win rather than after a loss when frustration might lead to an increase in violent behaviour (Wann et al, 2001). White et al attributed the increase in violence against women after a win due to an increase in power motivation that some males might feel. Disputes that might normally be resolved through negotiation are instead settled by force (Wann et al, 2001). Despite the significant amount of time and meaning that individuals give to sport fandom, it is surprising that social psychology in particular has given limited attention to this. This is despite the numerous psychological frameworks that are relevant to sport fandom (Donavon, Carlson, & Zimmerman, 2005). In particular, as Dimmock & Grove (2006) noted, a social identity framework could help provide explanations for sport fans’ attachments. Since this framework has been used at various levels of abstraction (e.g. role of self-esteem, self-enhancement and uncertainty reduction motivations, as well as social identity status threats) in previous research for explaining sport fandom, the social identity framework and how it relates to sports fans is therefore the focus of the next section.

2.3. Social identity and sport fandom

For sport fans, the team they support is an important part of their self-concept and for the most part fans are highly involved with a strong emotional attachment to 27

their team (Branscombe & Wann, 1990; James, 2001; Shank & Beasley, 1998; Wann & Branscombe, 1993; Wann, Melnick, Russell, & Pease, 2001). Consequently, sport fans like members of other social groups give emotional significance to their group membership, as well as gain positive well being and perceived social identity value as part of belonging to a group (Hirt, Zillman, Erickson, and Kennedy, 1992; Murrell & Dietz, 1992; Wann & Polk, 2007; Wann, Inman, Ensor, Gates, & Caldwell, 1999). Another important but often overlooked aspect of fandom, is the role that fellow fans play. Self-categorisation theory for example, explains how aligning one’s self to a group’s prototypical behaviour helps provide self-definition and reduce uncertainty about one’s self-concept in any given context (Hogg, 2000; Reid & Hogg, 2005; Turner, 1999), such that the fans are self-defining in terms of how they should behave. Therefore, part of being a fan involves acting ‘appropriately’ as a fan of a specific team. For example, Davies, Mayer, Shearer, Hall, Thomson, & Hall (2008) showed that Welsh regional rugby fans defined themselves in terms of their commitment to their team, but also described their in-group as being the most committed in comparison to other Welsh regional fans. Team performance however, has been the focus of psychological research even though the team provides very few clues about group prototypical behaviour (Branscombe & Wann, 1994; Hirt et al, 1992). Although the positive benefits of group affiliation and social support other group members provide has been highlighted (see Stangor, 2004 for a review; see also Wann, Dimmock, & Grove, 2003), it is surprising that social psychology literature about sport fans’ behaviour has remained focussed on the outcome of games. As such, one aim of the thesis is to examine how important the in-group is for fans’ social identity. For example, the

28

extent to which fans use their in-group to gain self-esteem is examined (see Chapter Four). Despite sparse findings, it has been found that the success of a team can have both a strong effect on the fans’ evaluations of their own attributes and influence feelings of personal success and failures (Wann et al, 2001). For example, when a team wins fans may experience and share a sense of achievement, and when they lose fans may experience a feeling of personal failure. Hirt et al (1992) showed that fans reported negative moods, low self-esteem, low self-efficacy, and negative evaluations about themselves at various cognitive tasks after a team loss. In contrast, after a win, fans reported increased social well being, collective self-esteem (selfesteem derived from a positive evaluation of one’s team), and a decrease in loneliness (Wann, Dimmock, and Grove, 2003). Accordingly, as Wann et al (2001, p4) pointed out, ‘When it happens to my team it happens to me’. It is clear that fans’ attachments are important for their social identity and overall self-concept.

2.3.1 A social identity perspective As discussed in Chapter One, one approach that has been useful for understanding fans’ attachments to a sport team and intergroup behaviour is the social identity perspective (see Dietz-Uhler & Murrell, 1999; Dimmock & Grove, 2006; Donovan et al, 2005; Ellemers et al, 1997; Murrell & Dietz, 1992; Tajfel & Turner, 1979, Tajfel, 1986; Turner et al, 1987; Wann, 2006a, 2006b; Wann & Polk, 2007). This is because the extent to which a fan identifies with a team, and with other fans as one social group, has been found to be similar to other social groups (e.g. ascribed groups, family groups; see Stangor, 2004 for a comprehensive review). For example, the relationship between identification, status, self-esteem,

29

and the degree of intergroup discrimination has been demonstrated in both sport fans and members of other groups, such as work groups and ethnicity (Bergami & Bagozzi, 2000; Branscombe & Wann, 1994; Forbes & Hallier, 2006, Gaunt, Sindic, Leyens, 2005; Hewstone, Rubin, and Willis, 2002; Wann, 2006a). In the social identity perspective’s original conception as social identity theory, group memberships were defined as social identities: ‘the part of an individual’s self-concept which derives from their knowledge of their membership of a social group together with the value and emotional significance attached to that membership’ (Tajfel, 1978, p63). When a social identity is made salient through categorisation, and is important to individuals in that group (in-group identification), there is a need for self-enhancement and positive distinctiveness (Hogg et al, 2004). In order to enhance their group’s image, members must evaluate one’s group positively by making favourable social comparisons of the ingroup (Abrams et al, 1990; Crocker & Luhtanen, 1990; Crocker & Major, 1989, Oakes, 2002, Turner & Reynolds, 2003). This is done by using whatever statusdefining dimension is contextually relevant and important and often through intergroup discrimination (Ellemers et al, 2002; Hewstone, Rubin, & Willis, 2002; Hogg et al, 2004; Oakes, Haslam, & Turner, 1994; Rubin & Hewstone, 1998, Tajfel, 1974; Tajfel & Turner, 1979). This allows the group to achieve, maintain, or enhance a higher or preferred status over relevant out-groups (Hogg & Terry, 2000; Hornsey et al, 2003; Reid & Hogg, 2005). Therefore, group members experience positive affect from group derived self-esteem (collective self-esteem, Brown, 2004, Luhtanen & Crocker, 1992; Rubin & Hewstone, 1998). Increased self-esteem from belonging to a high status group has also been found to contribute

30

to generic well being of group members (Joiner, Hollar, & Van Orden, 2006; Stangor, 2004; Wann et al, 2003; Wann & Pierce, 2005). Relating a social identity perspective back to sport fans, they should be motivated to maintain or enhance a positive self-image by increasing the positivity or decreasing the negativity of the group. In order to achieve superior status and positive distinctiveness over out-groups, the most obvious dimension of social comparison has traditionally been the team’s performance. Unsurprisingly, the majority of research examining social identity processes in sport fans focuses on the impact of competition between teams, such as game outcomes (e.g. Hirt et al, 1992; Wann & Grieve, 2005). Social identity theory then, would suggest that team performances will influence the valence of a fan’s social identity. When their team wins, fellow fans are evaluated positively and therefore more favourably (in-group bias 1) than fans of the team that lost. Therefore, a team win provides a favourable comparison and consequently results in a positive evaluation about fans’ image and elevated levels of self-esteem. In contrast, fans of a losing team are likely to have a less favourable comparison and therefore need to engage in image management strategies to decrease their group’s negative image (low status) and negative selfconcept (Tajfel & Turner, 1979; Wann, 2006a). Since teams appear to directly reflect their fans’ identity, according to social identity theory fans should therefore experience a team’s loss as a social identity ‘threat’, equating to a low status (cf. Murrell & Dietz, 1992). This is because game outcome appears to be the most obvious and most relevant comparative dimension that fans would use to evaluate their status, and thus positive distinctiveness from relevant out-groups. As such, the impact of team 1

It is important to note that intergroup discrimination is used for both in-group bias as well as outgroup derogation. In-group bias is not a necessary extension for out-group derogation, and they can both occur relatively independently (see Hewstone et al, 2002 for a review)

31

outcome as a form of comparative dimension, and thus social identity threat, has been demonstrated in early research examining reactions to team losses (e.g. Branscombe & Wann, 1990; Branscombe & Wann, 1994; Cialdini & Richardson, 1980; Hirt et al, 1992; Mann, 1974; Murrell & Dietz, 1992). Sport fans are constantly competing for a higher status through team performances and thus their positive social identity is perpetually threatened through possible and actual losses. A supporter of a constantly unsuccessful side cannot claim their group to be of high status, and is likely to experience a chronic threat to their social identity. As social identity theory proposes, we strive to achieve, maintain, and enhance a positive social identity motivated by selfenhancement, as this will have the most benefits to our self-esteem. In addition to engaging in intergroup discrimination to produce favourable evaluations and feelings about the in-group (whether from in-group favouritism or out-group derogation), social identity theory also posits a number of self-enhancement strategies available to group members to reduce the impact of a social identity threat. These strategies are said to protect self-esteem and positive intergroup distinctiveness (Tajfel & Turner, 1979; see also Hornsey & Hogg 2000), but are confined to a belief system that individuals have about the social reality of their intergroup relations (Haslam, 2004; Hogg et al 2004). Social beliefs include the perceptions that group members have in particular about the permeability and relative status of the group, the stability and legitimacy of their group’s status relations, as well as the possibility of achieving and sustaining an alternative status (Hogg et al, 2004; Hornsey, Leeuwen, & Van Santen, 2003). These beliefs as Hornsey & Hogg (2000) point out, are based on ideological contests between groups rather than reflecting a social reality. The

32

combination of these beliefs determines a variety of image management strategies used by group members to pursue self-enhancement and manage a positive social identity, and consequently enhance or protect group members’ self-esteem. Group members can use individual mobility when the group is of low status, but whether a group member ‘jumps ship’ is complicated by other factors (Doosje & Ellemers, 1997; Ellemers et al, 1999a). For example, members also pay attention to how the overall social mobility of their group is seen in terms of the permeability of the boundaries and/or the legitimacy of the status award (Jackson, Sullivan, Harnish, & Hodge, 1996). When individual mobility is not feasible, group members can use image management strategies such as social creativity. This can involve creating new comparative dimensions and externalising the threat (Crisp & Turner, 2007; Murrell & Dietz, 1992). Another strategy is social competition and this involves members directly competing with the out-group to bring about social change and can also include out-group derogation. (Branscombe & Wann, 1994; Brown, 2004; Cialdini & Richardson, 1980; Crisp et al, 2007; End et al, 2003; Prislin & Christensen, 2005; Tajfel & Turner, 1979). Since team performances appear to play an important role in enhancing their social identity and self-esteem (e.g. Hirt et al, 1992), in line with social identity theory, fans too have been found to possess several image management strategies For example, Murrell & Dietz (1992) proposed that because group members have a cognitive arsenal to protect their positive distinctiveness, poor team performance may not have as strong an effect on one’s self-esteem. Although these strategies have so far been used to demonstrate reactions to team losses, they can also be used to support the present thesis’ key arguments. This is because if fans use alternative means to protect their self-esteem it clearly indicates that there

33

may be other dimensions available for social comparisons with rival fans. In particular, rather than risk leaving the group, fans might be likely to change the comparative dimension to something that they can better control. This would allow fans to be responsible for achieving a favourable comparison over rivals rather than having to rely on their team’s performance. One comparative dimension that has not been previously examined is the in-group’s value and characteristics (where team performance has been used to determine the in-group’s value), and in particular fans’ commitment. Although attending matches has been found to make fans feel like integral participants in the game (Bairner, 2001; Neville, Balmer, & Williams, 1999; Novak, 1976; Wann et al, 1994, Wolfson, Wakelin, & Lewis, 2005), attendance may also serve other social identity benefits such as demonstrating one’s commitment and model behaviour as an in-group member. Research by Wann & Dolan (1994a) for example, showed that fans simply externalise their team’s defeat so that fans of a losing team do not feel responsibility or blame for the loss. Therefore, it is argued that, because fans cannot control the outcome of a game, more personally relevant comparative dimensions are ones that can be controlled by fans and contribute to their positive image. Since fans can control their image as supporters, it may be that this is the most realistic dimension (i.e. we are better fans because we stay with our team after defeat) that fans use to gauge their social identity’s status over rival fans. Branscombe & Wann (1995) found that fans manipulated the perceptions of fellow fans’ behaviour to present a more appropriate explanation that suited the ingroup’s ideological image of a model in-group member. Fans showed more perceptual alignment and in-group bias when rating a fellow fan’s behaviour at a

34

game (e.g. they were passionate not aggressive), than when rating the same behaviour by a rival fan (then they were aggressive). Despite game outcome being the most obvious comparative dimension, being seen to be a good fan by demonstrating commitment may be more important for enhancing one’s social identity and positive distinctiveness. This may especially be the case since fans have control over their image as committed fans. For example, Wolfson, Wakelin, & Lewis (2005) reported that fans tend to accentuate being superior fans to rival fans. Similarly, Wann (2006a) suggested that where the team may fail on one dimension (i.e. performance), fans can express their superiority creatively on another dimension (i.e. being better fans). In a recent study, Davies et al (2008) reported that fans tended to express superiority over rival fans in terms of commitment. As such, it appears common for fans to evaluate fellow and rival fans in terms of their commitment (e.g. die-hard and fair-weather fans, see Wann & Branscombe, 1990; and Wann & Branscombe, 1993). The use of in-group characteristics however, has not been demonstrated empirically as a comparative dimension and this is achieved in Chapter Four. Fans might actually distinguish themselves further within their in-group. Inline with the social identity perspective, fans will experience an acceptance threat if they acted atypically to the in-groups’ prescribed norms (Branscombe et al, 1999). As we have already discussed, CORFing is one behaviour that is frowned upon and as such, being seen to be a good supporter amongst fellow fans might also be as important as being sent o be a good supporter by rival fans. In addition, what is particularly interesting about examining fans commitment is that, to show their commitment to both in and out-group members, sport fans require their team to lose occasionally in order to demonstrate that they

35

and fan group are committed; a good example of when to be a good fan (cf. Ellemers & Spears, 1997). Consequently, fans changing the comparative dimension to accentuate their commitment after a team loss might therefore act to elevate their social identity image and therefore self-esteem. As such, team performance may only be an additional enhancement to fans’ social identity. As discussed earlier, self-esteem is clearly important in the social identity and sport fan literature, and is also an important theme in this thesis. Therefore, the next section will discuss the role of self-esteem further in sport fans, and in particular, with reference to how sport fans manage their social identity and thus self-esteem when their social identity has been threatened.

2.3.2 Social identity and self-esteem As previously discussed, according to social identity theory, one reason that we are motivated to identify with a group is for self-enhancement. This is because it is assumed a distinct positive self-image will have the greatest benefits to our self-esteem (Hogg & Williams, 2000; Tajfel, 1978). Furthermore, the impact of social identity threats on our self-esteem has been well documented (e.g. Abrams & Hogg, 1988; Abrams & Hogg, 1999; Brown, 2004; Crocker & Major, 1989; Ellemers, Spears & Doosje, 1997; Ellemers et al, 1999a; Ellemers et al, 1993; Foels, 2006; Long, Spears & Manstead, 1994; Mullin & Hogg, 1998; Oakes, Haslam & Turner; 1994). In particular, the relationship between intergroup discrimination and self-esteem has been elaborated through the self-esteem hypothesis (Abrams and Hogg, 1988; see also Hogg & Abrams, 1990). This hypothesis assumes that intergroup discrimination is motivated by a desire to see

36

one’s group in a positive light and therefore engaging in intergroup discrimination elevates self-esteem. The literature on the role of self-esteem and the self-esteem hypothesis in social identity theory to predict intergroup discrimination has been described as confusing and has yielded mixed results (Brown, 2004; Hunter, Kypri, Stokell, Boyes, O’Brien, & McMenamin, 2004; Rubin & Hewstone, 1998). According to the two corollaries described in the self-esteem hypothesis, on the one hand a threatened social identity and thus lowered self-esteem motivates members to engage in intergroup discrimination because of a need to enhance self-esteem. Selfesteem in this corollary is an independent variable. On the other hand, engaging in successful intergroup discrimination when a threat is not present maintains or elevates self-esteem, and thus in this corollary, self-esteem is a dependent variable based on specific intergroup discrimination (Hogg & Abrams; 1990; Rubin & Hewstone, 1998). Despite self-esteem’s role in intergroup discrimination, the impact of amount of identification with a group on our self-esteem is generally agreed to be a positive one; the more identified one is with a group, the more self-esteem an individual will gain from their membership in a specific group (Hogg et al, 2004; Stangor, 2004; Wann & Pierce, 2005). Therefore, it is unsurprising to find that sport fans generally gain high amounts of self-esteem from supporting a sport team (e.g. End et al, 2002; Davies et al, 2008; Wann, 1995; Wann, 1996; Wann et al, 2003; Wann & Pierce, 2005; Wann, Tucker, & Schrader, 1996; Wann & Polk, 2007). Additionally, the role of self-esteem in sport fans has been important in other disciplines (such as consumer behaviour) and has been reported as an

37

antecedent of fan behaviour and identification with a team (Funk & James, 2004; Gibson, Willming, & Holdnak, 2002; Kwon & Trail, 2001; Mahony, Nakazawa, Funk, James, & Gladden, 2002, Spinard, 1981). Marketing literature for example, has also established that self-esteem and strength of team identification are highly correlated (Kwon & Trail, 2003; Trail & James, 2001, but see also Wann, 1995; Wann & Pierce, 2005). These studies provide additional evidence for the link between social identity and self-esteem in other contexts than have previously been studied in social psychology literature. In terms of the effects of game outcome on fans’ self-esteem, with the exception of a few studies (e.g. Bizman & Yinon, 2002; Hirt et al, 1992; Murrell & Dietz, 1992), the majority of the research is speculative about the link between game outcome and fans’ self-esteem. Specifically, the link between self-esteem and game outcome explains fans’ reactions in terms of image management strategies. Most studies (e.g. Snyder et al, 1986) assume that these strategies help fans protect or maintain high self-esteem without directly establishing a need for protection or maintenance of their self-esteem. Researchers (e.g. BernacheAssollant et al, 2007; Crisp et al, 2007; End, 2001; Wann & Branscombe, 1990) frequently return to image management strategies suggested by Cialdini et al (1976) and Cialdini & Richardson (1980). They found fans strongly associated with their team by Basking in Reflected Glory (BIRGing). Later, Richardson and Cialdini (1981) speculated the opposite may also be true, namely that fans dissociate with unsuccessful others. However, they did not actually employ a measure to observe the effects of these strategies on fans’ self-esteem following a game, but rather the authors speculated that bolstering one’s public image by BIRGing, may serve to bolster one’s self-esteem.

38

Following this, the opposite effect proposed by Cialdini et al was discussed by Snyder, Stucky, and Higgins (1983), and then demonstrated by Snyder, Lassegard, and Ford (1986). Participants dissociated with an unsuccessful group by Cutting Off Reflected Failures (CORFing). This was presented as an image protection strategy where individuals severed negative associations with others who have failed in order to protect their self-esteem. Again, no measure was taken of self-esteem but rather a more ambiguous measure of negative emotions was used that has been taken to mean self-esteem (cf. Hirt et al, 1992). In social identity theory however, this would be a natural conclusion to make. Self-enhancement and protection of one’s public image would reflect an individual’s evaluation of their group’s image and therefore their levels of selfesteem. Two later studies by Hirt et al (1992) and Murrell & Dietz (1992) used social identity theory to provide some evidence for the effect of game outcome on fans’ self-esteem. Both studies observed a positive effect on self-esteem after a team had won and a negative effect on self-esteem after a team had lost. This research, however did not explicitly address BIRGing and CORFing and so it is difficult to draw conclusions about the use of these strategies to directly enhance or protect self-esteem. Returning to the conceptual framework of social identity, there is good reason for researchers to make the link between self-esteem and game outcome. Therefore, considering the effect of game outcome on fans’ social identity, image management strategies such as individual mobility, social creativity, and social competition, should logically also apply to sport fans in a similar manner as other group memberships (cf. End et al, 2002; Murrell & Dietz, 1992). Furthermore, these image management strategies may support the argument made in this thesis

39

that team performance may not be that important or beneficial for a fan’s social identity given that fans are well equipped to deal with this threat by using image management strategies. Rather, fans may use other more personally relevant and controllable comparative dimensions that are non-performance related such as their in-group’s characteristics (e.g. commitment) and other self-defining features. Therefore, the the next section focuses on the image management strategies that are employed by sport fans to protect their social identity and self-esteem after a sport team loses. By doing so, it is intended to show that having a variety of strategies makes team performance less important than previously thought.

2.4 Image management in sport fans

As discussed in previous sections, social identity theory posits that group members have various strategies to cope with social identity threats to their group status. Their choice of strategy is also largely dependent on their social belief system about their groups. The majority of research into group strategies has focused on how choice of strategy is moderated by the level of group identification (see Ellemers et al, 2002 for a review). In addition to identification, other factors determining responses include the social context of the threat, such as source and content, the variability of the threat, and the characteristics of the group (Doosje & Ellemers, 1997; Doosje, Spears, & Koomen, 1995; Forbes & Hallier, 2006; Hutchinson, Jetten, Christian, and Haycraft, 2006). Furthermore, some researchers have looked at in-group members’ accountability and anonymity to in-group and out-group members in reacting to threats (e.g. Barreto & Ellemers, 1998; Barreto & Ellemers, 2000; Ellemers, Van Dyck, Jacobs, & Hinkle, 2000). In addition, high identifiers have been found to emphasize group strategies (e.g. social change) 40

whereas a low identifier’s emphasis is to ‘bail’ on the group by using individual tactics such as dissociation or upward mobility into a group superior in status (Ellemers, Spears, & Doosje, 1997; Gibson, Willming, & Holdnak, 2002; Prislin & Christensen, 2005; Snyder et al, 1986; Wann, Hamlet, & Wilson, 1995). However, the literature about strategy choice is complex, and it is still unclear which parameters predict the strategy types members will use, and which strategy low status group members favour under different social identity threats (Brown, 2004; Ouwerkerk et al, 1999; Van Knippenberg & Ellemers, 1990; 1993). Similarly, several complex image management strategies have been found in sports fan research, but as with the social identity perspective, it is also unclear which strategies will be used and when. What is clear is that amount of identification has been found to influence individual or group level strategies to bring about social change (Ellemers et al, 2002; Ouwerkerk et al, 1999), and suggests some similarities between these frameworks. Hence, the following strategies will integrate the strategies found in sport fan research and embed them within a social identity perspective.

2.4.1 Individual mobility According to social identity theory, when the status relations are stable, the disadvantage is perceived as legitimate (such as when a team loses), and group boundaries are permeable, group members are most likely to use distancing tactics as this provides a line of ‘least resistance’ psychologically (Brown, 2004; Van Knippenberg & Ellemers, 1993; Tajfel, 1978) for protecting the self-image. Although the most individualistic of image management strategies, social identity theory also suggests that individuals that are lowly identified with their group are

41

likely to employ this strategy. This is because, for low identifiers, the group’s identity is unimportant for self-referencing and their self-concept, and therefore not important for providing a positive identity or for gaining high self-esteem (Ellemers et al, 2002; Tajfel, 1981; Turner & Reynolds, 2003). When an individual is highly identified, responses to identity threats should motivate members to enhance and maintain their self-esteem and group image with group level responses such as social creativity and social competition (see Dietz-Uhler & Murrell, 1998, 1999; Ellemers et al, 1993; Hogg et al, 2004; Tajfel & Turner, 1979; Tajfel & Turner, 1986). Sport fans however, are interesting because, although they are largely unknown to each other and there are no formal rules for joining or leaving the group, leaving particularly after a defeat would seem unthinkable to the die-hard committed fan (Branscombe & Wann, 1990; Ellemers, Spears, & Doosje, 1997). Therefore, many fans are reluctant to leave or dissociate completely with their team when they lose. As such, the group is perceived as being impermeable, at least psychologically, and it is proposed in this thesis that fans may in fact prefer to be seen as ‘committed’ rather than being mislabelled as ‘fair-weather’ glory supporting fans. The salient social identity threat is therefore to acceptance from other members as well as to the group’s value (see Ellemers et al, 1999) in terms of team performance. One reason may be that fans have numerous strategies they can use to cope with a team loss, which might be more difficult to employ if the ingroup is implicated since the fans are responsible for their image. Although dissociation is the traditionally observed tactic in social identity literature (e.g. Ellemers, Wilke, & van Knippenberg, 1993, Jackson, Sullivan, Harnish, & Hodge, 1996, Trafimow, Armendariz, & Madson, 2004), distancing

42

tactics are bidirectional. Just as members dissociate to protect their image and selfesteem, members associate to enhance their image and self-esteem. Therefore, in line with social identity theory’s self-enhancement motive, individuals should also show in-group favouritism. Fans enhance their social identity by choosing to associate with successful groups and therefore elevate their self-esteem, even if that group is not an important social identity to an individual (see Abrams & Hogg, 1988; Cialdini et al, 1976; Hogg & Abrams, 1990; Murrell & Dietz-Uhler, 1999; Tajfel & Turner, 1986; Tajfel, 1984). Like other group members, sport fans strategically manipulate their associations with their team based on the positivity or negativity of their team (Bernache-Assollant et al, 2007; Crisp et al, 2007; Richardson & Cialdini, 1981; Wann, 2006a). Two distancing tactics have dominated the sport fan literature that describes individual mobility in both directions. These are the BIRGing and CORFing tactics described in the previous section, the former being an image enhancement strategy whilst the latter an image protection strategy, and these tactics are the focus of the following subsections.

2.4.1.1 Basking in reflected glory Underpinning the BIRG phenomenon according to Cialdini and Richardson (1980) was the notion that people publicly showed their associations with a successful other, even if they had little to do with the success. In Cialdini et al’s (1976) research, they found that university sport fans stressed their association with their team and fellow fans the day after a win by wearing university apparel. Furthermore, in a follow up study fans used more collective pronouns when describing their team’s victories. In contrast, they also observed that fewer students wore apparel after a defeat, suggesting fans dissociated from an unsuccessful team.

43

By associating with a successful team, fans may believe that they will also be perceived as successful and have a positive identity. Cialdini and colleagues speculated that associating was one way individuals enhanced their self-esteem. This assumption is in line with the assumptions of social identity theory where group members are motivated to achieve positive identity through favourable comparisons, and to show in-group favouritism when a group is successful at achieving high status (see also Murrell and Dietz, 1992; Spears, Doosje, & Ellemers, 1997). This will then have the greatest benefits for elevated self-esteem (Abrams & Hogg, 1999; Crisp et al, 2007; Foels, 2006; Reid & Hogg, 2005; Rubin & Hewstone, 1998). In a later study, Wann, Royalty, and Roberts (2000) showed that highly identified fans were more likely to mention their team associations to out-group members. This was particularly the case for highly identified fans with high selfesteem, suggesting self-referencing acted as confirmation for a positive social identity (cf. Scheepers, Spears, Doosje, & Manstead, 2003, Turner & Reynolds, 2003). However, BIRGing by both highly and lowly identified provide fans with an incentive to maximise group differentiation. In other words, fans are able to distinguish their successful team from less successful teams, and therefore distinguish their in-group from rival out-groups, a process of positive differentiation (see Hogg et al, 2004). Other research has found that fans can BIRG using several other methods. These included painting their faces, BIRGing privately using websites and chat forums, displaying apparel in their cars and on their desks, and travelling great distances to get to a game (Boen, Vanbeselaere, & Feys, 2002; Cialdini & Richardson, 1980; End; 2001; End; et al, 2002; Pan & Baker, 2005). In addition,

44

BIRGing has also been found to be more apparent for unexpected wins, but can also be found in situations where a positive outcome is likely (Boen, Vanbeselaer, & Feys, 2002; End et al, 2002). BIRGing then, gives fans the opportunity to use their team as a strategy for positive identity confirmation by announcing their group’s superiority and reinforcing their membership within a superior group. Team performances however, as we will see in the following sections, may be only an additional method for further ‘positive’ reconfirmation of their social identity when the team wins.

2.4.1.2 Cutting off reflected failures Since fans of a losing team cannot declare superiority over fans of the successful team, losing fans have been found to dissociate with their team (Bizman & Yinon, 2002; Crisp et al, 2007). Fans could sever ties physically by BIRGing either with another high status team (out-group favouritism) or psychologically by cutting off reflected failures. Snyder, Lassegard, & Ford (1986) found that group members protected their positive image and self-esteem by dissociating with an unsuccessful group. Students who received negative feedback for a group task showed less association with the group by not taking or wearing their work team’s badge in comparison to group members who received positive or no feedback. Although disloyal fans are frowned upon by more committed supporters (Berache-Assollant, Lacassagne, & Braddock, 2007; Bizmon & Yinon, 2002; Ellemers et al, 1997; Wann & Branscombe, 1991; Wolfson, Wakelin, & Lewis, 2005), some research has found that disloyal members in assigned groups (those that CORF) who successfully migrated into a higher status group, were rated more positively in comparison to unsuccessful members (Charlton & Bettencourt, 2001).

45

Furthermore, the more impermeable the boundaries in both groups were believed to be, the more positively group members were rated. However, since fans emphasise being loyal as an important dimension of fan behaviour (Wakelin et al, 2005), as well as seeing their fan group as impermeable, it may be unlikely that migrating fans will be seen more favourably or indeed accepted by the opposition’s fans (i.e. having a disloyal fan may negatively affect the group’s image). Both social identity and sport fan research have proposed that CORFing is less likely to occur for highly identified members, even those of low status groups (Berache-Assollant, Lacassagne, & Braddock, 2007; Doosje & Ellemers, 1997; Gibson et al, 2002; Wann et al, 2001; Wann & Branscombe, 1990; Wann & Branscombe, 1993). Research in social identity for example, has found that lowly identified group members of a low status group decreased their self-stereotype of a typical in-group member in an attempt to dissociate from the group. In contrast, highly identified members increased self-stereotyping when their social identity was threatened (Doosje, & Ellemers, 1997; Doosje, Spears, & Koomen, 1995; Spears, Doosje, Ellemers, 1997). In addition, Branscombe, Wann, Noel, & Coleman (1993) found for example that fans who were seen to CORF by highly identified in-group members were seen as atypical of the in-group, and were described as disloyal and fair-weathered. In other words, fans dissociated themselves from the atypical in-group member, and similar behaviour has been reported in other social psychological research about deviant in-group members and research about the ‘black sheep effect’ (see Castano, Paladino, Coull, & Yzerbyt, 2002; Marques, Yzerbyt, & Leyens, 1988; Matthews & Dietz-Uhler, 1998).

46

These findings help support the key proposal of the thesis where part of being a good fan is showing strong commitment, particularly after a loss. However, in contrast to the above research, Bizman and Yinon (2002) investigated fans before they had left the stadium, and found that after a loss highly identified fans did CORF, but to a far lesser extent than lowly identified fans. They also suggested that highly identified fans returned to previous levels of allegiance very quickly after the threat has dispersed. This finding has also been found in more general group memberships, particularly when a low status group is considered to have relatively sharp boundaries (Ellemers et al, 1997; Ellemers, Wilke, & Van Knippenberg, 1993; Ellemers & Van Rijswijk, 1995; Van Knippenberg & Ellemers et al, 1993; Jackson et al, 1996). Furthermore, even though there are no formal rules for joining or leaving the group for sport fans, and detection of deviant fans is likely to be very difficult, being disloyal would nonetheless be an unthinkable act. Therefore, psychologically at least, the boundaries may appear impermeable to fans. Another way fans may CORF is presented in the social identity literature where members re-categorize their membership to a more inclusive superordinate group. Re-categorisation can reduce identity threats and can attribute potentially higher status out-group members to the current in-group (Crisp, Stone & Hall, 2006, Gaertner et al, 1993; Gaertner, Mann, Murrell, and Dovidio, 1989; Seta, Seta, & Culver, 2000; Van Leeuwen, Van Knippenberg, & Ellemers, 2003). This then would allow movement within a larger relevant social identity and provide additional benefits to one’s self image by offering a more stable and legitimate way of associating with their previous higher status groups (Roccas & Brewer, 2002; Seta & Seta, & Culver, 2000; Van Leeuwen et al, 2003). For example, Davies et al,

47

(2008) found that after several Welsh rugby teams had merged with traditional rivals, fans dissociated with their new regional social identities. Rather than build new rivalries, fans re-categorised their social identity to an already existing and legitimate superordinate social group by using the national team. Furthermore, since the merger had extinguished over 100 years of traditional intergroup histories and rivalries, fans’ commitment to the Welsh national team had increased to a higher level than before the mergers where commitment had been far higher to their traditional teams. The authors speculated that although national rugby was important to fans, Welsh rugby fans now had a shared common interest in terms of losing their beloved traditional team. Banding together as ‘Welsh rugby fans’ was likely to be in ‘protest’ to those involved in illegitimately (unfair and unjustly disadvantaged) taking away their traditional, stable, and legitimate group identities. In summary, the majority of the evidence suggests lowly identified fans show distancing tactics, but there is also some evidence that highly identified fans show a similar pattern. In particular, these strategies help distinguish lowly identified fans from highly identified fans after a loss. For example, in order to protect their image and self-esteem, lowly identified fans are more likely to be disloyal and jump ship after a loss when they believe their group boundaries are perceived to be permeable. In contrast, highly identified fans perceive the boundaries to be impermeable and leaving for these group members would be unthinkable, as well as risk being labelled a fair-weather fan. Highly identified fans are then left with still trying to resolve a social identity threat caused by their team’s loss. Social identity theory suggests that highly identified fans of a low status group are likely to engage in group serving tactics and it these strategies that are discussed next.

48

2.4.2 Social creativity strategies Since most fans are highly identified and have strong emotional attachments to their team (Shank & Beasley, 1998; Wann et al, 2001), being a prototypically loyal and committed in-group member is a natural extension of an individual’s attachment to their social identity (Ellemers et al, 1999c; Reid & Hogg, 2005; Tuner, 1999; Turner et al, 1987). Therefore, switching teams when they lose is not always a psychologically or physically possible option. This is particularly difficult when that team is an important extension of a fan’s selfconcept, especially since highly identified fans perceive a high risk of detection as a fair-weather fan (Wann & Branscombe, 1990). In this way, the thought of leaving may create an additional threat at least for highly identified fans. As such, highly identified members in a low status group are the least likely to opt for individual mobility when their social identity is unfavourable (cf. Barreto & Ellemers, 2000; Cicero et al, 2007; Ellemers et al, 1997, Ellemers et al, 1999a; Ellemers & Van Rijswijk, 1995; Hutchison et al, 2006; Jackson et al, 1996; Prislin & Christensen, 2005). Instead, when an individual is highly identified, responses to social identity threats should motivate members to enhance and maintain their self-esteem and group image with group level responses. These strategies include social creativity and social competition (Dietz-Uhler, 1999; Ellemers et al, 1993; Hogg et al, 2004; Murrell & Dietz, 1992; Tajfel & Turner, 1986) that are intended to bring about social change, and therefore enhance their group’s status. Social identity research has shown that highly identified members of low status groups tend to change the out-group comparator or change the status-defining dimension so that a more favourable social comparison could be made towards the in-group (see Brown, 2004; Hogg & Terry, 2000; Hornsey et al, 2003; Jackson et al, 1996; Wann,

49

2006a). This has also been described as ‘compensatory’ bias (Tajfel, 1974). In other words, fans may use downward social comparison with a lower status team (we did not beat team X but at least we beat team Y). Another method, and one that this thesis focuses on, is changing the primary comparative dimension to something that will allow a more favourable comparison. This can be done by changing and manipulating the values assigned to the group so that previous negative qualities are perceived as positive (e.g. “they don’t play dirty, they’re just playing aggressively”). Alternatively, fans could manipulate the importance of the comparative dimension in their favour (e.g. intergroup history, “we may have lost today, but we are more successful historically”). In this thesis, since commitment is an important part of being seen to be a good fan, it is hypothesised that fans would change the comparative dimension to the in-group’s characteristics. This is because some research has highlighted that fans accentuate their commitment (Wakelin et al, 2005). Therefore, since they have control over their image as committed fans, it might be that this dimension will be valued more than their team’s performances (e.g. “we may have lost, but at least we stick with our team”). Finally, rather than change the value, fans are able to engage in attributional biases to explain away unfavourable comparison such as a team’s loss (e.g. poor refereeing, weather). Of the creative strategies, attributional biases have been the most extensively covered in sport fan literature and these are discussed first.

2.4.2.1 Attributional Bias Apart from BIRG and CORF strategies, attributional bias is the second most developed area in sport fan literature. Since highly identified fans choose to stick with their team due to their strong attachments and the perceived

50

impermeability of the group boundaries, they are unable to engage in distancing strategies. These fans may protect their self-esteem by engaging in self-serving attributional biases to account for performances. These self-serving biases have been demonstrated with sport fans in several experiments and can be observed with several identity relevant dimensions such as game outcome, biased recall, and perceptions of fans (e.g. Wann & Dolan, 1994a; Wann & Schrader, 2000). In one of the first studies to investigate attributional bias in sport fans, Mann (1974) interviewed Australian football fans from both winning and losing teams and found that biases were observed in fans of a losing side. These fans tended to exaggerate the amount of free kicks the winning team had, reported seeing the opposition play a ‘dirtier’ game, and externalised their team’s failure by implicating uncontrollable factors such as luck or the refereeing in an attempt to explain away their team’s defeat. Similarly, in a series of studies examining team success/failure biases, Wann & Dolan (1994a, 1994b), Wann, McGeorge, Dolan, & Allison (1994), and Wann & Schrader (2000) demonstrated how fans’ perceptions altered after a loss to protect their self-esteem. For example, Wann & Dolan (1994a) showed that highly identified fans of a college basketball team internalised causes for their team’s success (e.g. the fans influenced the game, the team’s players) and appeared to take partial credit for the team’s win to enhance their social identity. In contrast, fans externalised causes for their team’s failures (e.g. blaming opponents’ playing style and the refereeing) to protect their self-esteem, but these biases decreased as levels of fan identification decreased. In addition, Wann et al (1994, study 1) showed that highly identified fans believed that they played a key role in controlling the outcome of home games, and their perceived ability to influence the game increased as their identification with the team

51

increased. Similarly, Wolfson, Wakelin, & Lewis (2005) showed that English football fans internalised successes because they believed they were responsible for manipulating the referee’s decisions and intimidating the rival team’s players, but after a loss fans blamed their team and showed that they had very little to do with the performance. These studies highlight the importance of the amount of perceived control that fans have. Since they do not really have control over the team it is easier to externalise the blame. However, since they have complete control over their image as fans, it may be much more difficult to externalise failures if the comparative dimension is based on the in-group’s characteristics. Moreover, in a later study, identification has been found to moderate fans’ perceptions of their teams’ success/failure in terms of whether the success/failures are controllable and stable (Wann & Schrader, 2000). They found that after a win, highly identified fans, in addition to using internal biases, also included stable and controllable dimensions (e.g. team ability and effort level). After a loss, fans were likely to externalise failures through unstable and uncontrollable dimensions (players were tired from a road trip or ill). Other research on more generic group memberships by Dietz-Uhler & Murrell (1998) found that when highly identified members were given internal attributions for their group’s success and then external attributions for their group’s failure, they had higher self-esteem in contrast to lowly identified members. If the member cannot control the evaluation of the group or blame the negative evaluation on an external source then it affects members’ feelings of self-esteem. This is an underlying assumption in this thesis in terms of the image of the in-group. Therefore, if fans cannot externalise blame for a low in-group image (i.e. a negative social identity), it might be that fans’ selfesteem would be more affected than when their team loses.

52

Some research however, has found that when fans are not with each other, or their behaviour was not being publically evaluated, they showed decreased levels of attributional bias. For example, End, Eaton, Campbell, Kretschmar, Mueller, & Dietz-Uhler (2003) examined the generation of success/failure biases on an Internet forum after a win or loss. They found that biases disappeared, presumably as there was no need to enhance or protect one’s social image in public. In addition, the authors did find that fans generated internal and controllable attributions for rival team’s victories because they may be cutting off their own team’s future failures (COFFing, Aitken, Campbell, & Kent, 2004; Wann, Hamlet, & Wilson, & Hodges 1995). More in line with the main proposal of this thesis, some research has examined attributional biases about in-group fans’ behaviours. Wann & Dolan (1994b) for example, gave students that were fans of their university basketball team a fictitious account of either a highly identified fellow or rival fan at an important match with their biggest rivals. The hypothetical match was manipulated to be a crucial situation with the match tied in points and only a few minutes left for either team to win. In both accounts, the fan’s behaviour was the same. After reading the scenario, fans were asked to give their impression about the hypothetical fan’s behaviour. Fans that had received the account of a fellow fan showed more in-group favouritism and positive self-stereotyping describing the supporter as committed and prototypical of their fan group. Fans who were told the fan was a rival did not show out-group derogation, instead explaining the rival’s behaviour as an example of the out-group membership. In-group members were motivated, according to Wann & Dolan, to view other in-group members positively in order to claim superiority over out-group fans. Although they also recognised

53

the rival fan’s level of commitment, fellow fans were still seen as superior. This study is one of the first to show that fans may value other dimensions of their social identity other than team performances. Similarly, in a later study, Wolfson et al (2005) found that in-group members reported more positive self-stereotypes, such as fellow fans being more committed, loyal, and passionate about their team than rival fans, regardless of their team’s performance. More recently, Wann & Polk (2007) found that fans high in identification, in addition to showing increased in-group favouritism, reported trusting fellow fans and having a general belief in the trustworthiness of other individuals not related to their team. However, the authors speculated that it would be unlikely this trust would transfer to specific rival fans. These studies however, did not present any status manipulations of identification and the in-groups’ level of trust, instead the study was correlational in nature. Therefore, by using fans’ attributions such as trust, it is not known whether trust as an in-group attribute (and therefore, a valued comparative dimension) is any more valued than team performance. In this thesis, fans’ commitment is manipulated experimentally to investigate whether commitment as a comparative dimension is valued more so than team performance.

2.4.2.2 Changing the comparative dimensions In another attempt to devalue the inferiority caused by a team loss, fans can affirm other positive aspects of their social identity to compare themselves with the out-group in order to enhance and protect their self-esteem. One way would be to change the primary status-defining dimension to another comparative dimension such as the in-group’s worth (e.g. we have lost but we are still superior fans). This is a central argument of the current thesis. This is because the team is not actually

54

the in-group, and therefore may not provide the most important comparative dimension to the in-group members. In line with social identity theory, in this thesis, it is proposed that following a loss, fans will try to distinguish their in-group from other out-groups on more group relevant dimensions. By demonstrating their superiority in another way, such as showing more cohesion and loyal patterns of support (also potentially to avoid acceptance threats e.g. Ellemers et al, 2002), fans will therefore portray a more accurate account of their social identity rather than one based on their team’s performances. In addition, part of this thesis intends to investigate other comparative dimensions that fans may use since social identity research has recently pointed to several other types of threats (Branscombe et al, 1999). Therefore, fans may also be socially creative with the comparative dimensions after their team’s defeat. Fans may devalue the importance of their team’s loss on their image by reassigning a more favourable status-defining dimension for comparison (Doosje, Ellemers, & Spears, 1995; Ellemers & Van Rijswijk, 1997; Hornsey et al, 2003; Lalonde, 2002; Oakes, 2002; Tajfel & Turner, 1979; Wann &Grieve, 2005; Wann, 2006a). For example, another way fans achieve a positive social identity is via in-group favouritism. In this thesis, it is suggested that fans, in line with other social identity research (e.g. Ellemers et al, 2002; Jetten, Schmitt, Branscombe, & McKimmie, 2005; Spears et al, 1997; Wann & Branscombe, 1993; Wann & Branscombe, 1995), may assign a higher value to their in-group’s characteristics (e.g. commitment) to create a more favourable comparison after their team loses. This is especially the case since fans have been shown to produce in-group favouritism towards fellow fans rather than towards their team (Wann & Branscombe, 1995; Wakelin et al, 2005). As such, fans align

55

their perceptions with the group’s norms that provide the in-group’s self-defining characteristics. Alternatively fans may rely on other character values assigned to the team or fans such as playing style (Murrell & Dietz, 1992), or may over exaggerate their opponent’s abilities to give them a new category in which relevant and contextual comparisons do not hold much value (e.g. they play at another level to other teams) (Crisp & Turner, 2007). Another tactic mentioned earlier would be to show downward comparisons by comparing the in-group’s team with less successful teams for a more favourable comparison, (Ellemers, Wilke, & van Knippenberg, 1993; Murrell & Dietz, 1992; Prislin & Christensen, 2005). Despite the awareness by researchers of multiple comparative dimensions that could act as status threats (Branscombe et al, 1999; Ellemers & Van Rijswijk, 1997; Ellemers et al, 1999b; Ellemers et al, 2004; Hogg et al, 2004; Murrell & Dietz, 1992; Oakes, 2002), and the obvious value in exploring them, progression in examining additional comparative dimensions has been slow. This is particularly the case in sport fan literature. One of this thesis’ aims therefore, is to address this issue by examining a other comparative dimensions that fans may use to evaluate their social identity and self-concept; this is specifically achieved in Chapter Four.

2.4.3 Social competition As we saw in earlier sections, according to social identity theory, in-group favouritism was one way group members achieve a positive social identity. Since fans cannot truly influence their team’s performances, when a team loses, the function of the team for providing a positive social identity ends. The outcome of a losing performance however, may remain as a strong reflection of the fans’ status

56

in their minds (Bizman & Yinon, 2002). When no other cognitive alternatives are available and treatment of the in-group is considered to be unstable and illegitimate, highly identified fans are likely to compete with the high status outgroup by displaying increased in-group cohesion and self-stereotyping (DietzUhler, 1996; Ellemers, Barreto, & Spears, 1999). For these fans, managing their image and self-esteem is achieved by accepting their lower status following defeat and competing anyway. This reaffirms their positive image as a good fan and positive self-esteem in the face of defeat in attempt to bring about social change and thus a superior identity (Hogg and Abrams, 1990; Hewstone et al, 2002; Rubin & Hewstone, 1998). In marketing literature for example, this is termed basking in spite of reflected failure (Campbell et al, 2004). In this thesis, it is expected that if fans cope with their team’s defeat by changing the comparative dimension to their in-group, then manipulating their in-group may limit the image management strategies available to them. As such, it is expected that fans would increase their in-group cohesion and loyalty regardless of their lower image by competing with their higher status rivals anyway in an attempt to reaffirm their image as a positive one. For fans, part of the social competition also includes derogating rival fans (on any comparable dimensions) as an attempt to gain perceived high status by the in-group. Out-group derogation in sport fan literature is also known as ‘blasting’ (Cialdini & Richardson, 1980; Richardson & Cialdini, 1981) and is particularly prevalent among highly identified fans after a team loss as an attempt to convince onlookers and rival fans that they are still superior (End, 2001; Wann, 2006a). Blasting then, is an attempt to regain superior status and protect self-esteem, and can be seen in a range of post match contexts. These have included amongst the in-

57

group by derogating the source of threat (i.e. the rival’s team, or the refereeing), face to face with rival fans, as well as on the Internet, although fans on the Internet also displayed derogation of lower status groups (see Branscombe & Wann, 1992; 1994; End, 2001; Wann, 2006a). According to Cialdini & Richardson (1980), by blasting opponents of a winning team, losing fans are tying to convince others that although their team lost, they should still be seen in a more positive light unlike their opponents. According to the Self-Esteem Hypothesis, these types of behaviour are an attempt to bring about social change and compete with the outgroup (Hewstone et al, 2002), despite the status differentiation. For example, in one study investigating intergroup discrimination in sport fans, Branscombe and Wann (1994) showed members who were highly or lowly identified with their country (USA) a video of a boxing match between an American fighter and a Russian fighter. Highly identified members who saw the Russian win showed greater out-group derogation (as well as lowered self-esteem) compared to those with low identification or those who saw the Russian lose. For highly identified fans, engaging in out-group derogation was linked to elevated self-esteem and positive evaluations about their social identity. Another study by Wann & Branscombe (1995) also highlighted the role of in-group bias and derogation for social competition. Participants were separated into highly or lowly identified fans of a university basketball side. Fans were then asked to describe their fellow fans and then to describe their rival team’s fans. The most favourable descriptions were given by highly identified fans describing fellow fans, who also gave the most negative descriptions of rival fans. The amount of bias reduced as identification decreased. Another recent study by Davies et al (2008) found similar trends in Welsh rugby fans. Fans from across several regions

58

in Wales described fans from other teams in a similarly negative way (e.g. inferior, uncommitted), but all described their own group in a similarly positive manner (superior, committed). This suggests that fan differences may be more perceptual, rather than actual, in-group/out-group differences. However, the reliance on commitment in this study does highlight the importance and value of being seen to belong to a superior in-group for fans’ status. Moreover, in terms of derogation turning to violence, highly identified fans are likely to further their out-group derogation and behave aggressively to fans of other teams (Branscombe & Wann, 1992; 1994). In two studies for example, fans reported being willing to anonymously injure a player of another team, as well as engage in aggressive behaviours to help their team (Wann, Peterson, Cothran, & Dykes, 1999; Wann, Culver, Akanda, Daglar, DeDeivitis, & Smith, 2005). There is also evidence that fans will derogate fellow in-group members who bring the group’s status into question, and that lowly identified members are more likely to socially compete and engage in out-group derogation than highly identified fans when given the opportunity. For example, Branscombe, Wann, Noel & Coleman (1993) asked students who were fans of their university basketball team to evaluate a false newspaper article written by a fellow or rival fan and to assess their level of loyalty. In the loyal condition, the commentator expressed devotion to the team, and in the disloyal condition, their expressed loyalty was dependent on recent form of the team. Home team fans rated a disloyal in-group member more negatively than a disloyal out-group member regardless of whether the team won or lost. According to the authors, in-group fans did this in order to protect their group’s positive image and therefore their self-esteem. Unexpectedly, lowly identified fans gave the most negative evaluations of disloyal members. The

59

authors further speculated that since team performance is peripheral to lowly identified fans’ self-concept there is no need to protect their in-group’s value, even after a loss. Rather, because the relationship is less complex with out-groups (us versus them) and therefore differences are maximised, lowly identified group members are likely to engage in perceived in-group stereotypical behaviour. This is likely to include out-group derogation in order to gain acceptance (see Branscombe et al, 1999), as well as boosting self-esteem. This would therefore support corollary one of the Self-Esteem Hypothesis (see Hewstone et al, 2002, and Rubin & Hewstone, 1998). In a later study by Noel, Wann & Branscombe (1995), it was found that when low status members believed fellow group members would see their responses towards the out-group, they were much more likely to derogate the outgroup in order to try to secure a position in the desirable in-group. In contrast, high status members showed derogation both publicly and privately but to a lesser degree of extremity than lowly identified fans. This finding provides support that group members are likely to experience other threats such as acceptance threats, as suggested in this thesis and by social identity perspectives (see Branscombe et al, 1999). Similar findings have also been demonstrated in other social groups in what is known as the black sheep effect (see Marques, Yzerbyt, & Leyens, 1988). Members who were found to dissociate and deviate from perceived group norms were disliked more by fellow group members (see Castano, Paladino, Coull, & Yzerbyt, 2002; Dietz-Uhler, 1996; Marques, Abrams, & Serodio, 2001; Matthews & Dietz-Uhler, 1998). Ostracising these group members stops the group’s status from being dragged down in comparison to the out-group, and therefore the ingroup can maintain a positive social identity.

60

Protecting one’s self-esteem and positive identity is clearly important to fans and the elaborate strategies fans use to protect and enhance their social identity and self-esteem illustrate this point. These strategies also provide strong evidence for the role of social identity theory in explaining sport fans’ responses and that team performance may be less important to them since fans can easily manage their identity and self-esteem after their team loses. However, there are some limitations to past research examining sport fans’ management of their selfesteem; this will be the focus of the next section.

2.4.4. Measuring self-esteem In previous sections, it was suggested that one benefit of committing to a team was to elevate one’s self-esteem. The amount of self-esteem a fan gains from their team appeared to be related to the amount of identification a fan has with their team. Similarly, the extent to which fans use image management strategies in order to enhance and protect their self-esteem is clearly complex and reported to be related to the amount of identification fans have with their team. These relationships have been well established in sport fan literature and the impact it has on their social well being and sense of self (Branscombe & Wann, 1994; Dietz-Uhler & Murrell, 1999; Murrell & Dietz, 1992; Wann et al, 2003; Wann & Pierce, 2005; Wann, Roberts, & Tindall, 1999; Wann, Royalty, & Roberts, 2000). It is unsurprising to find that similar relationships have also been shown in more general work on psychological well being and group membership in social psychology and related fields of psychology (e.g. Andreopoulou, & Houston, 2002; Bairner, 2001; Bettencourt, Charlton, Eubanks, Kernahan, & Fuller, 1999; Blaine & Crocker, 1995; Crocker & Park, 2004; Stangor, 2004).

61

Although the impact of social identity for one’s self-esteem has been well documented (Abrams & Hogg, 1999; Crocker & Major, 1989; Crocker, Major, & Steele, 1998; Crocker & Park, 2004; Ellemers, Spears, & Doosje, 1997; Spears, Oakes, Haslam, & Turner, 1994; Oakes, Ellemers, & Haslam, 1997), several authors argued that past research attempts have been flawed. This was because past attempts assessed personal rather than social self-esteem, self-esteem derived from knowledge of one’s membership within a group (Aberson & Howanski, 2002; Abrams & Hogg, 1988; Crocker, Blaine, & Luhtanen, 1993; Luhtanen & Crocker, 1992; Ellemers et al, 20002; Rubin & Hewstone, 1998; Turner, 1999). The social identity perspective thinks about self-esteem in a different way, namely that it is not fixed as a psychological state (Hunter et al, 2004). Instead, this perspective distinguishes between personal and collective self-esteem. In further support of a collective approach to self-esteem, Long, Spears, & Manstead (1994), and Long & Spears (1998) demonstrated that those with low personal self-esteem displayed less in-group bias, and those with low CSE showed more in-group bias, therefore underlining the importance of distinguishing between the two. Luhtanen & Crocker (1992) developed the 16-item Collective Self-Esteem scale (CSES) to investigate social identity derived self-esteem. This scale consisted of four subscales (discussed further in Chapter Three). The main argument they propose is that most self-esteem scales fail to examine the amount of self-esteem and positive identity conceptualised within social identity theory. This is because an individual’s social identity is an evaluation of one’s social groups, as well as how one perceives others’ evaluations of those groups. They suggested that collective self-esteem operates in group contexts and personal self-esteem in individual contexts. For example, personal self-esteem moderates the extent to which an

62

individual responds to a personal failure, while collective self-esteem moderates the extent to which one responds to a group failure, which may also feel like a personal failure. In other words, in the same way that social identity distinguishes between personal and social identity, self-esteem according to Luhtanen & Crocker (1992), can be distinguished at a personal and social level too, both of which have implications on an individual’s self-concept. As discussed in Chapter One, Tajfel & Turner (1979) proposed that those with strong ties are more committed and tend to evaluate their groups more positively, while those who distance themselves from the group are more negative in their evaluations. However, as we have also seen it is possible to remain in a negatively evaluated group due to such strong ties, for example when group members engage in social competition to try and bring about social change or accept their low status as it differentiates them from other groups (e.g. Barreto & Spears, 1999; Ellemers et al, 1997; Mlicki & Ellemers, 1996). In this context, since sport fans go through elaborate strategies to protect their overall self-concept, assessing selfesteem has been limited in relation to team performances. For example, our social identities and commitments are only questioned when one’s identity is on trial (Branscombe & Wann, 1994), and for fans this has been after their team loses. Consequently, we have a narrow understanding of how self-esteem is affected in fans, which is limited to fans’ post game reactions. In addition, considering the impact status threat has on one’s collective selfesteem and commitment (Ellemers & Van Rijswisjk, 1995; Hornsey et al, 2003), relatively few studies have adopted the CSES to assess fans’ group derived selfesteem. Studies that have used this scale used its original wording that may be of little meaning to fans (Murrell & Dietz, 1992; Wann, 1994). Other studies have used the

63

scale inconsistently (Aberson & Howanski, 2002; Foels, 2006). This has been for example, by using only one of the subscales rather than all of the scale to represent one’s collective self-esteem (Aberson, 1999; DeCremer, 2001; De Cremer, Van Vugt & Sharp, 1999; Foels. 2006; Ethier & Deaux, 1994; Gramzow & Gaertner, 2005, Wann, 1994). Furthermore, the psychometric properties of the CSES have also recently been questioned (Utsey & Constantine, 2006). For example, few studies have gone beyond reliability testing of the scale when it has been adapted. This is important considering that social groups differ in nature (e.g. Deaux et al, 1995; Deaux & Ethier, 1998; Ethier & Deaux, 1994; Stangor, 2004) and therefore how we evaluate our groups’ positivity may also differ between groups. In terms of psychometric properties, the factor structures may also differ in that different aspects of self-esteem may be more prominent in different groups. Recently, Utsey & Constantine (2006) found that for their sample of Black Americans, the CSES did not adequately fit the original four-factor model suggested by Luhtanen & Crocker. They suggested that CSE is likely to be more complicated than originally thought, and that the factor structure of CSE may differ between groups. These authors however, did not adapt the scale and therefore the global version may not have necessarily been an accurate measurement for Black Americans. This could explain their inconclusive results about the scale’s psychometric properties and suggests that further investigation of an adapted version of the CSES’s psychometrics is warranted. Relating this further to the thesis, because self-esteem has been found to be an important consequence of fan commitment, and is strongly related to one’s positive social identity (e.g. Hogg & Abrams, 1990; Hogg, Terry, & White, 1995; Houston & Andreopoulou, 2003; Hewstone et al, 2002), it is important to have a reliable measure of fans’ self-esteem that was conceptualised by social identity. Therefore, another aim

64

of this study is to investigate the psychometric properties (e.g. construct validity) of an adapted version of the CSES for use with sport fans, and this aim is addressed in Chapter Three.

2.5 Non-performance status threats and sport fans

One theme that has run throughout the thesis is that previous literature about the social identity of sport fans is limited because researchers have relied predominately on the reactions of fans to team performances. Also, as we saw in the previous section, one issue strongly related to team performance within sport fan literature has been the importance of self-esteem, and particularly in relation to protecting and enhancing self-esteem using image management strategies. Since sport fans have a large number of strategies to cope with the threat of a team loss to their self-esteem, it may be logical then to presume that team performances do not actually have as big as effect on sport fans’ self-esteem as previously seen (see also Murrell & Dietz, 1992 for additional comments). Studies mentioned earlier for example, showed that although poor team performance can decrease some fans’ self-esteem temporarily (e.g. Bizman & Yinon, 2002), fans’ overall self-esteem remains relatively high regardless of team performance (Branscombe & Wann, 1995). As such, strong attachments to the team may operate independently from team performances. This is because group members are also likely to favour other group members (in-group bias) as a way of enhancing a positive social identity, and might be to the extent that they assume that the attributes and actions taken by the group are the correct ones (Abrams, Thomas, & Hogg, 1990; Abrams et al, 1990; Oakes, 2002; Turner, 1999). For example, part of social comparison involves exaggerating differences between in-groups and out-groups in order to maximise 65

positive distinctiveness (Dimmock & Grove, 2006, Hogg et al, 2004) and therefore self-esteem. Thus, fans may gain self-esteem by perceiving their groups’ actions and behaviour as ideological fan behaviour. Fans then, in addition to their team, may use and exaggerate their fellow fans’ characteristics as a way of enhancing a more positive social identity. For example, fans may exaggerate the importance of being committed and being seen as good fan, and therefore being seen to belong to a good supporter group. This is supported in the social identity literature where highly identified group members are more likely to think and act in terms of their social identity compared to lowly identified group members (Hogg et al, 2004, Wann et al, 2001), in this case fans of a specific team. This is despite recent research, which has found that sport fans in general are similar to each other (Davies et al, 2008). The literature about sport fans still largely limits itself to assessing one type of status threat (i.e. the group’s value, where the status-defining dimension has been game outcome). This is surprising despite recent social identity literature that identified several types of possible threats that can be experienced by group members. A taxonomy of social identity threats that may be experienced by group members was provided by Branscombe, Ellemers, Spears, and Doosje (1999), and Ellemers, Spears, and Doosje (2002, but see also Wann, 2006b) and these threats included:

1. Threats to the group’s a) value (that their team is not as good as others), and b) morality (that fans of one group are morally superior on qualities such as ‘sportsmanship’)

2. Acceptance threat (fans of their team do not view fellow fans as real or die-hard fans. In other words, an atypical group member) 66

3. Categorisation threat (fans categorise an individual as a fan of an alternative team, or a sport they do not feel a part of).

4. Distinctiveness threat (that fans of their team are seen as no different from fans of other teams). This thesis does not look each type of threat, rather it continues a long line of research that has explored group value threats based on status comparisons. In social identity research, despite this promising direction, threats to the groups’ value has been to a certain extent saturated (Ellemers et al, 2002), and this is particularly true of research with laboratory based groups. Similarly, sport fan literature examining the group’s value has been the focus of research, but has been done so based on a single comparative dimension, team performances. Therefore, elaborating on this limitation is one focus of this thesis. Although team performance is the most obvious and contextually relevant comparative dimension, social identity theory proposes that depending on the nature of the social identity threat, highly identified members are likely to be threatened by several dimension that are contextually and identity relevant (Abrams et al, 1990, Ellemers et al, 2002, Tajfel & Turner, 1979; Turner & Reynolds, 2001). This therefore, leaves a large gap in the literature about other types of comparative dimensions that may threaten the group’s value. In addition, such a large gap also raises questions. For example, how independently do status threats operate from each other, and similarly are group value threats and group moral threats mutually exclusive? Furthermore, another line of inquiry may be to examine how valued each type of threat is for fans’ social identity, and as such, to consider whether there is a hierarchy of status threats. For example, are comparative

67

dimensions that implicate group morals more important than dimensions that implicate the group’s value? According to Brown (2002) however, one comparative dimension is not more valued than another, regardless of whether the evaluative outcome is favourable or not. In this thesis, the focus remains on the group value threats, and in particular examining different types of comparative dimensions that may cause social identity threats. However, in contrast to Brown’s suggestion, it is argued in this thesis that because the in-group’s characteristics (such as being seen to be committed) may be one dimension that fans may use, that when negatively manipulated, group members will find this more threatening than team performances. This is because threatening such ideological fan behaviour, a comparative dimension that they can control, might mean fans feel more responsible for the positivity of their social identity. Consequently, fans may feel more threatened when this dimension is manipulated, particularly since they cannot use it as a creativity strategy suggested by Wann (2006) (i.e. we may have lost but we are still committed fans). Although much of the research about sport fans’ reactions to game outcome have yielded similar responses to other groups in social identity literature (e.g. ingroup bias, Branscombe & Wann, 1994; in-group favouritism, Dimmock, Grove, & Eklund, 2005; Wann et al, 2002; Wann & Grieve, 2005), as well as image management strategies (Wann & Branscombe, 1990), these have been based on team performances. This is despite several studies that have found fan’s value their ingroup as well as their team (Branscombe & Wann, 1995) and are keen to declare their fans’ superiority over rival fans (Wakelin et al, 2005; Wann & Branscombe, 1990; 1995). Based on Branscombe et al’s taxonomy of threats, it is likely that there are other comparative dimensions that fans may use to evaluate their groups’ value at

68

a game. It would be hard to imagine a sport fan who would not feel threatened when their groups’ commitment is questioned, particularly when fans often have consensual agreement about their group’s identity and prototypical features (cf. Davies et al, 2008; Dimmock & Grove, 2006; Hogg et al 2004; Oldmeadow et al, 2003; Reid & Hogg, 2005). Given that social creativity allows the use of several dimensions to validate an in-group’s superiority and positive distinctiveness over a rival group when they are inferior on another (Barreto & Ellemers, 2000; Ellemers, Barreto, & Spears, 1999), fans then may also be creative in their social comparative dimensions. This thesis proposes that one way they do this is by accentuating their group’s commitment to their team, especially after a loss. Social identity research has found that one way group members ‘collectively resist’ and defend their group’s value under threat is by increasing self-stereotyping (Doosje & Ellemers, 1997; Ellemers et al, 1997; Jetten, Postmes, & McAuliffe, 2002; Verkuyten & Nekuee, 1999). Therefore, if the in-group are used as a valued comparative dimension, questioning this will also threaten the group’s value. Therefore, using the in-group as a comparative dimension establishes a novel context in sport fan literature. Additionally, there is further justification for examining the in-group’s value for fans’ social identity and self-esteem. Unlike team performance, the content of the threat is rather different. For example, fans commitment is a much less explicit and observable threat compared with game outcomes. Further, comparison of fans’ commitment is particularly salient after a loss, when fans can explicitly state and demonstrate their continued commitment to their team. In addition, fans are only likely to encounter such threats to their group’s value via out-group derogation or appraisal from other fans. In this way, questioning

69

commitment in this context is potentially as explicit for a fan as viewing their team lose. Questioning fans’ commitment after a loss however, would be experimentally difficult. This is because team performance (particularly a loss) would act as a confounding variable since both a team’s loss and the in-group’s commitment would be associated. Researchers are left with separating the effects from team performance from threats based on the in-group’s commitment on whichever construct (e.g. selfesteem). Additionally, since questioning the in-group’s commitment is likely to come from an out-group (the object of identity threat), rival fans also act as the source of the threat as they can intentionally direct the derogation and appraisal to the ingroup. Therefore, relating this to social identity theory, rival fans may be seen as a more stable (e.g. having expert knowledge) and legitimate source of threat to the fans than the rival’s team (see also Branscombe, Ellemers, Spears, & Doosje, 1999, p46). Other studies have found that certain social reality constraints can affect the degree to which a threat is experienced. In one study, the importance of an audience increased perceptions of accountability for the group’s image and was moderated by amount of commitment to the group. Ellemers, Van Dyck, Hinkle, & Jacobs (1998, study one) tested sport fans who were given ratings about their group’s performance. Their study showed that those in low status groups favoured the out-group ‘publically’, but favoured the in-group privately. As such, they found that members still had confidence in their groups’ worth whilst acknowledging the out-group’s superiority on the particular status-defining dimension. In their second study, Ellemers and colleagues measured fans’ identification as a moderator of strategic responses to group ratings. They found that high identifiers acknowledged an inferior status with other members (intragroup), while they denied the low status publically (intergroup context). Furthermore, highly identified fans after acknowledging a low

70

status (intragroup context) that needed to be resolved, were more likely to exert effort and incur personal costs to improve the group’s status and uphold the group’s worth in front of an audience (intergroup context). Lowly identified members however, still emphasised the in-group’s worth instead of focusing on improving status, but were not prepared to maintain this in front of an audience. Therefore, these fans acknowledged the in-group’s inferior status in an intergroup context. The authors concluded that lowly identified fans refrained from in-group favouritism in intergroup settings out of strategic consideration (e.g. positive image, self-esteem), rather than lack of trust in the in-group’s worth. For sport fans, proving one’s worth as a die-hard fan is an important criterion (Wann & Branscombe, 1990), whereas the team realistically has little to say about the worth of the fans. As we will see later this notion may be one reason why fans bask in their team’s glory (Cialdini & Richardson, 1980) and then lowly identified fans cut off reflective failure (Snyder, 1983). Rather than leaving their team, fans are adapting their responses to the context, and in this thesis, it is argued that fans are operating in terms of group normative behaviour rather than using their team as a reflection of their social identity status. In another study, the size of the source of threat has also been found to affect fans’ threat perceptions of threat (see Doosje, Spears, & Koomen, 1995). Another reason the team may be less important is because of the size of the ‘team-group’ in comparison to the volume of in-group members that come to watch the team. For example, Doosje, Spears, & Koomen (1995, study 2), found that the reliability and variability of evidence for an unfavourable comparison with an out-group was affected by the size of the sample the evidence came from. A small sample is less compelling for the status quo than evidence from a large group, and hence a small

71

group puts more constraint on in-group member’s social reality about the intergroup differences. In-group participants depicted as having an unfavourable comparison with an out-group, were unlikely to accept that differences were typical of their group when the evidence came from a small sample of their in-group. When applied to sport fans, the threat from a rival team’s performance (a team of rugby players has 15 players on the pitch) is overwhelmingly small in contrast to the amount of fans in attendance (Welsh premier rugby games can have attendances of 6000 fans and more). Therefore, following a loss, the losing team’s fans may depict to the rival fans, that although their team is inferior it does not reflect that in-group fans are inferior. In an effort to cope with the threat to their social identity from team performances, fans align their identity with the in-group. Therefore, although their group may be inferior on one status defining dimension (the team), the fans are superior in other ways such as commitment, and this is something that they can prove to rival fans (cf. Wakelin et al 2005; Wann, 2006).Moreover, to contrast further with the content of team performance as a social identity threat, threats to the in-group’s value differ in relevance. A threat to the group’s characteristics and commitment would be much more personally relevant since the team act as an extension of their in-group. As previously discussed, and perhaps crucially, because fans have a large degree of control over their image as supporters, increases their sense of responsibility for their social identity. It may therefore, be much more difficult to explain away or dissociate with their low ‘in-group’ status by external factors (e.g. poor refereeing, lack of effort; see Mann, 1974; Wann 2006a) than when the failure is their team’s. Lastly, unlike for team performance, threatening the in-group directly could affect other types of threats. For example, in Branscombe et al’s (1999) taxonomy of

72

social identity threats, each type is more directed at fan behaviour and is separate from any threat that could be caused by the team. This highlights the limitation of using team performance because it is only relevant for one type of threat, group value. By examining other types of comparative dimensions, and in particular the ingroup’s characteristics, this may concurrently also threaten fans’ acceptance, categorisation, and distinctiveness as proposed by Branscombe and her colleagues (1999). After a loss, for example, fans may wish to dissociate with their team but doing so may lead to detection as a fair weather fan and potential ostracism from the in-group (acceptance threat).This can also lead to being categorised in the first place as fair-weather, or not a true sports fan (categorisation threat). Furthermore, by the group having an uncommitted fan in their midst, this might threaten the in-groups’ positive distinctiveness (distinctiveness threat) since fans accentuate their commitment to differentiate themselves from fans of other teams (Davies et al, 2008; Wakelin et al, 2005). Having an uncommitted fan as part of your group would mean being seen as similar to other inferior supporter groups. Therefore, the need to affiliate with a group of ‘good’ fans may be stronger than the need to affiliate with a ‘good’ team and fans may actually be committing to their group’s values (the social identity) rather than the team (the object of their social identity). To avoid additional threats, and in line with the key proposal of this thesis, it would be more beneficial and less threatening for fans to be seen as good supporters, rather than supporters of a good team, as this may potentially enhance the image of the in-group. Only one study to date has thus far examined comparative dimensions other than performance related threats. In a recent study, Wann & Grieve (2005) investigated whether, in addition to game outcome, fans would be threatened when their home ground was visited by rival fans (game location). The authors found that

73

both in-group and out-group fans displayed in-group bias/favouritism and this intensified after the in-groups’ team had won (identity confirmation, see Scheepers, Spears, Doosje, & Manstead, 2003). Furthermore, the home team fans felt greater threat to their social identity compared to visiting fans, such that home fans perceived a greater number of visiting fans intruding their personal space. The most threatened fans were those who were highly identified, whose team had won, and who were playing at home. These fans displayed the most in-group bias. The authors concluded that the use of in-group favouritism is an important coping strategy amongst highly identified fans, as this would have the most benefits to enhancing one’s self-esteem for this group. However, the authors suggested that the salience of their social identity coupled with being in the sporting environment in general elevated in-group bias and that the additional threats from game outcome and game location intensified levels of bias. From the above study, it can be seen that fans displayed strong group biases regardless of who they supported. In line with social identity theory, displaying ingroup favouritism and bias may provide an alternative strategy to ensure a positive evaluation of the group. This has been supported in the social identity literature using both minimal group paradigms and real world intergroup contexts (Billing & Tajfel, 1973; Fein & Spencer, 1997; Jetten, Spears, Hogg, & Manstead, 2001). Where fans may accentuate in-group favouritism for favourable comparisons (e.g. Jetten et al, 2005; Wakelin et al, 2005) after a loss, when the features of in-group favouritism are questioned, this may also act as a source of group value threat. Using the in-group’s features as a source of threat may actually accentuate the threat considering that fans may use their in-group as part of a socially creative strategy (changing the comparative dimension; cf. Jetten et al, 2005) to obtain positive distinction from

74

other fans after a team loss. Therefore, using the in-group as an independent variable rather than team performances, is a novel aspect of this thesis. However, since the role of commitment and its use as a status-defining dimension is the recurrent theme throughout the thesis, understanding exactly what commitment is, and what it means to fans, is crucial to this thesis. Therefore, the next section introduces literature about the nature of commitment. In particular, it will focus on commitment in terms of how it has been conceptualised and measured in previous research.

2.6 Commitment and sport fans

For many fans, being committed is an important component for their fandom (Davies et al, 2008, Wakelin et al, 2005). Although there are no formal rules and regulations for joining a supporter group of a specific team, leaving one’s team particularly after defeat would almost be an unthinkable act for most committed sport fans (Ellemers, Spears, and Doosje, 1997; Wann & Branscombe, 1990). In addition, understanding fans’ commitment is also important for many sport clubs, particularly as their success as a club depends on maintaining and developing a committed supporter base. Research within consumer psychology has often concluded that fans’ attachments have the biggest influence on attendances (Kwon & Armstrong, 2004; Fisher & Wakefield, 1998; Mahoney et al, 2000; Sutton, McDonald, Milne, & Cimperman, 1997; Wann, Bayens, & Driver, 2004). Subsequently, much of the research in social psychology examining sport fan behaviours has distinguished between highly identified and lowly identified fans’ reactions. For example, in social identity theory, group members who are highly identified are likely to spend most of their time thinking and acting in terms of their identity (Hogg et al, 2004). Consequently, highly identified sport fans 75

derive much of their self concept and social identity from supporting a team (see Wann et al, 2003; Wann & Pierce, 2005). These fans also tend to enjoy games more, invest more time and money in their teams (Pan & Baker, 2005; Shank & Beasley, 1998; Wann et al, 2001), and remain highly committed to their teams after losses compared with lowly identified fans (Ellemers, Spears, & Doosje, 1997; Wann & Branscombe, 1993; Wann et al, 2001). This supports much of group membership research in social identity literature that has shown the importance of group commitment (e.g. Jetten et al, 2005, Ouwerkerk, et al, 1999). As discussed in the previous sections, highly identified and committed fans produced a variety of behaviours and feelings such as in-group bias/favouritism and out-group derogation (see Branscombe & Wann, 1994; Cialdini et al, 1980, End, et al, 2002; Wann & Grieve, 2005). The degree of identification affected fans’ feelings and self-esteem (Crisp et al, 2007; Hirt et al, 1992, Murrell & Dietz, 1992; Wann et al, 2000; 2003) as well as influenced how fans deal with social identity threats by engaging in self-presentational strategies (see Berache-Assollant, Lacassagne, & Braddock, 2007; Bizman & Yinon, 2002; Branscombe, Wann, Noel, & Coleman, 1993; Wann, 2006a). As not all fans identify to the same degree, one outcome of sport fan research within social psychology has been to measure how attached fans are to their teams. Commitment, in terms of social identity has been the label given to the amount of affective attachment individuals have with their groups (Cameron, 2004; Ellemers et al, 1999), but other perspectives have conceptualised commitment in different ways, and is the focus of the next section.

76

2.6.1 Conceptualising commitment Within a broader context in psychology, the concept of group commitment has been examined in relation to consumer behaviour to brands (e.g. Thatcher & George, 2006), in relation to one’s work team and organisation (e.g. Allen & Meyer, 1990; Cooper-Hakim & Viswesvaran, 2005; Meyer, Becker, & Vandenberghe, 2004), and in relation to sport participation (e.g. Alexandris, Zahariadis, Tsorbatzoudis, & Grouios, 2002; Backman & Crompton, 1991). It is therefore unsurprising to find that commitment has been defined in numerous ways. This has included being defined in relation to possessing behavioural, affective, and attitudinal components (see Bergami & Bagozzi, 2000; Ellemers, Spears, & Doosje, 2002; Iwasaki & Havitz, 1998; 2004; Mael & Ashforth, 1992; Pritchard, Havitz, & Howard, 1999). Commitment has also been defined within reference to other constructs such as identification and loyalty that often contain similar characteristics (see Bergami & Bagozzi, 2000; Doosje, et al, 1999; Edwards, 2005; Ellemers et al, 1999; Funk & James, 2004; Harris & Cameron, 2005; Wann, 2006b). Other definitions however, distinguish between different types of commitment (e.g. Allen & Meyer, 1990; 1996; Delobbe & Vandenberghe, 2000; Meyer & Allen, 1992). Despite the number of approaches, most of these definitions contain similar attributes where the outcome of commitment is to make some form of act, attitude, or feelings less changeable in response to conflicting information. According to Meyer, Becker, & Van Dick (2006), commitment is a force that binds an individual to a target (social or non-social) and to a course of action of relevance to that target. As such, commitment has also been described as one’s strength of attachment to the group and/or other associated objects (Cooper-Hakim &

77

Viswesvaran, 2005; Crosby & Taylor, 1983; Doosje et al, 1999; Kwon & Armstrong, 2004, Ouwerkerk, 1999; Wann et al, 2001). Defining

commitment

is

complicated

further

because

the

term

‘commitment’ is often interchanged with’ identification’ (see Edwards, 2005; Wann, 2006b), and these are assumed to be describing similar attachments, and to be synonymous with each other (Bergami & Bagozzi, 2000; Ellemers et al, 1999b; Ellemers & Rink, 2005; Funk & James, 2004; Miller, Allen, Casey, & Johnson, 2000). This has subsequently caused confusion in social identity and organisational literature that has seen different labelling for similar constructs, and similar labelling for different constructs (Edwards, 2005; Harris & Cameron, 2005; Meyer et al, 2006). This is complicated further by other research suggesting that both constructs are actually distinct from one another (Ashmore et al, 2004; Edwards, 2005; Harris & Cameron, 2005; Meyer et al, 2006; Wann & Pierce, 2003).Alternatively, some have described identification processes as an antecedent of commitment (Ashforth & Mael, 1989; Meyer, Becker, & Vandenberghe, 2004). Still however, there has been little agreement about the distinction between identification and commitment, or the relationship between them. Rather, some have suggested that it may be more fruitful to reconceptualise our understanding of group attachments by integrating both constructs (Meyer, Becker, & Van Dick, 2006). Although for the most part they may be related (Wann & Pierce, 2003), and for many sport fans may mean the same thing, the literature within psychology has seen little agreement between commitment and identification. As such, it is often unclear to what extent ‘highly identified’ and ‘highly committed’ refer to the same thing. Within sport fan research, borrowing from social identity theory, Wann

78

(2006b) defined sport fan identification as “the extent to which an individual feels psychologically connected to the team” (p332). This definition, however provides a more limited definition of social identity by encompassing all of the subcomponents (affective, cognitive, and evaluative) proposed by Tajfel, (1978) under ‘psychological connectedness’, and thus appears as a unidimensional definition. In one way, this makes understanding fan attachment easier but also detracts from the multidimensional nature of our social identities intended in social identity theory (see Ellemers et al, 1999b). The lack of consensus about commitment as a construct, is partly because of the role that social identity theory initially had in influencing conceptualisation of group attachments in work groups (e.g. Ashforth & Mael, 1988) and sport fans (Wann & Branscombe, 1993). However, different interpretations and lack of understanding social identity theory have been prevalent within research examining group attachments. This is not without reason, because until recently the conceptualisation of our social identities and how group members commit to them had also been given little attention within the social identity tradition (Cameron, 2004, Dimmock & Grove, 2006; Ellemers et al, 1999; Leach et al, 2008; Obst & White, 2005). Examining commitment in this thesis therefore has implications for how we conceptualise commitment to our identities in a wider context. Returning to a social identity perspective, group commitment as we saw in Chapter One, is conceptualised as the extent of people’s social identification with their group which determines their inclination to behave in terms of the group and engage in group serving strategies to enhance their positive distinctiveness (Doosje, Ellemers, & Spears, 1999; Ellemers et al, 1999a; Ouwerkerk et al, 1999). In an attempt to offer a resolution to this conceptual confusion, which to a certain

79

extent is due to semantic confusion, Ellemers, et al (1999b) return to the original conception of social identity theory. They suggested that in order for an individual to commit to a specific social identity, they must see themselves as part of that group, and more importantly, a strong emotional component that binds them (i.e. affective commitment. 2) must be present. This is because in their original conception, Tajfel & Turner (1979; p63) defined social identity as having three distinct components, or in other words, three types of identification (Dimmock, Grove, & Eklund, 2005). These included a cognitive identification (selfcategorisation; self-awareness of belonging to a group membership), evaluative identification (social comparison, positive or negative value connotations of the group), and an emotional identification (affective commitment, the amount of strength of their identification). In this sense, Ellemers and her colleagues argued that the strength of [social] identification is primarily used to refer to a feeling of ‘affective commitment’ to the group (the emotional component), rather than the act of identification to distinguish between members of other categories (the cognitive component), as well as having a cognitive awareness of membership in a collective. By separating these components, it becomes easier to understand the differences in degree of committed behaviour. For example, in sport fans this may be seen in fluctuating attendances or differences between ‘fair-weather’ and ‘diehard’ fans in contrast to those who support a team because of some basic categorical connection (e.g. a city team).

2

The emotional component is often referred to as affective commitment but also used interchangeably with social identification to indicate strength of emotional ties. Similarly, identification can also be confused with the cognitive component of social identity, which is the degree that one feels perceptually part of a specific group, and involves categorising or differentiating groups (Ellemers et al, 1999b; van Dick, Wagner, Stellmacher, & Christ, 2004). In addition, although both are related they are also distinct components within our social identities.

80

Social identity’s conceptualisation of commitment is supported by numerous studies showing that ‘affect’ is a good predictor of whether one will remain in a group regardless of a negative image (see Bergami & Bagozzi, 2000; Ellemers et al, 1997; Ellemers, De Gilder, & Haslam, 2004; Mlicki & Ellemers, 1996; Van Dick, Wagner, Stellmacher, & Christ, 2004; Wann & Branscombe, 1993). In addition, affect has been found to moderate which type of group members engage in group enhancement strategies rather than individual strategies as previously discussed (e.g. Branscombe & Wann, 1994; Doosje et al, 1999; Ellemers & Van Rijswijk, 1997; Jackson et al, 1996). Furthermore, high affective commitment has also been found to be a good predictor and moderator of other group behaviours and feelings, such as levels of self-esteem, in-group bias as well as out-group derogation (Branscombe & Wann, 1994; Dietz-Uhler & Murrell, 1998; Dimmock, Grove & Eklund, 2005; Jetten Spears, & Manstead, 2001; Jetten et al, 2005). These effects however, have not been found as strongly as a consequence of cognitive identification (Ellemers et al, 1999b; Ellemers, Spears, & Doosje, 1997; Van Dick et al, 2004). Therefore, where some may strongly identify with a group in terms of a cognitive identification as an attempt for self-definition and uncertainty reduction (see self-categorization theory, see Abrams et al, 1990; Cicero et al, 2005; Edwards, 2005; Hogg, 2000; Turner et al, 1987; but also uncertainty reduction hypothesis Reid & Hogg, 2005), the key difference for committing to the group is strength of affective ties. Commitment therefore, from a social identity perspective is represented by an affective component only. In addition, other research investigating the nature and construct of social identities has confirmed similar conceptualisations of commitment (e.g. Cameron, 2004; Dimmock, Grove and Eklund, 2005; Dimmock & Grove, 2006; Deaux,

81

1996; Jackson; 2002; Jackson & Smith, 1999; Obst & White, 2005). However, other research has indicated that additional components may also be present that may not be conceptualised by social identity’s initial definition. These may include for example, several behavioural components (Ashmore et al, 2004, Jackson, 2002). Despite conceptual variations and numerous definitions of commitment, much of the research presents a strong case that like social identity, commitment is also complex and multifaceted that can take different forms (Allen & Meyer, 1990; 1996; Fullerton, 2003; Harris & Cameron, 2005; Iwasaki & Havitz, 2004; Kyle, Graefe, Manning, & Bacon, 2004; Meyer, Becker, & Vandenberghe, 2004; Meyer et al, 2006; Pritchard, Howard, & Havitz, 1999). A multidimensional approach to commitment has strongest support within organisational psychology. For example, Delobbe & Vandenberghe (2000) distinguished between four types of commitment: internalisation, compliance, affective,

and

continuance

commitment.

In

addition,

a

more

popular

conceptualisation has been put forward by Allen & Meyer (1990) and Meyer & Allen (1991). According to these authors, three types of commitment can be distinguished. This includes affective commitment (strength of one’s emotional attachment, those that remain and go to work because they want to); normative commitment (feelings of obligation because they ought to remain with the organisation); and continuance commitment (refers to commitment based on the costs and benefits that employees perceive with staying or leaving). Relating this conceptualisation back to sport fans for example, fans that may have been brought up with their team may experience higher normative commitment than those who may have started their support much later in their life. Additionally, fans who attend with friends for social reasons may be higher in continuance commitment than those who do not. From this

82

conceptualising of commitment, it becomes much harder to state simply that one fan is more attached than the other, or highly or lowly committed, but rather may be committed in one or all three types of commitment. In comparison with a social identity perspective, Allen & Meyer’s conceptualisation of commitment has a strong overlap with and clear similarity, in that both perspectives have an affective component. Like social identity literature, affective commitment has been shown to be a strong predictor of group behaviour and feelings. For example, organisational literature based on Allen & Meyer’s conceptualisation has found affective commitment to be a robust predictor of work place relevant thoughts, job satisfaction, and intention of turnover (Allen & Meyer, 1996; Bergami & Bagozzi, 2000; Cole & Bruch, 2006; Ellemers DeGilder & Haslam, 2004; Meyer & Allen, 1997; Meyer & Hersocovitch, 2001). A multidimensional conceptualisation, therefore, shows that individuals can commit to groups in different ways, and it is unlikely that commitment is purely affective as suggested by the social identity tradition (e.g. Ashmore et al, 2004; Allen & Meyer, 1990; Delobbe & Vandenberghe, 2000; Edwards, 2005; Pritchard et al, 1999). Within sport fan research, a similar proposal was made by Park & Kim (2000). They proposed that fanship is complex in nature and that fans’ commitment to a team is likely to be multi-dimensional. For example, in line with other group and social identity researchers (e.g. Ashmore et al, 2004; Jackson, 2002), Kim, Scott, & Crompton (1997) suggested that more behavioural elements of commitment other than attendances need to be measured, given that lowly identified fans may still display committed behaviours (see also Wann et al, 2001 for a distinction between spectators and fans; p2).

83

Therefore, considering the conceptual differences about commitment further warrants investigation. This is particularly important as conceptual confusion may also lead to inconsistent interpretations of fan behaviours (Dimmock, Grove & Eklund, 2005; Wann & Pierce, 2003). This is specifically achieved in Chapter Three of this thesis where commitment is investigated using a mixed methodology to provide a more meaningful and conceptually accurate tool for assessing fans’ commitment.

2.6.2 Measuring sport fans’ commitment The availability of appropriate measures of fans’ attachments to their team is severely limited and generally, there is a paucity of measures examining fans’ attachment. This is similarly the case within social identity literature examining group members’ attachments to their social identities (Cameron, 2004; Jackson, 2002). For example, at the start of this thesis, there had been only one attempt in social psychology literature to develop a psychometric measure of fans’ psychological attachments toward a sport team (the Sport Spectator Identification scale; SSIS Wann & Branscombe (1993; appendix A). There have however, been several attempts to measure sport fans’ attachments and motivations in other academic disciplines (e.g. Funk & James, 2004; Kwon & Trail, 2003). In particular, one popular measure from consumer and marketing psychology is the Psychological Commitment to Team scale (PCT, Mahony, Madrigal, & Howards, 2000; appendix B). More recently, Wann & Pierce (2003) found that the SSIS and PCT were empirically related and purported to be measuring similar fan behaviours. This highlights the argument made earlier that different labels are used to measure similar constructs. Despite this, the PCT has

84

been used as a competing measure to Wann & Branscombe’s SSIS for assessing fans’ attachments, and has also helped to fuel the interchange of the terms commitment and identification discussed in the previous section. The SSIS, based on the social identity approach, assesses the degree of identification with a sport team (Wann & Branscombe, 1993). Wann & Branscombe (1993) proposed a seven item Likert scale where strength of identification lies on a continuum from low to high. In addition, this measure of fans’ attachment has been the most extensively used scale in examining sport fan attachments; it has been employed in over a hundred studies of fan behaviour (Wann, 2006b). This scale has been reported to have good psychometric properties (Wann & Pierce, 2003; Wann, 2006b) such as internal consistency, test retest reliability, and general construct validity (Theodorakis, Vlachopoulos, Wann, Afthinos, & Nassis, 2006; Wann. 2006b; Wann et al, 2001; Wann & Pierce, 2003). In contrast, the PCT is based on evidence from consumer attitudes towards brand preferences (see Haugvedt & Petty, 1992; Haugtvedt & Wegener, 1994) in which resistance to change was the critical factor underlying psychological commitment. In addition, the authors believed fans exhibited both behavioural and attitudinal loyalty and that using their proposed 14 item measure would allow prediction and segmentation (psychographic distribution) of sport fans’ commitment into four types. They distinguished between fans who had true loyalty (high

commitment

and

behaviour

consistency),

spurious

loyalty (weak

commitment but high behaviour consistency), latent loyalty (high commitment and low behavioural consistency), and low loyalty (neither high in commitment or behavioural consistency). In general, the PCT has also been shown to possess good

85

internal consistency, as well as construct and predictive validity (Mahoney et al, 2000; Wann & Pierce, 2003). However, despite initially good psychometrics, neither measure has been tested using more stringent psychometric methods such as confirmatory factor analysis. In contrast to Wann & Pierce (2003), Kwon & Trail (2003) in an attempt to evaluate the PCT more rigorously found that it did not have good construct validity and did not represent a good example of psychological commitment. In addition, although both the SSIS and the PCT developed in related yet separate disciplines based on different principles, Wann & Pierce (2003) found that the SSIS appeared to be a better predictor of overall ‘fandom’ but that both models were strongly related. This research shows that firstly, both measured a similar construct and secondly, researchers are unclear about which construct they are trying to measure. Furthermore, Wann & Pierce later suggested that both the PCT and SSIS also measure the three types of commitment developed by Allen & Meyer (1990) in the organisational literature. As such, Wann & Pierce (2003) claim that these models measure both social identity and commitment, but then also other types of commitment that are generally unrelated to social identity theory or consumer marketing literature. Given that Wann & Pierce found that both scales were highly related and overlapped, it may then be logical to question the construct validity of the SSIS, since Kwon & Trail (2003) found the PCT to have poor construct validity. These models then are in need of further scrutiny and analysis. For example, factor analysis and confirmatory factor analysis are now becoming a standard method for examining inventories for measuring psychological constructs (DeVellis, 2003).

86

Finally, in terms of the items,

according to Dimmock, Grove, & Eklund (2005), both scales have items that measure team identification indirectly by inferring it from other responses. It is clear that both scales are limited in the full scope of fans’ commitment. In particular, one problem with both of these models is that although both approach fan attachments from differing perspectives, both are unidimensional. This is despite social identity and commitment being conceptualised as multidimensional. Therefore, the usefulness of these scales has also been brought into question (Dimmock, Grove & Eklund, 2005, Kwon & Trail, 2003; Wann & Pierce, 2003). The unidimensional limitations of these measures makes it difficult to try and understand how fans get from high to low, and secondly which aspect of identification and commitment moderate fans’ behaviours and feelings. One alternative to these measures of fan attachments may be to use measures from more generic social identity research and then to adapt them. However, in doing so, researchers investigating sport fan attachments would also be limited to unidimensional scales. This is because most popular scales are derived from Brown, Condor, Mathews, Wade, & Williams’ (1986) measure of social identity (Cameron, 2004). These scales often denote item directionality (positive or negative scoring) rather than construct dimensionality (Obst & White, 2005) and we are left then with having a limited conceptualisation of a multidimensional construct when realistically some highly identified individuals may score differently on different dimensions (Dimmock Grove & Eklund, 2005). Although offering an alternative measure for fans’ attachments was one purpose of this thesis, recently Dimmock, Grove, & Eklund (2005) provided a multidimensional model of fan attachment based on social identity theory. They also recognised the paucity of suitable measures assessing sport fans and the lack

87

of rigorous psychometrics employed in the available measures. Instead, Dimmock and his colleagues used items from several similar social identity measures to represent individual components of social identity proposed by social identity theory (see Tajfel & Turner, (1976, p63). Using confirmatory factor analysis they demonstrated that fan attachment to their team could be conceptualised as a three dimensional construct consisting of a cognitive, affective, and evaluative component proposed by the social identity perspective. This finding also supports similar conceptualisations and measures of social identity (e.g. Cameron, 2004; Ellemers et al, 1999b; Obst &White, 2005), which is not surprising considering Dimmock and his colleagues based their scale on these generic tri-partite scales of social identity. However, although this model of identification is a substantial improvement on previous attempts with strong theoretical groundings, there are two limits to the measure. Firstly, and crucial to this thesis, is that sport fans are self-selected and as such joining these groups are more voluntary. Fans then are likely to have a very strong emotional bond with their team unlike members in other assigned groups (Ellemers et al, 1999b). This measure of social identification did not take self-selection into consideration by asking fans what it means to be identified and committed to their team. Rather, this measure was strongly theory based, despite research having conflicting views about our social identities and commitment (Obst & White, 2008). Considering fans are likely to differ between individual components (e.g. Ashmore et al, 2004; Deaux, 1996; Deaux et al, 1995; Dimmock et al, 2005; Ellemers et al, 1999b), then they may also differ qualitatively between their meaning and understanding of their attachments. Secondly, although grounded in social identity theory, there is strong evidence that

88

commitment is also multidimensional. Therefore, this measure of fans’ commitment is limited to measuring commitment based on social identification (affective commitment) rather than the wider gamut of commitment types proposed in other disciplines (see Allen & Meyer, 1990; Backman & Crompton, 1991; Meyer et al, 2006; Pritchard et al, 1999). To extend previous research into fans’ commitment to their team, one part of this thesis presents and develops a competing model of attachment. Instead of adopting a specific approach (e.g. social identity theory), a more novel approach of asking fans about their commitment was adopted. In doing so, it was hoped this method would yield more valid and conceptually accurate responses in which to develop a tool for assessing fans’ commitments to their teams. As previously discussed, for many fans just saying you are committed is not enough, and for many of them being seen to be committed may be more important than strong affective commitment to the team. In the thesis, the model developments are presented first because the second research chapter involves employing the newly developed measures in the field.

2.7 Summary The main objective of this chapter was to elaborate the main arguments and key proposals presented in Chapter One. Throughout both chapters, the importance of the construct commitment was highlighted, and particular attention was given to outlining the limited understanding of commitment in social identity and sport fandom literature. Specifically, the social identity approach and sport fandom literature have conceptualised the construct commitment as unidimensional. This approach is seemingly at odds with the majority of literature from other fields of

89

psychology that have found that commitment is more likely to be a multidimensional construct (e.g. Allen & Meyer, 1990). Furthermore, social identity theorists (see Ellemers et al, 1999a; Luhtanen & Crocker, 1990; Rubin & Hewstone, 1998) generally agree that other components of social identity, such as collective selfesteem are multidimensional, but have neglected to explore whether this may also be the case for group members’ commitment. To resolve these limitations in the literature, Chapter Three explores the construct of commitment using novel methodology whilst addressing some of the limitations within sport fandom literature (e.g. an under researched social group in psychology, and the limited understanding of commitment) by using sport fans as the sample. Sport fans were also chosen since being committed has been identified and referred to as an essential component of being a good sport fan of a specific team (Wakelin et al, 2005; Wann et al, 2001; Wann, 2006a) but there is little evidence to substantiate these claims. In addition, sport fans have elaborate strategies to help them maintain their commitment with their team when they perform poorly. One strategy suggested involves fans changing the comparative dimension to the ingroup, and in particular, the in-group’s commitment (Dietz & Murrell, 1992; Wann, 2006a), but this has not been examined to any degree. If sport fans align their perceptions with the in-group’s perceived value, then fans clearly gain some social identity benefit from them. One reason might be that the in-group have more control over their own social identity whereas they cannot control their social identity if they solely rely on their team to define their status. Furthermore, it was proposed Chapters One and Two that if the in-group is a valued comparative dimension that is more relevant to in-group members, then the in-group might be more important for group members’ social identity than the team. To examine this proposal, Chapter Four

90

utilises the in-group’s status in terms of their attributes (e.g. commitment) as a comparative dimension and compared with attributes about their team’s status. Finally, the construct collective self-esteem is an important aspect of one’s social identity, but measuring this construct has proved problematic since the scale has been used inconsistently (Aberson, 1999). As highlighted in section 2.3.2 and 2.4.4, this has included using only parts of the scale, or leaving the scale items unaltered so that they do not reflect a specific social group. Furthermore, where some have altered the scale to suit the context, it has not been sufficiently assessed to ensure that construct validity has remained. This is an especially pertinent issue given the advancement in psychometric procedures in psychology (see DeVellis, 2003), and the importance of collective self-esteem as a fundamental tenet of social identity theory. In Chapter Three, to resolve these limitations, the CSES is adapted to suit the context of sport fans and several quantitative procedures are used to examine the psychometric properties (e.g. construct validity) of this popular scale.

91

Chapter 3 The Adaptation and Development of Two Multifactor Models: The Collective Self-Esteem scale and the Sport Fan Commitment scale

3.1 Introduction

In this chapter, research questions one and two are addressed specifically, and therefore this chapter aims to fulfil two objectives. The first objective is to provide a conceptually accurate and psychometrically robust measure for assessing fans’ commitment. Secondly, a contextually specific version of the Collective Self-Esteem scale (CSES; Luhtanen & Crocker, 1992) will be developed. These objectives are addressed by using both qualitative and quantitative procedures outlined by several authors (Braun & Clarke, 2006; DeVellis, 2003; Luhtanen & Crocker, 1992; and Smith, 2003). To develop a measure of the construct ‘commitment’, a combination of steps were used. In general, a conceptual guide for inventory development by DeVellis (2003) was adopted in conjunction with quantitative procedures suggested by Braun & Clarke (2006), and Smith (2003). This combination approach was adopted particularly for the initial development and item generation to contrast specifically with previous measures of fan commitment. For example, rather than develop an item pool traditionally based on expert opinion and literature reviews as suggested by DeVellis (2003), focus groups were held and a thematic analysis was conducted using procedures suggested by Braun & Clarke (2006), and Smith (2003). This mixed methodology approach to conceptualising a construct and item generation was chosen since it contrasted with scales that had been previously

92

developed (e.g. Cameron, 2004; Dimmock et al, 2005; Mahoney et al, 2000; Wann & Branscombe, 1993). As discussed in Chapter Two, some of these scales were unidimensional and had poor psychometrics. These reasons therefore, warranted a more novel approach to inventory development to assess fans’ commitment. By asking fans what they believe is important in being a committed fan, it is anticipated that more meaningful and conceptually accurate items may be generated. In addition, much of the literature presented in the previous section about commitment suggests there is good reason to understand commitment as a multifactor construct. It is therefore unlikely that commitment to a sport team can be conceptualised purely as an affective component suggested by a social identity perspective (Bergami & Bagozzi, 2000; Ellemers et al, 1999b; Ouwerkerk et al, 1999). It was expected that sport fans, when asked, would produce several themes that are important or make up their commitment. However, since mixed method procedures are used specifically for the development of the commitment scale, they are outlined in more detail in the appropriate section of this chapter. The second objective of this chapter is to validate the construct validity of an adapted version of the CSES (Luhtanen & Crocker, 1992) for specific use with sport fans. This scale was chosen because of its frequent use in evaluating the positivity of our social identities. In particular, the CSES has been used regularly with real world groups (Houston & Andreopoulou, 2003; Twenge & Crocker, 2002; Utsey & Constantine, 2006). Subsequent research has also shown that the CSES is easily adapted to assess self-esteem in members of specific social categories (Crocker, Luhtanen, Blaine, & Broadnax, 1994). However, little consideration has been given to the consequences for adapting the scale’s items. In particular, the scale’s psychometric properties after adaptation have not gone beyond tests of reliability and

93

internal consistency until recently (see Utsey & Constantine, 2006). In terms of scale development and psychometric theory (e.g. Item Response Theory, see Baker & Kim, 2004), adapting a scale’s items may have consequences for the effective interpretation of results (DeVellis, 2003, see also Byrne 2006). This is because adapting items may affect the characteristics of an item and potentially result in misinterpretation by an individual (Baker & Kim, 2004). Therefore, to provide a contextually specific version of the CSES that has undergone psychometric scrutiny, exploratory and confirmatory factor analysis was performed on the adapted CSES using the same procedures as the scale’s original authors for comparative purposes. Since this scale is also multifactor and its construction is well documented, it is therefore able to act as an additional guide to DeVellis’s (2003) framework for developing and testing a new model of sport fans’ commitment. Another reason this scale was chosen was that collective self-esteem is considered to be highly correlated with strength of attachments (Dimmock, Grove, & Eklund, 2005; Long, Spears, & Manstead, 1994; Trail & James, 2001, Wann et al, 2003). The scale then may be used to provide additional criterion validity for the new commitment scale proposed. For these reasons, it is necessary to adapt and present the development of the adapted CSES first in this chapter. This chapter is separated into two sections; section A presenting one study outlining the adaptation of the collective self-esteem, and section B that presents five studies outlining the development and testing of a new scale to assess fans’ commitment with their team.

94

Section A Adapting the Collective Self-esteem scale 3.2. Introduction Given that self-esteem has been linked to fan attachment as outlined in Chapter Two, the Collective Self-Esteem scale (Luhtanen & Crocker, 1992) was used to assess sport fans’ self-esteem in a manner envisaged by social identity theory. This scale is also multidimensional developed to measure self-esteem derived from group membership. This is in line with social identity theory’s conceptualisation for distinguishing between personal and social or collective identity discussed in Chapter One and Two. The scale consists of four subscales that measure different aspects of group members’ self-esteem derived from the groups to which they belong. The subscales include ‘private CSE’ (the extent to which individuals feel positively about their group), the ‘public collective self-esteem’ (the extent to which members believe others out-group members feel about their group), the ‘group membership’ (the extent to which members believe that they are good members of their groups), and finally, ‘importance to identity’ (the extent to which members believe their groups are an important part of their social identity and selfconcept). In further support for distinguishing between personal and collective selfesteem, Long, Spears, & Manstead (1994), and Long & Spears (1998) revealed that those with low personal self-esteem displayed less in-group bias, and those with low CSE showed more in-group bias. These findings underline the importance of distinguishing between personal and social self-esteem in a manner envisaged by social identity theory (Tajfel & Turner, 1979). Therefore, the CSES was developed to

95

compliment the distinction between personal and social identity made by social identity theory (SIT; see also Foels, 2006; Rubin & Hewstone, 1998; Turner, 1999 for a review). To this end, the scale is now thought of as a reliable and robust measure of the collective self-concept and positivity of one’s social identity derived from their group memberships (Hunter et al, 2004). For the reasons stated in the previous section (i.e. item response theory; Baker & Kim, 2004), and since this thesis uses sport fans, a more contextually specific version of the CSES is needed. Despite the scale’s popular use with researchers of social identity, and particularly stigmatized and racial groups (Houston & Andreopoulou, 2003; Twenge & Crocker, 2002; Utsey & Constantine, 2006), members of self-selected groups such as sport fans, may not gain self-esteem from their groups in the same way as members of ascribed or assigned groups. As Ellemers et al (1999b) pointed out, as group members in self-selected groups for example, might not demonstrate as consistent social identity processes (e.g. intergroup discrimination) as those in assigned groups. Subsequent research has also shown that the CSES is easily adapted to assess the self-esteem with specific social categories (Crocker, Luhtanen, Blaine, & Broadnax, 1994). In addition to item response theory, item adaptation may also have consequences in terms of structural equation modelling (e.g. altering the structure of a latent construct). Altering the characteristics of items may potentially alter the construct validity of the scale (Reise, Waller, & Comrey, 2000; Wirth & Edwards, 2007). The ability to replicate a measure’s integrity across contexts and different populations would allow researchers to evaluate whether the measure in question is still a reliable tool in an alternative context (Reise, Waller, & Comrey, 2000). Therefore, changing the context and wording of items without further tests of

96

validity, could allow the potential to disrupt the measure’s ability to assess the purported construct (Baker & Kim 2006; Byrne, 2004). For example, a sport fan may not necessarily extract a similar meaning from items in a global version of a measure than from the same items in a more specific version, and as such, the measure might mean two different things in the same contexts. Researchers then should strive to ensure good research practice when any model is adapted to suit a specific population, but also go beyond minimal psychometric tests. Other procedures such as factor analysis and confirmatory factor analysis have become popular methods of scale development for example (DeVellis, 2003). However, unlike factor analysis, confirmatory factor analysis is used to examine the extent to which a data sample matches an a priori model. Furthermore, confirmatory factor analysis only measures the strength of factor covariation for those items and factors defined by the model (Biddle, Markland, Gilbourne, Chatzisarantis, & Sparkes, 2001). This is done by specifying the desired pattern of constraint on the factor loadings such that items can only load onto to their specified factor (cf. Biddle et al, 2001; Byrne, 2006). As such, it has become the contemporary standard for testing the generalisability and construct validity of measures as well as to test the extent to which data supports a hypothesised relationship between variables. Some however, have argued that when one is dealing with a hypothesised model such as the CSES, that the first test of factorial validity should be a test of confirmatory rather than factor analysis (Schutz, 1994). This is because factor analysis is primarily a theory generating tool (Lane, Harwood, & Nevill, 2005). In this study however, factor analysis is used in conjunction with confirmatory factory analysis for comparative purposes with Luhtanen & Crocker’s (1992) procedure. In addition, DeVellis (2003, p132) also states that although factor analysis is mainly a theory generating procedure

97

for scale development, it provides a more stringent analysis of a scale’s replication. For example, the chance that the data from several different samples on different occasions producing the same factorial results being due to a ‘recurring quirk’ is slim. The argument here in other words, is that repeated factor analysis could provide stronger confirmatory evidence of a strong factor structure for a particular construct. Confirmatory factor analysis in contrast confirms a good model fit, and in addition provides cues as to how the model should turn out. Following a similar procedure to this current study for example, Terry, Lane, Lane, & Keohane (1999) adapted the Profile of Mood States (POMS; McNair, Lorr, & Droppleman, 1971) to use amongst adolescents involved with sport. In addition to providing a more contextually relevant scale, they were able shorten the number of items from 65 items to 24 items whilst retaining construct validity and the scale’s integrity by using both factor analysis and confirmatory factor analysis. In the current study, the 16 item CSES was adapted to fit the context of sport fandom. For each item, words such as ‘member’ were changed to ‘supporter’, and words such as ‘group’ were amended to ‘team’. For example, ‘I am a worthy member of the social groups I belong to’ was amended to ‘I am a worthy supporter of my team’, and amended items can be seen in Table 4. These items were then distributed to fans of Welsh regional rugby before a rugby match. For comparative purposes, data were then analysed in a similar way to that used in the original development of the CSES.

98

Study One 3.2.1 Method Participants Participants were 328 (280 males and 48 females) fans of Welsh regional rugby from five different regions in Wales. 3 Participants were given incentives to complete the surveys such as tickets to see their international team compete in forthcoming competitions. The mean age of participants was 42.2 years old (SD = 17.6, age range 10 – 82 yrs), and they had followed their team for 30.6 years (SD = 15.6; range 2 - 72 yrs).

Procedure The amended CSES was administered at each of the regional teams’ home grounds before a European Cup competition match. These fixtures were chosen due to the importance of these matches to the clubs and their fans and it was anticipated they would have higher attendances. Fans were handed surveys and told the survey was about attitudes toward their rugby team. In return for their time, participants were entered into a draw to win tickets to see their national team play in the forthcoming international games. Participants were also given a freepost envelope to return their responses to the university.

Data Analysis To examine whether the adapted CSES scale retained internal consistency and reliability the scale was examined using Cronbach’s Alpha (Cronbach & Meal, 1955). Following tests of reliability, the nature of the factor structure was then examined

3

The CSES in this study was also used as part of a larger ESRC funded project.

99

using exploratory factor analysis in the same manner described by Luhtanen & Crocker (1992). This is because one assumption underlying alpha is that a set of items are unidimensional, and that each item measures a common underlying variable (DeVellis, 2003). Additionally, a set of items does not necessarily constitute a scale. Therefore, factor analysis allows identification of one or more latent variables underlying an item set rather than one variable assumed by tests of reliability such as alpha. Since adaptation of items may affect the characteristics of each item and therefore its content and construct validity, re-examination of the scale using factor analysis is critical. Once a common factor structure was identified for collective selfesteem, to test the relationship of the latent variables, confirmatory factor analysis was administered on four hypothesised models of collective self-esteem indicated by Luhtanen & Crocker (1992).

3.2.2 Results Mean scores and standard deviations were computed and compared to Luhtanen & Crocker’s initial scores for the CSES and these can be seen in Table 1. Mean scores show that overall, fans of Welsh regional rugby seemed to have similar collective self-esteem scores as Luhtanen & Crocker’s sample. Their sample’s scores provides a basis for comparison in later research chapters. However, one of the subscale’s (public collective self-esteem) mean score were lower for fans of Welsh regional rugby than participants in Luhtanen & Crocker’s original study. Since there was also a large gender difference in the sample, separate scores were calculated for males and females. In this study females had a slightly higher mean score (X = 81.4, SD = 13.9) compared to males (X= 77.5, SD = 16.3). However, the data was tested for homogeneity of variance and Levene’s assumption

100

of equal variance was not violated (F

(1)

1.57, p > 0.05). Therefore, to test between

gender differences further, an independent t-test was employed and no significant difference between males and female collective self-esteem scores was found (t (324) = 1.52, p > 0.05).

Table 1. Mean scores for the collective self-esteem scale

Mean Scores (SD) Me

Pr

Pu

Id

CSES

Adapted CSES

21.7 (4.9)

23.4 (5.1)

17.5 (5.7)

16.2 (6.2)

77.7 (16.3)

Original CSES

22.6 (3.4)

22.8 (3.5)

21.4 (4.0)

18.7 (5.0)

85.6 (11.7)

Key:Membership (Me); Private collective self-esteem (Pr); Public collective self-esteem (Pu); Importance to identity (Id); Collective self-esteem scale (CSES)

Reliability To examine internal consistency and reliability, Cronbach’s alphas were computed. Scores were calculated for each of the four subscales and the total CSES and presented in Table 2. For comparison, Luhtanen & Crocker's alphas are also presented. Analysis indicated the adapted CSES scale was internally consistent with substantial alpha for the total scale of 0.85. This result was similar to that found by Luhtanen & Crocker’s initial study, which can be seen in Table 2. As Cohen (1988, but see also Cronbach & Meahl, 1955) suggests, alpha levels of above 0.80 are considered to show a scale with good reliability. Alphas were also computed for the subscales, and each was substantial. These alphas were very similar to those found in Luhtanen & Crocker's results (0.70 for membership and 0.80 for private CSE). Subscales with an alpha level above 0.60 are considered substantial for subscales with four items (Loewenthal, 2001). Therefore, 101

subscales in the adapted CSES suggest each one has strong reliability and internal consistency. Item total correlations ranged from 0.43 to 0.69 for the total scale compared to a range of 0.49 to 0.62 in Luhtanen & Crocker's results. Items that loaded lower did not increase the overall alpha if they were deleted.

Table 2. Alpha scores for the collective self-esteem scale

Alpha Levels Me

Pr

Pu

Id

CSES

Adapted CSES

0.70

0.80

0.74

0.76

0.85

Original CSES*

0.73

0.74

0.80

0.76

0.85

*Luhtanen & Crocker’s (1992) original scores

To further examine construct validity of the adapted CSES, correlations between the subscales and the total scale were also calculated and are presented in Table 3. Subscales were moderately correlated with each other, showing that they were related but were also tapping different aspects of collective self-esteem as suggested by Luhtanen & Crocker (1992). The highest correlation was between membership and private CSE subscales (r = 0.60, p < 0.01), and the lowest between public CSE and identity subscales (r = 0.15, p < 0.01). This pattern was also found in Luhtanen & Crocker's initial study (group membership and private CSE, r = 0.59, p < 0.001; public CSE and identity, r = 0.23, p < 0.001). Subscale scores were also correlated with the total scale, the highest being private CSE (r = 0.78, p < 0.01), and the lowest being public CSE (r = 0.61, p < 0.01), again mirroring the pattern found by Luhtanen & Crocker (private CSE; r = 0.80, p < 0.001; and public CSE; r = 0.69, p < 0.001). 102

Table 3. Correlations for the adapted collective self-esteem scale

Subscale Correlation

Membership Private CSE

Pr

Pu

Id

CSES

0.60**

0.29**.

0.46**

0.77**

0.39**

0.43**

0.78**

0.15**

0.61**

Public CSE Identity

0.72**

** p < 0.01

Factor analyses Some researchers argue for using confirmatory factor analysis rather than exploratory factor analysis to test a models factor structure when it is a priori (Schutz, 1994), and additionally it better allows the testing of theory driven approaches to inventory design (Biddle, Markland, Gilbourne, Chatzisarantis, & Sparkes, 2001, Lane et al, 2005). Since the CSES is an established multifactor model, confirmatory factor analysis was employed to examine the relationship between the factors. However, as items were adapted and the sample in this study are qualitatively different in contrast to Luhtanen & Crocker’s original study, it was deemed to be good research practice to first ensure that a similar factor structure existed using exploratory factor analysis. Following factor analysis, the items were specified into their appropriate variables (i.e. each subscale) and their relationships tested using confirmatory factor analysis. This procedure is in line with DeVellis’s (2003) order of testing for scale development. For example, after item generation and an appropriate measurement scale (i.e. Likert scale) have been decided, researchers should still confirm factorial structure of the hypothesised model. This should be

103

done by first using exploratory factor analysis before running a confirmatory analysis that then tests the relationship between the hypothesised factorial structures. To examine the four components of the CSE, the data for the 16 items were subjected to a principal components analysis using varimax rotation in same manner as Luhtanen & Crocker. Keeping to the same procedure as Luhtanen and Crocker, the four factors were not specified in the factor analysis. Results of the factor analysis indicated, as with the original CSES, a four-factor structure for the adapted version explaining 63.08% of the variance; loadings can be seen in Table 4. Most of the items loaded on the appropriate factors but some unexpectedly cross-loaded to a slightly higher degree than Luhtanen & Crocker reported. A closer examination of the cross loading shows that the items that cross loaded were in fact from the subscales that most strongly correlated; the ‘private collective self-esteem subscale’ and the ‘membership subscale’ (see Table 3). Both of these scales have been found to be more similar with each other than the other subscales (see Bergammi & Bargozzi, 2002; Rubin & Hewstone, 1998)

Confirmatory factor analysis To examine construct validity further and the relationship between the common factors, tests of confirmatory factor analysis using Amos 7.0 (Arbuckle, 2006) was performed on the data. For comparative purposes each of the hypothesised models originally tested by Luhtanen & Crocker were examined. These models included a one factor model, a four factor uncorrelated model, a four factor correlated model, and four factor hierarchical model; examples of these types of models can be seen in figures 1 - 4.

104

Table 4 Factor loadings 4 for the adapted collective self-esteem scale

Factor Loading a Private CSE I often regret being a supporter of my team a

.40

In general, I’m glad to be a supporter of my team

.75

Overall, I often feel that being a supporter of my team is not worthwhile

a

.34

I feel good about being a supporter of my team

.77

Membership

1

I am a worthy supporter of my team

2

3

4

.78

I feel I don’t have much to offer other fans of my team

a

.59

I am a cooperative supporter of my team I often feel I’m a useless supporter of my team

.35

.77 a

.67

Public CSE Overall, the fans of my team are considered good by others

.85

Most people consider fans of my team to be more

.31

ineffective than fans of other teams a In general, others respect the fans of my team

.87

In general, others think that the fans of my team are unworthy

a

.48

.62

Identity Overall, being a supporter of my team has very little to do with how I feel about myself a Being a supporter of my team is an important reflection of who I am Being a supporter of my team is unimportant to my sense of what kind of person I am a In general, being a supporter of my team is an important part of my self image a.

4

Item was reversed for coding

In line with factor analysis, factors were extracted based on eigenvaluies greater than 1 (Field, 2007),

and for clarity, only factor loadings equal to or greater than .30 are indicated (Tabachnick & Fidell, 2001).

105

.77 .49 .74

.56

e1 e2 e3 e4 e5 e6 e7 e8 e9 e10 e11 e12 e13 e14 e15 e16

GM1 GM2 GM3 GM4 PR1 PR2 PR3 PR4 PU1 PU2 PU3 PU4 ID1 ID2 ID3 ID4

CSES

Figure 1 example of a one factor hierarchical model

e1

GM1

e2

GM2

e3

GM3

e4

GM4

e5

ID1

e6

ID2

e7

ID3

e8

ID4

e9

PU1

e10

PU2

e11

PU3

e12

PU4

e16

PR4

e15 e14

PR3

e13

PR1

F1

F2

F3

F4

PR2

Figure 2 example of four factor uncorrelated model

106

e1

GM1

e2

GM2

e3

GM3

e4

GM4

e5

ID1

e6

ID2

e7

ID3

e8

ID4

e9

PU1

e10

PU2

e11

PU3

e12

PU4

e16

PR4

e15

PR3

e14

PR2

e13

PR1

F1

F2

F3

F4

Figure 3 example of four factor correlated model

e1

GM1

e2

GM2

e3

GM3

e4 e5

GM4 PR1

e6

PR2

e7

PR3

e8

PR4

e9

PU1

e10

PU2

e11

PU3

e12

PU4

e13

ID1

e14

ID2

e15

ID3

e16

ID4

F1

F2 CSES F3

F4

Figure 4 example of a four factor hierarchical model

107

An assumption underlying confirmatory factor analysis is that the data are normally distributed. Therefore, two precautions were taken to minimise the effects of non-normality suggested by Lane, Harwood, Terry, & Karageorghis, (2004); and Tabachnick & Fidell, (2001). As well as using a large sample, it was decided to use the scaled χ2 Santorra-Bentler method as this method is reported to control for overestimation of χ2 and underestimation of fit indices. The use of these precautions was supported by preliminary analyses that indicated that the data did not show multivariate normality. The fit of the four models of collective self-esteem were evaluated using several different criteria (goodness of fit indexes) recommended by Luhtanen & Crocker (1992) and other researchers (Bollen, 1989; Byrne, 2006; Kline, 1998; Lane, et al, 2004, Obst & White, 2005). These included the chi-square goodness of fit, the chi-square / degrees of freedom ratio, the Normed Fit Index (NFI; Bentler & Bonett, 1980), the Robust Comparative Fit Index (RCFI; Bentler, 1995), and the Non-normed Fit or Tucker-Lewis Index (TLI; Tucker & Lewis, 1973). The results for all goodness of fit indexes for this study and Luhtanen & Crocker’s initial study are presented in Table 5. A non-significant Chi-square statistic (χ2) indicates an excellent model fit and suggests that the observed relationships between the variables do not depart from the hypothesised model and the empirical model. This statistic however, is sensitive to sample size (see Byrne, 2006, Tabachnick & Fidell, 2001) and therefore the χ2 / df ratios are also presented. A value of less than 1 indicates an over fit, a value of 1 but less than 2 indicates a good fit, and a value of between 2 and 3 indicates an acceptable fit. Moreover, sometimes a poor fit does not necessarily reflect a poor model, and according to Bollen (1989), a poor model to data fit indicated by Chi-

108

square is not uncommon as sample size increases. This is one reason why other indices (CFI & NFI) must be used in conjunction with Chi-square goodness of fit since they are less sensitive to sample size (Bentler, 1990). Other indices used in this study examine the hypothesised models in relation to how much better the model fits the data in relation to where no relationships exist at all in the data. These indices however, differ in the way they examine the data fit (Bentler, 1990, 1995; Byrne, 2006). Moreover, a good or poor fit of a model is likely to produce consistent results across these indices (Tabchnick & Fidell, 2001), and therefore provides another reason for using several criteria to confirm model fit. Values ranging from 0 to 1 for these indices estimate the degree to which a hypothesised model fits the data in comparison to a null model. Consequently, the higher the index values, the better the fit. Hypothesised models that produce fit indices of less than 0.90 could be substantially improved, and ideally values should approach 0.95 to demonstrate a model with good model to data fit (Arbuckle, 2006; Byrne, 2006). In addition to these fit indices, MacCallum & Austin (2000) strongly recommend the root mean square error of approximation (RMSEA; Steiger, 1990) be used. This is because the RMSEA is sensitive to model misspecification and yield appropriate conclusions of model quality (MacCallum & Austin, 2000). For the RMSEA however, values of less than .10 indicate a moderate to good fit and generally the lower the score the better the hypothesised model will fit the data (Hu & Bentler, 1999; MaCallum & Austin, 2000; Steiger, 1990) Results of the CFA showed that the χ2 values were significant for all four models. This indicates that there was a significant difference between the observed and specified models, and can be seen in Table 5. This suggests that there is a less

109

than adequate fit to the data for all of the models in the adapted CSES, mirroring Luhtanen & Crocker's results. As mentioned above this is not uncommon with larger samples and therefore we examine other indices to provide better estimates of model fit. Values obtained from the CFI, NFI & TLI are also reported in Table 5, and all four models demonstrated less than adequate fits to the data. Values for each model loaded substantially under .95 on all indices, again mirroring Luhtanen & Crocker’s analysis. In addition to these values, the RMSEA values indicated that each model was a less than adequate fit. Of the four models however, the four factor correlated model yielded the highest values and had best model to data fit in comparison to the others. This was not the case for Luhtanen and Crocker’s original study where the one factor hierarchical model had the best fit.

Table 5. Goodness of fit indices for the collective self-esteem scale χ2

df

χ2 / df

RCFI

NFI

TLI

RMSEA

One factor

1614.09

104

15.52

.68

.67

.63

-

Four factor uncorrelated

1309.71

104

12.59

.75

.73

.71

-

Four factor correlated

425.67

98

4.34

.93

.91

.92

Hierarchical model

441.82

103

4.55

.93

.91

.92

-

One factor

785.72

104

7.55

.64

.61

.58

.14

Hierarchical

785.72

104

7.55

.64

.61

.58

.14

Four factor uncorrelated

730.37

108

6.76

.67

.64

.63

.13

Four Factor correlated

354.31

102

3.47

.87

.82

.84

.09

Model Original CSES*

Adapted CSES

*Luhatnen & Crocker’s (1992) original indices using confirmatory factor analyis Note (1) Models are ordered from poor model to data fit (higher figures) to best model to data fit (lower figures); (2) CFI = comparative fit index; TLI = Tucker-Lewis Index; RMSEA = root mean square error of approximation (3)All chi square values (χ2) are significant at p < .001 for the CSES and adapted CSES

110

3.2.3 Discussion This study aimed to validate the construct validity of an amended version of Luhtanen & Crocker’s (1992) Collective Self-Esteem scale for the specific use with sport fans. To do this, exploratory factor analysis was used to examine whether the amended version replicated the same four factor structure as the original scale, and confirmatory factor analysis was used to examine the structure or best fit of the variables. In the current study, the population were a self-selected group. This is in contrast to the original development where the population were members of an ascribed group (ethnicity). In a comparison of descriptive statistics with Luhtanen & Crocker’s study (mean CSES score, X = 86), fans of Welsh rugby were of a similar level in overall collective self-esteem with a mean score of 78 from a possible score of 112. It was interpreted that generally fans of Welsh rugby were as high in collective self-esteem as members of ascribed groups in Luhtanen & Crocker’s study. This suggests that both groups hold their membership in these groups of particular importance to their collective self-concept, and evaluated their collective identities as generally positive. The scale in general was found to be as reliable as Luhtanen & Crocker’s initial global scale showing the same alpha level (0.85) as the original scale. This shows that generally the scale was easily adapted without affecting its internal consistency and supports previous research advocating the scale’s adaptability (Crocker, Luhtanen, Blaine, & Broadnax, 1994; Utsey & Constantine, 2006; Verkutyen, 1997). Correlations between the subscales and the overall scale confirmed that each subscale, despite adaptation, is likely to still be measuring the overall construct of collective self-esteem. As such, it appears from the reliability and internal consistency examinations that the scale is measuring the same construct as

111

the global version of the CSES. Altering the scale, therefore, did not detract from its reliability of the scale as a measuring tool for examining specific groups’ collective self-esteem. However, to examine the scale’s construct validity further, factor analysis was performed to examine whether adapting the items altered the four factor structure demonstrated in the original scale’s developments. In keeping to the original procedure of factor analysis by Luhtanen & Crocker, it was decided to not specify the four factors so that the items were free to load. If items loaded into their correct factor then this would be evidence of the construct validity. The current study’s data replicated the four factor structure found in the original scale suggesting that the amended scale retained its construct validity. This is despite the qualitative differences between these types of groups. This suggests that sport fans, and in particular fans of Welsh rugby, gained self-esteem from their group membership in a similar way as members of ascribed groups. Furthermore, this may be initial evidence that collective self-esteem might be stable across a variety of different groups (i.e. ascribed, assigned, and self-selected groups). In particular, the sample size in the current study was under half of that used in the original development of the scale suggesting that despite the smaller sample size, the four factor structure was stable (see Byrne, 2006; DeVellis, 2003 for a review on sample size and scale development). Similar findings were also reported in a recent study by Utsey & Constantine (2006) that also used ascribed groups and a similar sample size to examine the CSES’s psychometric properties. A closer examination of the subscales showed that two items cross loaded. These items belonged to the membership subscale, and both cross-loaded onto the private collective self-esteem subscale. This is unsurprising given that these two

112

subscales were strongly correlated in the current study, and in Luhtanen & Crocker’s original study. In terms of the social identity literature, the cross loadings and strong correlations may not be that unexpected since there is disagreement about the nature for these two subscales. For example, previous chapters noted the scale’s inconsistent use in the literature for examining collective self-esteem. Some suggest that only one subscale was needed to measure collective self-esteem, whilst others believe all four subscales were required. For example, Aberson (1999) and Bergammi & Bagozzi (2002) used the membership subscale to measure group members’ collective selfesteem, whilst Rubin & Hewstone (1998) reported that it was the private collective self-esteem subscale that represented group members’ collective self-esteem. Since both subscales have been suggested to be measuring the same thing (i.e. collective self-esteem), this further supports the strong correlation between these subscales (Table 3) and therefore the cross loadings that were observed in the factor analysis (Table 4). Whichever is the case however, is an issue for future research since the whole scale is used in this thesis. In terms of confirmatory factor analysis, four hypothesised models were examined. In the original scale’s development, a one factor hierarchical model was found to explain the data fit best, but this was only of a moderate level. Similarly, Utsey & Constantine (2006) found a moderate fit for the models tested, but their study found that the one factor hierarchical model produced the best fit. In contrast, the current study showed that a four factor correlated model produced the best fit of the models tested, but similarly this was only of a moderate level. This is unsurprising considering the correlations observed between each subscale. This further indicates that each subscale measures some aspect of collective self-esteem, and therefore each subscale forms one aspect of one underlying construct i.e. collective self-esteem.

113

Although confirmatory factor analysis did not prevail a superior four factor structure to explain the relationship between the variables, Luhtanen & Crocker fail to expand on what this lack of a clear factor structure means for the construct. However, Luhtanen & Crocker do add in passing that most measures cannot be adequately subsumed by one underlying construct [pg 327]. In a similar vein, Utsey & Constantine (2006) concluded that collective self-esteem may be more complicated than previously thought [p178]. Another reason may be that since there is strong evidence of a four factor structure in the CSES, even after items have been amended, this may weaken the model’s ability to sustain a common factor structure. For example, as discussed in previous chapters, some groups differ in their characteristics. This may mean that each group could gain self-esteem from their membership in a specific group in different ways. Therefore, it is unlikely that a common factor structure and relationship can exist for all groups. Fans of Welsh regional rugby may differ in their collective self-esteem to other self-selected groups, strengthening the case for even more specific versions of the CSES. However, in this study, for the most part, fans of Welsh regional rugby seemed to gain as much selfesteem from their group membership as those in Luhtanen & Crocker’s original sample. In comparing the current study with the original study, it is clear that despite a difference in the sample size and type of group (self-selected), construct validity was upheld and suggests that collective self-esteem may be a relatively stable construct. In the original study, the sample was predominately white and consisted of university students taking introductory psychology lessons. Although Luhtanen & Crocker did not specify the age of their students, typically this sample is likely to be between the ages of 18 and 21. In the current study, a large age range (10 to 87 years old) was

114

observed. Considering self-esteem has been reported to change and develop throughout our lifespan (Crocker, Robins, Trzesniewski, Gosling, & Potter, 2001), the adapted scale’s psychometric properties remained very similar to the original findings. This finding provides some initial evidence that collective self-esteem might remain relatively stable throughout our lives, at least in self-selected groups. In the current study, there was a large difference between the numbers of male and female participants. Luhtanen and Crocker (1992) however, had a more balanced sample of male and females in their original study, but they did not test for gender differences. Despite this current study’s sampling bias, results showed that there was no significant difference in collective self-esteem scores between males and females, and therefore fans of Welsh rugby. Based on the current study, collective self-esteem appears to be a relatively constant four factor construct, regardless of gender. This is interesting because in sociology men are more likely to be sport fans (Gantz & Wenner, 1991). This is because of the way that males experience socialisation in contrast to females, and therefore being a fan is likely to be important to a male’s identity in general. However, in terms of becoming a fan, this is not the case in psychology. Despite evidence for a clear gender difference in attendance patterns (Dietz-Uhler & Murrell, 1999; Wann et al, 2001), differences between male and female fans are considered perceptual (cf. James & Ridinger, 2002). Rather, it appears that highly identified fans (both male and female) are likely to attend more games, gain self-esteem from their team, and be more emotionally attached to their team (Shank & Beasley, 1998; Wann et al, 2001). Therefore, despite the sampling issue in the current study, and although men and women may think of their relationship to their teams differently, being highly identified might be a more important factor in determining one’s amount of self-esteem gained from supporting a

115

team. Distinguishing actual gender differences is therefore, one area that could be investigated in future research since there does not appear to be any research that can corroborate any real gender differences in their collective self-esteem. As discussed previously, one reason for this is because the majority of research using the CSES has utilised ethnicity as a grouping variable (see Utsey & Constantine, 2006 for a review). Although initially there seems to be little difference between the amended and original versions of the CSES, there may be some additional methodological implications that may have affected the amended version. In the original study, participants responded immediately whereas in the current study fans were able to respond by freepost. Consequently, fans’ responses may have been affected by a time lag. For example, if fans responded after the game then identity salience may have been less as they were no longer at the game. Social identity posits that our identities change rapidly in response to our social contexts (Ellemers et al, 2002; Hogg et al, 2004; Turner et al, 1999). Therefore, if fans were not operating in their social identity (as a fan of a specific team), they may have been operating at a more personal level of their identity. Responding to items about a social identity that is not particularly salient may have affected their responses. In terms of affecting the CSES, lowered identity salience should have therefore produced a lower amount of collective selfesteem (Luhtanen & Crocker, 1992; Hogg et al, 2004). This assumption might be partially supported in the ‘important to identity’ subscale as scores were much lower in contrast to other subscales. However, it is unlikely that time lag or lowered identity salience did have a significant impact since both Luhtanen & Crocker (1992) and Utsey & Constantine (2006) observed lowered ‘importance to identity’ subscale scores in their studies. Therefore, it is unlikely to have had an impact on the factor structure of the model in this study. Additionally, if some fans responded after a

116

game, then potentially their team’s performance may have affected their responses to the items. For example, if their team had won, fans self-esteem may have been boosted and therefore their collective self-esteem in general is much lower than reported. Alternatively, a fans’ team may have lost, therefore fans' collective selfesteem may have been lowered. As such, their self-esteem may actually be higher than reported in this study. Since the outcome of the games used in the current study was not monitored, the extent of the effect of performance on their self-esteem is not known. However, since the results observed were similar to those obtained by Luhtanen & Crocker (1992) and Utsey & Constantine (2006), it is likely that this was not the case in the current study. Finally, some consideration about the sampling method in terms of fans’ positioning in each ground should be acknowledged. Unlike football, rugby in general enjoys unsegregated crowds where both home and away fans can interact. Care was given to avoid overpopulating the season and family ticket areas with the surveys in order to avoid a biased responses from a specific demographic within the fan population. However, it is likely that those that respond to surveys, much like those that attended the focus groups, are likely to be high in commitment or motivated by incentives, and as such this might affect interpretation of how committed Welsh rugby fans are. In summary, the CSES was found to be a valid and reliable scale that has strong adaptability. Similarly, the construct validity in terms of factor structure remained stable, but confirmatory factor analysis suggested that the nature and reliability of the four factors remains unclear, a finding also confirmed by Utsey & Constantine (2006). Although this does not mean that the factors do not measure the construct collective self-esteem, it does suggest that further studies are needed to

117

investigate the nature of collective self-esteem and the relationships between the factors more thoroughly.

118

Section B Developing a commitment scale: A factorial model of commitment to supporting a sport team 5 3.3. Introduction Given that there is lack of adequate scales to assess fans’ commitment, the main objective of this section is to develop a psychometrically robust measuring tool for assessing fans’ commitment. In addition, the conceptualisation of commitment from a social identity perspective is currently understood as unidimensional, this is despite research in other disciplines suggesting commitment is best understood as a multidimensional construct (Allen & Meyer, 1990; Cooper-Hakim & Viswesvaran, 2005; Delobbe & Vandenberghe, 2000; Pritchard, Howard, & Havitz, 1999; Van Dick et al, 2004; Wann & Pierce, 2003). In order to understand commitment further and to contrast with previous approaches to scale development of commitment (e.g. Cameron, 2003; Wann & Branscombe, 1993), a mixed methodology approach was chosen. Focus groups were conducted to investigate what sport fans thought and felt about their commitment to a specific team. It was hoped that fans would provide a more accurate and meaningful insight into how their commitment can be conceptualised. Items were then generated from the focus groups and subsequently measured quantitatively. This was conducted in a similar manner employed in Study One using exploratory and confirmatory factor analysis. This combined approach incorporated guidelines provided by several authors (e.g. Burke & Clarke, 2006; DeVellis, 2003), and these are explained in detail in the appropriate studies. 5

The studies used in the development and validation of the SFCS were conducted within a larger ESRC funded project. Parts of this section have been presented at a number of conferences and published (e.g. Davies, Mayer, Shearer, Hall, Thomson, & Hall, 2008).

119

Commitment from a social identity perspective is explained as an affective dimension (affective commitment) and an indicator of one’s strength of identification with a group (e.g. Brown et al, 1986; Cameron, 2004; Ellemers et al, 1999b; Dimmock, Grove & Eklund, 2005; Obst & White, 2005). However, for many fans their commitment may include other behavioural elements of their involvement (e.g. effort in attending, buying team merchandise). Many of these behaviours may be seen by fans as demonstrating commitment, but may not necessarily mean that they are strongly attached to their team. As discussed in Chapter Two for example, fans may simply wish to be seen by other group members as committed to increase their acceptance by other members (cf. Ellemers et al, 1999a; 2002). Lastly, behavioural elements of our social identity and involvement with our groups has also been suggested previously in group identity and sport fan literature (Ashmore et al, 2004; Donovan et al, 2005; Jackson & Smith, 1999; Jackson, 2002; Kim et al, 1997; Kwon & Armstrong, 2004; Meyer et al, 2006; Moutinho, Dionisio, & Leal, 2006). Therefore, it is unlikely that commitment is purely affective. With commitment being a strong predictor of fan behaviour, and despite a social identity perspective largely relying on affective components to explain commitment, more appropriate measures of this construct are needed to advance the understanding of commitment in sport fans (Dimmock, Grove, & Eklund, 2005; Wann & Pierce, 2003). Two competing, but limited unidimensional measures of fans’ commitment: the Sport Spectator Identification scale (SSIS; Wann & Branscombe, 1995) and the Psychological Commitment to Team scale (PCT, Mahony, Madrigal, & Howard, 2000) have therefore also been questioned (cf. Kwon & Trail, 2004; Wann & Pierce, 2003). Specifically, both measures initially appear to offer internal reliability and construct validity (Mahoney et al, 2000; Theodorakis et al, 2006;

120

Wann & Pierce, 2003; Wann & Polk, 2007). The PCT, however has been found to perform poorly after further tests of construct validity using factor and confirmatory factor analysis, and was also found to lack good criterion validity (Kwon & Trail, 2003). Therefore, in terms of inventory development it is unlikely to be measuring the intended construct of fan commitment (cf. Biddle et al, 2001; Byrne, 2006; DeVellis, 2003, but see also Kwon & Armstrong, 2004). Secondly, both scales have been found to be empirically related (Wann & Pierce, 2003), with one based on principles grounded in social identity theory (the SSIS, see Wann, 2006b) and the other, (the PCT, see Mahoney et al, 2000) based on consumer psychology. Indeed, Wann & Pierce (2003) compared both scales and found that they were highly correlated, suggesting they were measuring a similar construct, although the SSIS appeared to be a better predictor of overall ‘fandom’. If they are both empirically related and measure a similar construct , then given that the PCT has been found to be a poor measure of the intended construct, should logically also undermine the SSIS’s psychometric properties. Schutz, (1994) pointed out that there are many scales in existence that provide evidence for their validity, but only for another to show its limitations. According to Schutz, one reason for this is that researchers have not used or followed any sort of criteria for inventory development such as those suggested by DeVellis (2003), and employed in this thesis. Only one study to date has attempted to develop a sport-specific commitment scale that has content, factorial, and construct validity (See Dimmock et al, 2005). This attempt however, was based on a social identity conceptualisation of commitment and therefore limiting commitment to a unidimensional construct. Since there is strong evidence that commitment is multidimensional from organisational psychology literature, another problem with the SSIS and PCT is that

121

they are both unidimensional. Although some items on both scales appeared to measure each form of commitment proposed by Allen & Meyer’s (1990) model, Wann & Pierce (2003) suggested that unidimensional scales such as the SSIS and PCT limited the full scope of measuring fans’ commitment. Thus, the aim of the following series of studies is to develop a measure of sport fans’ commitment to their team, and to extend previous knowledge and understanding of the nature of the construct ‘commitment’. As a starting point, focus groups were conducted in Study Two using fans of Welsh regional rugby. Thereafter, a qualitative analysis was used to examine how sport fans understood their commitment.

Study Two 3.3.1 Method To help develop a measure of commitment, and to address research question one, a qualitative methodology was employed for the current study and involved several qualitative techniques that included focus groups and open coded thematic analysis. This was performed in a manner similar to that of several authors’ (e.g. Braun & Clarke, 2006; Wilkinson, 2003) recommendations. Focus groups were chosen as the main method of collecting data from participants as it allows a large amount of data to be collected in a relatively short time in contrast to conducting several single participant interviews. In addition, focus groups allow interaction between group members that are more naturalistic and involve a wide range of communicative processes such as jokes, storytelling, boasting, and disagreement (Wilkinson, 2003).

122

In the current study, focus groups were held at two Welsh Premier rugby clubs.

6

The clubs sent out general invitations asking fans if they were interested in

taking part. As well as hosting the focus groups, the clubs offered incentives for the fans, such as complimentary tickets, stickers and merchandise to participating fans.

Participants In total, 45 fans (29 male and 16 female, aged between 12 and 62) attended one of seven groups. These included 14 fans who held an adult season ticket, 12 adults who held family season tickets, 13 children who held season tickets, and six occasional attendees.

Some of the questions asked included “what makes a

committed a supporter?”, and “why is committing to your team important?” The focus groups were recorded using video equipment and digital voice recorders so that they could be transcribed later.

Data analysis To help organise the large volume of data and understand what commitment meant to fans, transcriptions of the discussions were subjected to open coded thematic analysis. Open coded thematic analysis allowed the generation of codes derived from the participants’ responses rather than any pre-conceived codes, themes, and theory. DeVellis (2003) suggests the second step of inventory development is to generate items based on the literature or expert opinion. Since the literature is less than adequate and expert opinion across multiple disciplines conflicting in terms of defining the construct, it was deemed that a more novel and insightful approach to

6

A strong acknowledgement and thanks to Cardiff rugby football club, Newport rugby football club, and Llanelli rugby football club for their assistance in conducting and hosting the ESRC project on their premises, of which this researchwas also part.

123

inventory development was necessary by asking fans what they thought and felt about their commitment. Firstly, thematic analysis was conducted using a similar procedure to one recommended by Braun & Clarke (2006) for identifying and organising themes. These procedures included six phases of thematic analysis. Phase one involved familiarising oneself with the data. This described the process of transcribing the data, reading, and re-reading the data, noting down initial ideas. Phase two involved generating initial codes. This phase included coding interesting features of the data in a systematic fashion across the entire data set, collating data relevant to each code. Phase three involved searching for the themes. This phase involved collating codes into potential themes, gathering all data relevant to each potential theme. Phase four involved reviewing the themes. For instance, this included checking if the themes work in relation to the coded extracts and the entire data set. Phase five included defining and naming themes. This included an ongoing analysis to refine the specifics of each theme, generating clear names for each theme. Phase six involved producing the report. This phase was described as the final opportunity for analysis that included selecting appropriate extracts, providing a thematic ‘map’, and discussion of the analysis. In addition to Braun & Clarke’s procedures, after thematic analysis was conducted, transcripts were then re-read and subjected to independent cross validation by five researchers. 7

3.3.2 Results Phases two to five revealed four recurring superordinate themes each composed of two subthemes. These are shown as a thematic map in Figure 5. The 7

As part of this research was part of a larger ESRC funded project, researchers refers to the other grant holders, and a strong thanks and acknowledgement to them for peer reviewing the transcripts

124

first of the superordinate themes was labelled affective loyalty and consisted of two subthemes: ‘resistance to change’ and ‘emotional attachments’. The second theme was labelled involvement and consisted of two sub-themes labelled ‘camaraderie’ and ‘effort’. These subthemes overlapped most since being a worthy fan came about through putting in effort, whilst effort to get involved allowed transition to a ‘worthy’ status. The third theme was labelled BIRGing and consisted of two subthemes labelled ‘distinctiveness’ and ‘acceptance’. Finally, the fourth theme was labelled self-affirmation and contained two subthemes labelled ‘knowledge’ and ‘selfverification’.

3.3.3 Discussion The main aim of this study was to explore how fans thought and felt about their commitment to their team. This was carried out because commitment as conceptualised from a social identity perspective has been limited to an affective nature, and therefore unidimensional. Similarly, scales (e.g. SSIS and PCT) that have attempted

to

assess

fans’

commitments

are

limited

to

unidimensional

conceptualisations. This is despite strong support for commitment being best understood as multidimensional (e.g. Allen & Meyer, 1990; Delobbe & Vandenberghe, 2000; Pritchard et al, 1999). Results of a thematic analysis from several focus groups using fans of Welsh regional rugby showed promising evidence that fans too, understood their commitment to possess several dimensions.

125

Figure 5. Thematic map of sport fans’ commitment Commitment

Affective Loyalty

1.Resistance to

2.Emotional

change

Attachment

1a. There is only our side so every decision that goes against us is a wrong decision 1b. no matter what happens I’ll keep coming, no questions asked 2. What would stop you guys coming? Nothing…. Death? You’d have to scatter my ashes on the pitch afterwards. I’d still be here

Involvement

1.Effort

BIRGing

2.Camaraderie

1. One supporter comes from Nottingham every week (to Llanelli), he has a season ticket. He’s truly committed to the team and makes the effort. 2. they’re turning up to the home ground but, you listen, they’re not making effort to sing unless they’re winning. They go silent. They stopped going to away games as well. We don’t let up, you know

126

1.Distinctiveness

2.Acceptance

1. You have to wear the shirt, you want people to know who you’re supporting, even if you lost, it’s the rivalry and distinguishing ourselves.. I never want to be confused as a supporter of another team 2. I want to be a good supporter so I will buy their stuff.. y’know shirts and stuff. I want the other fans to see I am a good fan, especially when we win I go crazy, put pictures and that on my PC too.

Self-affirmation

1. knowledge

2.selfverification

1. To know the history of my team and town gives me a lot of pride – I know a lot about the team but the rugby can get a bit above me… 2. If we don’t do well this season, these people won’t be there next season. Same if star players leave. I’ll be here obviously but not everybody will.

Theme 1: Affective loyalty Of all the themes that emerged in the current study, this was the one which paralleled social identity’s conceptualisation of commitment. Fans in this study generally indicated that part of their commitment involved an overwhelming need to put their team as a priority in their life, regardless of other commitments or life events. This was expressed in two subthemes. Fans shared stories expressing their strong emotional attachment and their ability to resist changing their behaviour, to the extent that they arranged their life and thoughts around the team. Almost all fans had at least one story where they could easily recall an event that demonstrated their loyalty to their team. For example, one prominent story that illustrates the sub-theme ‘emotional attachment’ involved one man boasting about arranging his engagement and wedding around his team’s match. The fan recalls:

“Let me tell you, I missed my engagement party for the Swansea game and I was late for my wedding because of the Wales - Scotland game.”

Another fan explained that the team meant everything to him and that he could never stop coming. The fan, when told by another fan that death would stop him from attending games, replied by explaining how he would remain a fan of his team past his death. “you’d have to scatter my ashes on the pitch afterwards… I’d still be here”.

In terms of the subtheme ‘resistance to change’, many fans shared awareness that anything that questioned them or their team was incorrect, even when their team were clearly underperforming. This type of attitude clearly demonstrates a form of ingroup favouritism and cohesion that has been well documented in both social identity and sport fan literature (e.g. Abrams & Hogg, 1999; Branscombe & Wann, 1995; 127

Ellemers et al, 1997; Wann & Grieve, 2005). For example, a typical, and to a certain extent self-effacing statement, would describe their fans’ ‘one eyedness’. This is illustrated in the following statements:

“There is only our side, so every decision that goes against us is a wrong decision”.

In addition, another fan boasted that such ‘one eyedness’ was arrogant but they took much pleasure in this knowledge.

“We’re arrogant, conceited…we know our team is the greatest of all time we take that away with us because we only see one side”.

Another fan added that part of being a good supporter was still attending, despite a losing streak. This involved resisting the temptation to stay away because the team are underperforming.

“I remember when we had to travel away when they were bottom and they were struggling and they weren’t going to go anywhere… we still went no matter what”.

In general, fans took pleasure in sharing such apparent irrational and extreme behaviours, and this may have been in an attempt to gain additional praise from fellow supporters. In addition, many of these statements were preceded by discussing rival teams and their fans, and they generally shared a dislike of rival teams and their supporters. Therefore, to a certain extent, these types of stories helped distinguish between good and bad fans often termed in the literature as ‘fair-weather’ and ‘diehard’ fans (Wann & Branscombe, 1993). Although fans often interchanged loyalty and commitment, literature in other disciplines such as consumer and organisational psychology often describe loyalty as

128

pertaining to be a resistance to change attitude that can also be displayed behaviourally (Neale & Funk, 2006). Others however, have suggested that loyalty makes an important contribution to overall commitment (Funk & James, 2004; Funk, Ridinger, & Moorman, 2003). For the above reasons affective loyalty was chosen for this theme. However, despite the semantic overlap between commitment and loyalty, this theme is clearly similar in nature to affective commitment in the social identity literature. As discussed earlier in Chapters One and Two, group members that have strong emotional attachments are often those members that will remain committed, even when their group is questioned (Dimmock, Grove, & Eklund, 2005; Ellemers et al, 1999b; Ellemers, Spears and Doosje, 2002; Ellemers, Spears, Doosje, 1997; Ouwerkerk et al, 1999).

Theme 2: Involvement In line with other group identity research (Ashmore et al, 2004; Jackson, 2002, Park & Kim, 2000), a theme of involvement emerged. In this theme, fans generally believed that the level of involvement that one had with the team contributed to being a committed fan. This theme was particularly dominated by discussion over fans’ attendance at away games, although to a lesser extent, there was some discussion about getting involved with camaraderie at games. To fans of Welsh regional rugby, involvement was a very important part of evaluating and demonstrating ‘true’ commitment and their worthiness as a fan. This seemed to be a more practical way for fans to distinguish between fair-weather and die-hard fans, although fans generally had little way of knowing who these actually were. Despite this, fans generally regarded their fellow fans as more committed than rival fans, demonstrating further in-group favouritism. In addition, fans used their attendance as

129

a way of distinguishing themselves further from fellow fans. For example, a typical statement would involve explaining how ‘true’ fans go to away games and attend after a losing streak in comparison to fair-weather fans.

“If we (Newport RFC) don’t do well this season, these people won’t be there next season. Same if star players leave. I’ll be here obviously but not everybody will”.

Two subthemes emerged, termed ‘effort’ and ‘camaraderie’. Fans recalled situations where fellow fans made considerable effort to attend games despite clearly being inconvenienced. Although it could be argued that this represented a strong emotional attachment, fans used such examples to describe the concept of a good fan. Therefore, it appears that being seen to make the effort is clearly an important dimension of being committed. To illustrate this, one fan describes the attendance of one man who drove 400 miles return trip on a weekly basis to watch his team.

“One supporter comes from Nottingham every week (to Llanelli), he has a season ticket. He’s truly committed to the team and makes the effort”.

In another instance, a fan reported that attending was an essential part of commitment and that even though he would prefer his social network to attend with him, he would turn up on his own if not.

“It’s only 4 or 5 years that we’ve been part of the group. I’ve been 25 years without them…I’d still go watch the games without them… I’m die-hard”.

Fans from other teams also expressed such effort as an important part of being an ideological fan. For example, one fan expressed that ‘true’ fans can be seen at away games.

130

“We make every effort we can to go and watch our teams; especially away…we know everyone who goes and those who don’t go to away games”

Perhaps one fan summed up this subtheme simply. When asked what made a good supporter they replied with the following answer.

“One that turns up every game and makes an effort to turn up”.

The other subtheme that emerged was labelled camaraderie. This theme was not as strongly present as ‘effort’, but some fans felt that part of being committed included being involved with the camaraderie. For example, behaviours such as singing and booing the opponents were considered normal, and reflected one’s commitment. This was particularly illustrated by two fans from Llanelli. These fans took considerable pride in the effort they put into getting other fans to join in with the singing as well as distracting their team’s opponents. One fan explained how although rival fans turn up to away games (a committed behaviour), Llanelli fans are distinguished further because they continue to sing and attend even at away games.

“they’re (Swansea fans)turning up to the home ground but you listen, they’re not making the effort to sing unless they’re winning. They go silent. They stopped going to away games as well. We don’t let up, you know”.

Further, another fan took pleasure in explaining how fans of English rugby respected their commitment and model fan behaviour when they travelled away. Although the following account reflects Llanelli fans’ attendance behaviour, it also shows a strong cohesive relationship Llanelli fans share that has earned them respect. In his account, he states:

“We went to the shed (Gloucester rugby club) and we gained respect because it was on the Internet, and they were saying the Llanelli fans were the best sort of 131

fans they’ve ever had. It said we travel large and that we have the most committed fans of any Welsh club” For rugby fans then, being strongly involved in ‘fan like’ behaviour, such as attendance is a good indication of other fans’ commitment. In sport fan literature for example, Wann et al (2001) proposed that those that are highly identified are more likely to attend games. However, since lowly identified fans are able to attend, then using attendance and stereotypical behaviour may not be as practical as fans perceive. Although this theme emerged separately, it is likely that it would be strongly correlated with the affective loyalty theme. For example, Hogg et al (2004) suggested that highly identified members are more likely to act in terms of their group membership. Since affective loyalty in the current study overlaps with social identity’s component of affective commitment, then similarly those with strong affective loyalties may have strong involvement. However, in contradiction, Allen and Meyer’s (1990) conceptualisation of commitment suggests that fans may commit in different ways. Since this theme is strongly behavioural in nature, fans high in normative commitment may also be likely to be high in involvement, independent of one’s strength of emotional attachments. Attending for some fans, according to Allen and Meyer, is part of their obligation rather than their strong emotional attachment. This theme therefore, seems to be an additional component of commitment that is not previously considered by a social identity perspective. Additionally, there is some support for such a behavioural component of social identity and commitment. For example, Kwon & Armstrong (2003) distinguished between personal commitment (similar to affective commitment in social identity theory, see Ellemers et al, 1999b) and behavioural commitment (that involved members feeling compelled to carry out duties as a supporter) in sport fans’ attachments. Similarly, Jackson (2002), and more recently Ashmore et al (2004), 132

posit that social identities can be assessed behaviourally as well as using cognitive and affective indicators proposed by recent models of social identity (e.g. Cameron, 2003; Dimmock et al, 2005). In Ashmore et al’s model, it was suggested that social identity includes a ‘behavioural involvement’ component, which entailed group members participating in their group’s activities as an expression of their identity and commitment. Interestingly, one of the definitions also used by Ashmore and colleagues (p93) to describe behavioural involvement indicated that the wearing of apparel was also a good indicator of behavioural involvement. Although wearing apparel is undoubtedly behavioural, in this analysis, this behaviour became a separate theme as fans clearly distinguished between involvement via attendance and using their apparel. This is because many fans can wear apparel without attending or being actively involved. In addition, none of the previous measures that assess sport fans’ identification and commitment considers attendance as a form of commitment, even though these have been administered partly to predict and understand fans’ attendance.

Theme 3: BIRGing In some research examining sport fan behaviour (e.g. Cialdini, 1976; End, 2001), there has been a strong link between strength of attachment to team and the wearing of apparel. Similarly, wearing apparel has also been indicated in some items about sport fan motivation (see Funk and James, 2004; Kwon & Armstrong, 2003; Trail & Fink, 2002). This behaviour was discussed in Chapter Two and is known as Basking in Reflected Glory (BIRGing) (e.g. Cialdini et al, 1976; End, 2001). This theme was then named BIRGing to represent the behaviour of presenting oneself as a

133

fan of a specific team. For fans, wearing a team shirt appeared to function as one way to enhance the salience of the identity in question in an effort to show their distinctiveness from other fans and gain acceptance quickly into a group. In addition, more generic measures of group members’ attachment that utilise a social identity perspective do not incorporate many items that assess the group members’ behaviours which aid their acceptance and distinction as a member of a specific group (e.g. Cameron, 2003; Dimmock, Grove, and Eklund, 2005; Wann & Branscombe, 1993). Therefore, the ability to assess members’ identification or categorisation behaviours in social identity measures is also limited. In this theme, fans described the use of apparel to associate themselves with a specific team, as well as to identify other fans and gain acceptance. This was expressed in two sub themes, ‘distinctiveness’ and ‘acceptance’. In the other themes, fans wanted to express their commitment through accounts that proved their worth as a fan, by attending or putting the team first over other commitments. In this theme, fans used apparel to gain acceptance quickly without having to prove themselves through stories to demonstrate their commitment. In addition, this quickly distinguished in and out-groups. For example, one fan after describing the importance of wearing apparel ended his sentence with the following:

“I never want to be confused as a supporter of another team”

In this statement, the fan wanted to be both distinct from rival fans and part of the in-group. This function is elaborated through a Newport rugby fan who explained the importance of wearing apparel at an away game.

“The best bit’s going away – when you go to Bridgend, for instance, and in the small pub outside Bridgend’s ground, it’s absolutely packed in there with Black and

134

Amber jerseys, and you all walk in together – you don’t know each other but.. but you’re all part of the same thing …”

Clearly, the wearing of apparel heightens the salience of the identity through instant visual categorisation, and induces group cohesion very quickly. This emphasises the depersonalisation process suggested in social categorisation theory, where group members are able to ‘turn off’ their personal identity (Haslam, 2000; Turner, 1999) rapidly in response to a change in context. However, the use of apparel as a means of distinction and acceptance happens away from the match day as well, suggesting that fans may never leave their social identity completely. This can be seen in the form of car stickers, flags, computer avatars and screen savers. Pan and Baker (2005) showed how fans who had to travel to games in cars used this as an opportunity to display their apparel in their windows, and therefore to demonstrate that they were a committed supporter of that team. Additional use of apparel is illustrated in the following statement:

“I want to be a good supporter so I will buy their stuff... y’know shirts and stuff. I want the other fans to see I am a good fan…especially when we win… I go crazy, put pictures and that on my PC too”.

Additionally, another fan expressed that commitment was more than just turning up. He believed that wearing apparel lets the team know they are supported.

“Being a supporter is important and it’s important to show the team how much… if they you see you in their jersey then they’ll know you’re loyal and support them… it’s about showing loyalty as well, not just turning up”.

In addition to the involvement theme, this theme also appears to suggest that a large part of commitment is being seen to be committed. This supports one of the assumptions in this thesis. In Chapter One for example, it was suggested that part of

135

being a committed fan was sticking with your team when they lost. In order to do this, one would need their team to lose. Being able to demonstrate one’s commitment after a loss, also highlights to other fans that you are committed. Therefore, it may be more important to be seen as a committed fan, rather than a fan of a good team. One way fans can achieve this is by wearing their apparel. Fans for example may continue to wear their shirts and show their association despite their team losing in order to demonstrate their commitment. Since, most of the literature has labelled such behaviour as BIRGing (Cialdini et al, 1976; Hirt et al, 1992; Wann & Polk, 2007), naming this theme BIRGing was considered to be appropriate in relation to the literature.

Theme 4: Self-attribution A fourth theme labelled self-attribution was identified that appeared to indicate how fans might reduce the impact of an acceptance threat (see Branscombe et al, 1999) to their self-concept by focusing on and affirming their competence in other ways or comparative dimensions, such as club history, player and knowledge level, and tests of ficklness (i.e. I still made it to the game even though it was raining). This theme emerged when fans were asked about what types of things made a good supporter. In addition to involvement, fans felt that individual knowledge was also important, as well as individual need for self-verification and individual acceptance as a committed fan. These became two recurring subthemes. Although they are not natural opposites, they were discussed within the context of individual choices, where knowledge level was a positive domain and self-verification was a negative domain of self-attribution. For example, fans acknowledged and regarded

136

knowledge about their team as a positive attribute. This is illustrated in the following statement.

“To know the history of my team and town gives me a lot of pride”.

Although knowledge was held in high regard, this was not as important as showing your support at matches.

“Knowing things about the team is important but supporting them is more important … knowing how many points they lost by and stuff won’t help them win a game...Shouting will! Some stuff is good to know and shows to your mates you know stuff but in the terraces so what! You need to be a good supporter too and turn up”.

This type of connection with the team seems to be a more individual component of one’s commitment and is supported in more individualistic measures of self-identification, personality and individual differences within groups (e.g. NarioRedmond, Biernat, Eidelman, Palenske, 2004, Trail & Fink, 2004). In contrast, this type of behaviour was also indicated by Ashmore et al’s (2004) behavioural involvement component of collective identity, but fans in this research clearly indicated that it was a personal preference if one has the time to develop additional information about their team. Similarly, Shank & Beasley (1998) suggested that fans with strong attachments are likely to spend more time thinking and reading about their team. However, this appears to be something that is unimportant to showing one’s commitment at a game. Knowledge then is an important concept to being a committed supporter but it is not an essential component as it appeared to be a disposable attribute in terms of defining how committed one fan is to their team. In the focus group debrief for example, one fan suggested that highly knowledgeable fans were compensating for

137

not being regular attendees. However, there is some evidence to show that highly identified fans do have greater general knowledge about their teams than lowly identified fans (e.g. Wann & Branscombe, 1995; Wann Meltcalf, Adcock, Choi, Dallas, and Slaton, 1997), and as discussed, these fans are also more likely to attend (Wann et al, 2001). In contrast, Wann, Morris-Shirkey, Peters, and Suggs (2002) found that lowly identified fans were as knowledgeable about their team as highly identified fans, but that differences were in the type of knowledge being asked to recall. According to Wann and colleagues, highly identified fans were more knowledgeable with performance irrelevant information, but decreased in accuracy of performance relevant knowledge. According to the authors, this may be because highly identified fans are more motivated to perceive the team in a more positive and biased manner. In other words, they chose to not remember or recall their team’s losses as this may have increased their self-perceptions about the positivity of their social identity. In contrast, fans indicated a more negative form of self-attribution labelled self-verification. In many ways, this theme contrasts the most with the other themes identitfied. Fans that were did not seem to provide self-verification as a competent fan were seen as disloyal, uninvolved, unworthy, lacked commitment, and generally disliked. In other words, these fans were also seen as fair-weather fans. This theme may have emerged because fans low in involvement, BIRGing, and affective loyalty, might be considered atypical of the group and therefore ‘individualistic’. In terms of social identity literature, group members have been found to explain deviant members as’ black sheep’ (Dietz-Uhler, 1999). Similarly, other group members have been found to express their group’s heterogeneity when its image is questioned (Prislin & Christensen, 2005). This highlights the notion of the importance of a ‘shared’ social

138

identity for a group. These behaviours clearly represent CORFing behaviour described in the sport fan literature and discussed in the literature review. It is unsurprising to find that these fans were generally disliked (cf. Wann & Branscombe, 1990). For example, CORFing was illustrated by the following statements by fans.

“I think Newport fans are quite fickle in many ways – you could find 2000 or 3000 less at the gate..Depends who is playing… they can change their support”.

Additionally, other more individualistic behaviours were discussed where fans made pseudo-commitments to the team based on losing streaks or favourite players and coaches.

“If we don’t do well this season, these people won’t be there next season. Same if star players leave. I’ll be here obviously but not everybody will”.

These fans were strongly disliked, but more interestingly, it shows that fans can attach themselves to other objects associated with their team. One particular comment highlighted by younger male fans indicated that they disliked the involvement of female fans.

“some girls come and watch the team for the players..then they leave when the players leave. They spoil the game”.

As we saw in study one, there is some difference in motives for attendance, but actual amounts of commitment and identification have not been reported to be different between male and females (e.g. Dietz-Uhler & Murrell, 1999). Although such a statement may initially provide additional evidence for different motives in attending between males and females, the majority of younger boys felt that attending for other reasons other than to get a team win was not prototypical of a ‘true’ fan.

139

Other fans believed that support should continue when a fan’s personal circumstances change. In the involvement theme, one fan applauded the efforts of a man who travelled from Nottingham every week, and another fan explained his disappointment over a fellow fan they knew who had become less involved. This person had moved 20 minutes away and changed their support to a grassroots club.The following comment clearly highlights the voluntary nature of joining a team, as well as demonstrating the permeability of a fan group since membership is not highly regulated in this type of group. “My mate moved up the road ‘bout 20 minutes away, and he stopped coming. He disappointed me as I had to go on my own..,. he started going to his local team, Ton-yr-efail – you know, like a grassroots rugby team. He still supports Cardiff but he just doesn’t come anymore… he made the effort at first but I think he just changed what team he likes”.

3.3.4 Conclusion Despite the different ways fans expressed it, the overall concept of commitment was clearly important to the participants as they often criticised and distinguished themselves from other fans for being ‘fair-weather’ or uncommitted fans. These were people who would only attend matches when the team was winning or would change the team they supported regularly. Unsurprisingly, the participants all described themselves as generally more committed than other fans of their team and felt that their team’s fans were generally more committed than those of rival teams. This supports the notion of in-group favouritism hypothesised by a social identity perspective and the meta-contrast principles of self-categorisation theory (see Haslam, 2004; Hogg et al, 2004; Oakes, 2002; Oakes et al, 1994; Turner, 1999). Commitment therefore, seems to be used much more explicitly as an identity enhancement tool to make favourable comparisons over other fans, and therefore perhaps contributes to elevating one’s social identity and self-esteem. 140

The literature presenting commitment as having an affective element (see Allen & Meyer, 1990; Ellemers et al, 1999b), appears to be supported in the current study’s initial analysis. In addition to an affective element, fans indicated other components to their commitment that have not been addressed by other conceptualisations of commitment. Since this current study aimed to investigate how fans understood and perceived commitment, there seems to be initial evidence for the potential of a multidimensional construct. This would then contrast with social identity’s conceptualisation of commitment. Furthermore, many of the themes seemed strongly related to each other but at the same time addressed different aspects of commitment. Therefore, this may give more insight into the nature of the construct, and to examine this construct further, each theme was used as a basis for developing an inventory for assessing fans’ commitment.

Study Three 3.4.1 Method In the previous study, four main themes emerged from the data. These themes already appear to contrast with previous conceptualisations of commitment from both social identity perspectives and models of fan commitment in literature about sport fans. In this current study, the aim is to develop an inventory to assess the four themes. To do this, a combination of steps for inventory development proposed by DeVellis (2003) was used, and is outlined below.

1. Defining the construct In this stage, DeVellis suggests using theory to define a construct. Since there is conflict between disciplines about the construct of commitment, it makes defining 141

the construct difficult. In addition, since the conflicts prompted the development of a new model of commitment in the current study, the nature of commitment is still yet unknown. However, using novel sources in the previous study to examine commitment, additional factors that might be involved have been identified

2. Generate an item pool The purpose of this step according to DeVellis is to generate a large pool of items that are candidates for eventual inclusion into the scale. As part of item generation, theory and literature should provide the basis (Goldberg & Digman, 1994; Reisse et al, 2000). Since the literature is conflicting at best, it was felt that asking the intended sample about their understanding of commitment would be a more helpful and conducive approach. Therefore, a qualitative methodology was employed to examine commitment. Items were generated based on the themes described in the previous study. This resulted in a 19 item pool 8 reflecting the four themes. The items were also subjected to independent cross validation in a similar manner as the thematic analysis in the previous study (see footnote 8).

3. Determine the format for measurement For the development of the instrument, a Likert scale format was chosen.This is because it gives the participant an opportunity to respond to statements in gradations; in other words give an indication of strength of agreement. In addition, DeVellis suggests that Likert scales are particularly useful for assessing attitudes,

8

There is no one rule for coming up with a specific number of items, although it is generally accepted that more items increase a scale’s reliability (DeVellis, 2003). Since this study was exploratory, it was decided that number of items in the early stages was of less concern but rather content validity was more important. Items of similar content validity can be added and re-tested.

142

beliefs, and opinions. Since the previous study investigated what fans believed commitment to be, this scale sought to assess their beliefs about commitment, and therefore a Likert scale was deemed appropriate.

4. Have the initial item pool reviewed DeVellis (2003) points out that a jury of experts helps to confirm or invalidate the items, as well as evaluate their clarity. Given that the literature is sparse about sport fans (Wann & Hamlet, 1995; Wann & Pierce, 2003), obtaining experts to review the items was impractical. DeVellis recommends that regardless of having experts, that for clarity of the items, having at least one reviewer that has knowledge of scale development is essential. This is because experts without scale development knowledge may give poor advice about items. To resolve this issue, the same researchers were used from the previous study to provide independent cross validation of the items and ensure that they mapped onto each theme from the previous study. Two of the six researchers had significant knowledge of scale development.

5. Consider inclusion of validation items DeVellis points out that additional items from other sources may help establish construct validity. If the construct to be measured strongly relates to other constructs, then the performance of the scale vis á vis measures of those other constructs can serve as evidence of its validity. In addition, the resultant pattern of relationships could provide support for claims of validity, or clues as to why the set of items do not perform as anticipated. In the current study, items from the Psychological Commitment to Team scale (Mahoney et al, 2000) were used. This

143

was because, that of the scales used to measure fan commitment, this scale was most controversial in terms of psychometric properties (see Kwon & Trail, 2003; Kwon & Armstrong, 2004; Wann & Pierce, 2003). This therefore allowed the opportunity to examine the items in more detail. Furthermore, in its original development and in Kwon & Trail’s (2003) tests of construct validity, items from the PCT were administered to university students in a classroom. Since the current study will use fans at a game, this allowed the opportunity to examine the real world usefulness for some of these items in the new commitment scale on sport fans in the field. This would add further validation for the PCT.

6. Administer items to a development sample Spector (1992) recommends that the development sample should be as representative as possible of the ultimate population for which the scale is proposed. Since this scale is intended for sport fans, fans of Welsh rugby were chosen as it is a popular sport in Wales, and for consistency, the themes in the previous study used fans of Welsh rugby. There does however, seem to be disagreement on the size of the sample needed for administering the items. DeVellis (2003) recommends that the pool of items should be administered to a large sample of between 5 to 10 participants per item. Nunnally (1978) for example, suggests that a sample size of no less than 300 is adequate particularly for performing factor analysis. Costello & Osborne (2005) recommend having a participant to item ratio of no less than 10:1, and that a more reliable ratio would be 20:1. Generally, however, they suggested the more participants the better. Lastly, MacCallum et al (1999) suggest that a minimum of 500 participants should be used, particularly when the content and construct validity of the items have not been established with factor analysis. Using Costello &

144

Osborne’s 20:1 ratio as a baseline, the minimum number of participants needed for the 19 item commitment scale would be 380.

7. Evaluate the items According to DeVellis (2003), this stage is the heart of scale development, and is second in importance to perhaps the item development stage. In this stage, DeVellis suggests two key tests of item analysis. These include tests of content validity that include testing the scale’s reliability and internal consistency, and tests of factor analysis that determine how many latent variables underlie a set of items. Since the previous study showed that commitment may be multidimensional, Spector (1992) adds that factor analysis can be used to verify that the items empirically compose the intended subscales.

8. Optimize scale length In DeVellis’s last stage, an investigator should have a reliable scale and that the length of the scale should be optimised. This involves reducing the number of items without compromising the scale’s integrity. In general, this involves repeating step seven by removing items that may affect the scale’s content and construct validity (i.e. lower factor loadings and inter-item correlations). The benefit for repeating these steps, in addition to a more reliable and accurate scale of the construct in question is a more parsimonious scale. This is of benefit to both researcher and participant since it is likely to increase participant’s willingness to respond more diligently to a shorter scale. Therefore, it will avoid set response acquiescence whilst providing researchers with a valid data set (Coolican, 2007; see also Terry et al, 1999).

145

Participants To test the initial scale’s four themes, 998 (759 males and 177 females) fans of Welsh premier rugby9 were used. This sample had an age range between 8-86years (X = 45.5 yrs, SD = 16.5) and had supported their team on average for 43.25 years (SD = 15.3). Unfortunately, not all respondents had completed the entire scale, so this number was reduced to 842 who had provided complete data.

Item generation To examine the four themes identified in study two, 19 items 10 were generated based on the four themes identified (affective loyalty; e.g., “Nothing could change my allegiance to my team”, BIRGing; e.g. “I often wear my team’s colours”, involvement; e.g., “I watch every game I can that features my team”, and selfattrbution; e.g., “I don’t know much about the players and history of my club”), and these are shown in Table 6. Participants were asked to indicate to what extent each item applied to them on a 7-point Likert scale from 1 (strongly disagree) to 7 (strongly agree). This scale was chosen as scales larger than the traditional 3 or 5 point Likert scale are more reliable and allow for greater item-level factor analysis, and thus better psychometric properties (Reise, Waller, & Comrey, 2000).

Procedures To examine the construct of commitment, a 19 item commitment scale (the Sport Fan Commitment scale, SFCS) was administered at the home ground of each of

9

Since the thematic analysis showed that the amount of effort made in attending games was considered to be an important part of fans’ commitment, both home and away fans were used to develop the new Sport Fan Commitment scale. 10 In line with DeVellis’s step 5 of inventory development, some items from the PCT were also used to add content validity. Items seen to be appropriate for describing some of the themes identified from the thematic analysis were used.

146

the nine premier league Welsh rugby clubs over two weekends at the beginning of the season. The questionnaires were distributed to fans as they entered the ground, and fans in the clubhouse before kick off, as well as to those in the ground during half time. Fans were instructed that they could complete the questionnaire in their own time, and freepost envelopes were provided to return the completed scale. As an incentive, respondents were entered into a prize draw to win international tickets provided by the Welsh Rugby Union.

Data analysis To examine whether the SFCS possesses internal consistency and reliability, the scale was examined using Cronbach’s Alpha. Following tests of reliability, the nature of the factor structure was then examined using exploratory factor analysis and principal components analysis 11

as the main method of extraction in the same

manner described by Luhtanen & Crocker (1992), and used in study one.

3.4.2 Results Firstly, since there was a large gender difference, in the same manner as study one; an independent t-test was employed to examine the mean scores for male and females on the 19 item Sport Fan Commitment scale. As in study one, there were no significant differences (t (256) = .231, p > 0.05) between male (X = 113, SD = 2.6) and female’s (X = 115, SD = 3.1) mean scores on the 19 item Sport Fan Commitment scale.

11

Principal component analysis extraction method was used in the factor analysis because it is better than the preferred maximum likelihood method used in exploratory factor analysis at summarising and reducing variables of a construct in a more efficient manner. The result is a more parsimonious association between variables of common factors that contribute to a construct, such as commitment (for a review of extraction methods see Fabrigar, Wegner, MacCallum, and Strahan, 1999)

147

Reliability To examine internal reliability and consistency, Cronbach’s alpha was calculated for the total scale. Analysis indicated that the scale was internally consistent with all 19 items revealing a substantial alpha score of 0.79. Item correlation scores ranged from 0.37 to 0.74, and removing lower loading items did not significantly improve the scale’s alpha levels, therefore all items were kept for factor analysis.

Factor analysis As suggested by DeVellis (2003), the proposed four factor structure of the commitment scale was subjected to factor analysis. To do this, a principal components analysis using varimax rotation was used. To examine whether the items loaded into their appropriate factors, rather than force the analysis to explain the data within a set number of factors like confirmatory factor analysis, the number of factors were not specified so that variables were free to load on all factors (cf. Stevens, 1996). Such a method is in line with exploratory factor analysis where it is used to summarize and explain the data and relationship between variables as common factors (Coolican, 2007; Gorsuch, 1983; Schumacker & Lomax, 1996). In support of the thematic analysis, the factor analysis indicated a model including four factors and explaining 52.0% of the variance. These four factors complemented the themes that emerged from the focus group, with each theme loading as a separate factor: loyalty and involvement factors (both with four items), a BIRGing factor (three items), and a factor comprising self-attribution (6 items), and can be seen in Table 6. Some items cross-loaded into the self-attribution factor. This may be because that a fan with low loyalty, poor attendance, and low BIRGing is

148

Table 6. Factor loadings for the 19 item Sport Fan Commitment scale

Factor Loading a Loyalty

1

It would be unlikely for me to change my allegiance from my team to

2

3

4

.71

another I could never switch my loyalty even if my close friends were fans of

.77

another team Nothing could change my allegiance to my team

.70

I could easily be persuaded to change my team preference

b

.48

.46

Involvement If I were away from home I would make every effort to

.65

follow my team I travel to away games as much as possible

.70

I watch every game I can that features my team

.60

I would not go to games alone if friends or family couldn’t make it I am a not a very committed fan of my team

b

.41

b

.48

BIRGing I really dislike the biggest rivals of my team

.53

I often wear my teams colours

.70

I never display my teams name or logo at work on the car Being a fan of my team is not that important to me

b

.69

b

.42

Self-attribution I don’t know much about the players in my team b

.51

It would affect my loyalty to my team if players or a coach that I really

.56

disliked were employed b If I moved to another town, it’s unlikely that I would still support my team b My commitment to my team would decrease if they were losing consistently I am one of my teams biggest fans

.49 b

.62 .57

Commitments to others come before my commitment to my team

b

a as in study one, factors were extracted based on Eigenvalues greater than 1, and factor loadings equal to or greater than .30 are indicated b Item was reversed for coding

likely to be considered atypical, and thus a black sheep (cf. Dietz-Uhler, 1999). It might be that this subscale may be problematic since fans that score lowly in the other subscales would be considered an atypical group member, which would then make

149

the self-attrbution subscale redundant. However, there are disproportionate items in each factor, which may have an effect on the strength of any given factor.

3.4.3 Discussion From the preliminary analysis of the 19 item scale, there appears to be initial support for a multi-factor model of commitment that has content and construct validity. This promising initial result was enough to merit further exploration of the factor structure in terms of understanding the relationship between each factor. In DeVellis’s (2003) guide for inventory development, the next stage involves optimising the scale’s length. Indeed, one rationale for factor analysis and confirmatory factor analysis is to find the most efficient factor structure to explain a construct (Coolican, 2007; Kline, 1998; Schumacker & Lomax, 1996). Before doing this, however, analysis showed a disproportionate number of items for each subscale. This was because too many items were generated for some themes. It was therefore deemed necessary to first develop a more parsimonious scale to measure commitment that had equal items in each factor. This is because having more items in one factor can give the impression that a factor is stronger than it is, or that a factor is there when it may not be (Jensen, 1998; see also Terry et al, 1999). Having a larger number of similar items in one factor also increases the probability of items correlating and thus gives the impression of a strong factor (Reise, Waller, & Comrey, 2000), and thus a scale with content and construct validity. Additionally, over representation (psychometric sampling error, Reise et al, 2000) of one factor does not accurately portray a construct if items in one factor dominate a factor analysis. This is a problem commonly reported, for example, in general intelligence scales, where some items measuring certain types of intelligence (e.g. spatial ability, language, reasoning) have been over represented (see Jensen, 1998). However, one cannot simply delete items 150

that load poorly and infer that the model will be stronger. According to Lane, Harwood, Terry, and Karageorghis (2004) if items are unnecessarily deleted in the preliminary stages of model development, then this may affect the construct validity and factor structure of the scale. This would then result in a scale that is parsimonious, but may not be representative of the construct in question. Rather, it is best to delete items and repeat the analysis with a different sample. As part of making a scale parsimonious, items with low factor loadings are generally the first to be dismissed from a scale (cf. DeVellis, 2003). However, as the initial pool of 19 items was too few in this study, and since there was an unequal amount of items in the subscales, deleting lower loading items would still leave the scale disproportionate. One resolution for unnecessarily deleting items without significantly compromising the factor structure is to be over inclusive of items in each factor (Reise et al, 2000, Tellegen & Waller, 2001). According to Tellegen & Waller (2001), this allows factor analysis to provide better empirical evidence for what type of item content belongs in a subscale and what belongs in another. Secondly, it allows more powerful item correlations that are more reliable and valid for each common factor. Once this has been done, reliability and factor analysis of the items can be rerun, allowing a stronger case for deleting lower loading items. In addition to adding new items 12 for each common factor, some existing items were reworded (based on low inter-item correlation scores), and a new 29 item measure using the same 7-point Likert scale was developed.

12

Several additional validation items from the PCT that seemed appropriate for assessing some of the factors were used in a manner proposed by DeVellis’s (2003) stage five for inventory development.

151

Study Four 3.5.1 Method Participants Participants were 556 (462 male and 94 female) fans of Welsh regional rugby from five of the new 13 regional clubs. The mean age of the fans was 44.32 years (SD = 15.4) and on average they had supported rugby for 24.3 years (SD = 15.7).

Procedure Using the same procedure as study one, the new 29 item scale was distributed to fans at each of the Welsh regional clubs’ home ground at the beginning of the season. As with the CSES adaptation in the previous section, these fixtures were chosen due to the importance of the matches and the subsequent anticipated higher attendances. Fans were handed surveys that included the SFCS in a freepost envelope. In return for completing the survey, fans’ surveys were entered into a draw to win international rugby tickets.

Data Analysis To examine whether the SFCS possessed internal consistency and reliability, the scale was examined using Cronbach's Alpha. Following tests of reliability, the nature of the factor structure was then examined using exploratory factor analysis in the same manner described by Luhtanen & Crocker (1992). Once the common factor structure was identified for the SFCS, a test of confirmatory factor analysis was administered on five hypothesised models of commitment in a manner described by

13

At the time of this study the structure of Welsh rugby was being changed with the former nine premier teams being combined to form five regional teams.

152

Luhtanen & Crocker (1992). This was to test the reliability of the factor structure and the construct validity.

3.5.2 Results Since a different sample in this study was used to study three, and items were added to the scale, another independent t-test was performed to examine differences between male and female mean commitment scores. Again, there was no significant differences (t (621) = 1.34, p > 0.05 between male (X = 187, SD = 2.4) and female’s (X = 183, SD = 1.8) scores. Reliability To examine internal reliability and consistency, Cronbach’s alphas were calculated for the total scale and subscales. Analysis indicated that the scale was internally consistent with all 29 items revealing a substantial alpha score of 0.85. Total item correlation scores ranged from 0.12 to 0.60, and removing the lower loading items did not significantly improve the scale’s alpha levels, suggesting again that the scale was internally consistent. Therefore, all items were retained for factor analysis.

Factor analysis: 29 item SFCS To confirm the previous four factor model found in study two, a principal components analysis with varimax rotation was used to examine the 29 item SFCS. The results were similar to the previous four factor 19 item scale, however this time they explained only 40% of the variance (factor loadings are presented in Table 7) Three of these factors related to high commitment and were affective loyalty, involvement, and BIRGing. The final factor ‘self-attribution’ comprised of items

153

Table 7 Factor loading scores for the 29 item Sport Fan Commitment scale

Factor Loading a Loyalty

1

I really dislike the biggest rivals of my team

.52

I will occasionally support teams other than my own b

.55

Nothing could change my allegiance to my team

.54

I will never support another team

.73

My favourite players play for my team

.38

2

3

I could easily be persuaded to change my support to another team b I could never switch my loyalty even if my close friends were fans of another team It would be unlikely for me to change my allegiance from my team to another

4

.64 .45

.40

.50

Involvement If I were away from home I would make every effort to

.50

follow my team I travel to away games as much as possible

.62

I am a not a very committed fan of my team b

.48

I make every effort to go and watch my team play away

.40

.60

I would be unlikely to attend games by myself b

.39

I watch every game I can that features my team

.30

I make every effort to go and watch my team play at home

.39

BIRGing I would be unlikely to buy merchandise associated with my

.55

team b I never display my teams name or logo in public (e.g. at

.33

.62

work or in the car) b I often wear my team’s colours

.75

I am proud to wear my team’s shirt

.78

Being a supporter of my team is not that important to me b

.52

.38

Self-attribution I would switch my loyalty to another team if my favourite

.59

player moved b If I moved to another town, it is unlikely that I would still

.54

support my team b My commitment to my team would decrease if they were losing consistently b I am one of my teams biggest fans

.48 .32

.39

Commitments to others come before my commitment to my team b

.46

I don’t know much about the players from my team b

.41

I could easily name the coaching staff at my team

.51

I wouldn’t recognise players from my team if I saw them in the street b

.47

I would find it easy to name players in my team

-

-

-

a Only factor loadings equal to or greater than .30 are indicated b Item was reversed for coding

154

-

showing the amount of knowledge about the team and lack of willingness to engage in self-verification of loyalty, such as switching to a new team if individual circumstances changed. In addition, nothing over four factors or under four factors helped to explain the variance. After the four factors had been reconfirmed, the reliability of each subscale were also examined using Cronbach’s alpha. These were found to be moderate ranging from 0.54 (self-attribution) to 0.80 (BIRGing), and removing lower loading items did not significantly improve the subscale’s alpha levels. This suggests that like the main scale, each subscale was also internally consistent.

Table 8 Mean and alpha scores for the Sport Fan Commitment Scale.

Loyalty

Involved

BIRGing

Individual Total

17.9 (6.2)

16.7 (6.4)

20.0 (7.1)

23.7 (4.5)

78.1 (16.3)

Study 5

17.7 (7.3)

16.1 (7.3)

20.5 (6.8)

23.5 (5.6)

79.5 (15.6)

Study 6 – 3 factor

19.5 (5.8)

18.4 (7.0)

20.9 (5.7)

-

58.7* (15.9)

Study 6 – 4 factor

19.9 (6.7)

18.6 (6.6)

20.8 (5.7)

23.3 (4.8)

82.5 (17.5)

.69

.81

.83

.58

.81

Study 5

.88

.92

.79

.88

.83

Study 6 – 3 factor

.90

.93

.79

-

.91

Study 6 – 4 factor

.80

.86

.79

.74

.89

Mean score (S.D) Study 4

16 item

Chronbach Alphas Study 4

16 item

* Total SFCS score based on 12 rather than 16 items.

Construct validity To further examine the construct validity of the scale and subscales, and that each common factor measured a related but different aspect of commitment,

155

correlations between the new subscales and the total commitment scale were also calculated and are presented with correlations from the subsequent studies in Table 9 for comparative purposes. As a guide, Cohen (1988) suggests that moderate correlations (between 0.40 and 0.60) for subscales with four items show internal consistency and initial construct validity (Lowenthal, 2001). In this analysis, subscales showed that each factor was moderately correlated with each other factor.

Table 9.Correlations for the Sport Fan Commitment Scale

Support

Identify

Individual

Total

.19**

.24**

.19**

.60**

Study 5

.25**

.34**

.07

.56**

Study 6 – 3 factor

.44**

.50**

Study 6 – 4 factor

.39**

.56**

.34**

.78**

.43**

.20**

.70**

Study 5

.30**

.04*

.56**

Study 6 – 4 factor

.56**

.42**

.75**

Study 6 – 3 factor

.55**

Loyalty Study 4

16 item

.83**

Involvement Study 4

16 item

.83**

BIRGing Study 4

16 item

.31**

.77**

Study 5

.15**

.80**

Study 6 – 4 factor

.43**

.82**

Study 6 – 3 factor

.81**

Individual Study 4

16 item

.58**

Study 5

.35**

Study 6 – 4 factor

.69**

Study 6 – 3 factor

-

** p