UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Attitudes, interests and sport spectator profiling : variables affecting game attendance in a NHL franchise 2003

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_2004-901467.pdf
ubc_2004-901467.pdf [ 6.09MB ]
Metadata
JSON: 1.0077060.json
JSON-LD: 1.0077060+ld.json
RDF/XML (Pretty): 1.0077060.xml
RDF/JSON: 1.0077060+rdf.json
Turtle: 1.0077060+rdf-turtle.txt
N-Triples: 1.0077060+rdf-ntriples.txt
Citation
1.0077060.ris

Full Text

ATTITUDES, INTERESTS AND SPORT SPECTATOR PROFILING: VARIABLES AFFECTING GAME ATTENDANCE IN A NHL FRANCHISE by SANG WOO BAE B.Sc., Yonsei University, 1990 M.Sc, Yonsei University, 1992 M.H.K., The University of British Columbia, 1996 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (School of Human Kinetics) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA November 2003 © Sang Woo Bae, 2003 Abstract ATTITUDES, INTERESTS AND SPORT SPECTATOR PROFILING: VARIABLES AFFECTING GAME ATTENDANCE IN A NHL FRANCHISE This study examined the literature on spectator profiling and evaluated current survey methodologies being used to profile hockey spectators and their preferences regarding NHL hockey game attendance. The research had two inter-related objectives. One was to investigate the consumer profiles of sport spectators associated with a specific NHL franchise, the Vancouver Canucks, and to relate these to game attendance. The second objective was to test and evaluate a questionnaire designed to incorporate latest thinking on variables that affect professional sports game attendance. Survey data were collected from 907 attendees at four Vancouver Canucks games and from 203 students at the University of British Columbia during the 2001-2002 NHL season. Two separate studies were conducted using the data. Study 1 investigated reasons for game attendance, and respondents' activity patterns and interests. Data analyses for study 1 included a confirmatory factor analysis for testing the validity of the game attendance items (reasons for game attendance), and a decision tree analysis for classifying spectators based on their reported game attendance frequencies, activity patterns, and interests. Results of study 1 indicate that the respondents' game attendance rates were affected by income, use of hockey websites, playing baseball when young, job, being accompanied by co-workers, ethnic background, perception of T.V. as the best promotional medium, personal fan history, and playing hockey when young. Study 2 investigated the relationship between spectators' opinions about sports and their game attendance, and attempted to identify potential predictors of game attendance levels. A confirmatory factor analysis was used in Study 2 to test the validity of the sport opinion items (opinions about sports), and a cluster analysis was used to identify unique spectator groups. Three distinctive spectator segments were identified, and the three groups varied on a number of dimensions including game attendance rates, sports participation history, and their opinions about the benefits of sports, community pride, cynicism about professional sports, attitude toward professional athletes, perceived time costs, role of government, and attitude toward amateur sports. The research found evidence of national cultural differences between Canadian sports spectators and their American counterparts. It is recommended that future research investigate national and regional cultural differences as well as differences between major and minor league sports and competitive factors in local sports markets. iii Table of Contents Abstract ii Table of Contents iv List of Tables vi List of Figures vii Chapter 1. Introduction 1 Chapter 2. Review of Literature 4 Theories of Consumer Behaviour 4 Fishbein Model and the Theory of Reasoned Action 4 Means-End Theory 7 Game Attendance Variables Identified In Prior Research 8 Additional Key Variables 14 Dwelling Information 15 Psychographic Information 15 Chapter 3. Methodology 22 Subjects 22 Overview of Instrument Design and Validity 23 Defining Constructs and Selecting Items 26 Game Attendance 26 Sports Interests and Activities.. 33 Personal Background 33 Opinions about Sports 35 Review by Experts for Content Validity 37 Pilot on Small Sample for Clarity and Ambiguities 38 Test on Large Sample for Internal Consistency and Internal Structural Validity 39 Test on Large Samples for Convergent and Discriminant Validity 41 Test on a Large Sample for Criterion-Related Validity 42 Data Collection and Data Analysis 42 Collection of Data 42 Data Analysis 43 Chapter 4. Results 46 Descriptive Statistics 46 Response Rate 46 Demographics 46 iv Study 1 51 Psychometric Evaluation of Reasons for Game Attendance 51 Predicting Game Attendance Using Decision Tree Analysis 57 Comparing Low, Middle, and High Game Attendance Groups v. 68 Study 2 74 Psychometric Evaluation of Opinions about Sports 74 Cluster Analysis of Opinions about Sports 78 Predicting Game Attendance using Decision Tree Analysis 90 Chapter 5. Discussion and Implications 94 Examination of Study 1 94 Examination of Study 2 103 Cluster 1 104 Cluster 2 105 Cluster 3 106 Short-term Marketing Implications 109 The Use of Internet 110 Ticket Prices I l l Perceptions of Pro-sports and Athletes I l l Long-term Marketing Implications 112 Limitations 114 Future Research " 5 References 117 Appendix A 123 Appendix B 132 Appendix C 136 Appendix D 139 v List of Tables Table 1. Parts and Components 23 Table 2. Constructs and Items in Hockey-Related Studies 27 Table 3. Constructs and Items for Part 1 30 Table 4. Questionnaire Items for Part 2 and 3 34 Table 5. Constructs and Items for Part 4 36 Table 6. Questionnaire Forms and Components 39 Table 7. Data Sets and Component 44 Table 8. Data Sets and Components 46 Table 9. Descriptive Statistics for Demographic Variables 47 Table 10. Descriptive Statistics for Attendance Level 49 Table 11. Results of the Confirmatory Factor Analyses for Part 1 52 Table 12. Factors, Items, Means, and Loading scores of the Final Model 53 Table 13. Descriptive Statistics, Internal Consistency and Test-Retest Reliabilities....54 Table 14. Correlations between Latent Constructs with Zhang's Scale 56 Table 15. Correlation Coefficients for Criterion Validity 57 Table 16. Variables for the Initial Decision Tree Analysis 58 Table 17. Variables for the Second Decision Tree Analysis 63 Table 18. Gain Chart Defined by 9 Predictors 67 Table 19. Group Descriptive Statistics of Activity Pattern 69 Table 20. Demographic Profiles 72 Table 21. Results of the Confirmatory Factor Analyses for Opinions about Sports 75 Table 22. Factors, Items, Means, and Loading scores of the Final Model 75 Table 23. Descriptive Statistics, Internal Consistency and Test-Retest Reliabilities....77 Table 24. Inter-Factor Correlations 77 Table 25. The Number of Clusters and Cases per Cluster 78 Table 26. Cluster Centroids from the Hierarchical Cluster Analysis 80 Table 27. Overlap between K-Means and Ward's Methods 80 Table 28. Cluster Centroids, MANOVA Table, & Profiles of the Three-Cluster Solution 81 Table 29. Prediction Result from the Discriminant Analysis for Cluster Membership 84 Table 30. Cluster Membership and the Game Attendance Frequency 85 Table 31. Profiles of the Three Clusters on Sports Interests & Activities 87 Table 32. Demographic Profiles of the Three Clusters 88 Table 33. Descriptive Statistics of 33 Items for Part 1 95 vi List of Figures Figure 1. Steps in Questionnaire Validity Test 25 Figure 2. Factor Analysis Model for Game Attendance 45 Figure 3. Tree Diagram of the First Analysis 61 Figure 4. Tree Diagram of the Second Analysis with Data set 4 65 Figure 5. Tree Diagram of the Final Model with Data set 4 66 Figure 6. K-Means Method Clusters Plotted on the Principal Components 82 Figure 7. Cluster Profiles for the Three-Cluster Solution 83 Figure 8. Tree Diagram on Activity Pattern and Opinions about Sports 92 Figure 9. Tree Diagram on Activity Pattern and Opinions about SportsOl 107 vii Chapter 1. Introduction Spectators are integral to the business of professional sports. As well as contributing to the excitement and ambience of sport events, spectators are a vital component in the revenue stream of most professional sport franchises and leagues. Spectators typically make a direct contribution to revenues through ticket sales, but they also contribute indirectly to sale of broadcast and sponsorship rights, and to licensing. In the case of the National Hockey League (NHL), ticket sales comprise a major source of income along with broadcast rights and corporate sponsorships (Badenhausen & Stanfl, 2000). Although ticket sales for games and sale of broadcast and sponsorship rights are major sources of revenue for all major professional sport leagues in North America, their actual contribution to total revenue varies from league to league and from franchise to franchise. For example, the National Football League (NFL) in 1996 earned the major part of its revenue - 55.3 percent (US $43 million of total revenue US $77.7 million) - from media contracts, whereas in the same season (1995-96) the NHL earned most of its revenue - 60.5 percent (US $25.6 million of total revenue US $42.3 million) - from ticket sales (Badenhausen & Nikolov, 1997). The picture that emerges from recent figures is that the NHL is in reasonable shape financially, but lacks a large network deal on a par with the other major professional sports leagues, and therefore ends up relying primarily on game attendance for revenue (Canfora, 2001). The average total revenue per franchise for the NHL was a respectable US $69 million for the 2001-02 NHL season (Badenhausen & Ozanian, 2002), and the total revenues are estimated to be US $1.9 billion for the 2002-03 season 1 (LeBrun, 2003). Current NHL broadcast agreements with ESPN/ABC, CBC, and TSN are worth US $600 million, CD $300 million, and CD $100 million respectively over five years, and each NHL franchise has about US $5.7 million shared TV contract revenue for the 2002-03 season (LeBrun, 2003). These are significant contracts and reflect positively on the new media environment for the NHL where they can sell rights to several different networks. The relatively low value of hockey to broadcasters, however, shows up when these figures are compared with the other major leagues in North America. The National Basketball Association (NBA), for example, has a six year deal worth US $4.6 billion with ESPN/ABC and AOL and US $26.4 million per team for the 2002-03 season, and the NFL has an eight year deal worth US $18 billion with ESPN/ABC, CBS and Fox and US $77.34 million per team in the 2002-03 season (LeBrun, 2003). Badenhausen and Stanfl (2000) indicate that arena sponsorship revenue has increased for the NHL from the late 1990s. Nevertheless, arena sponsorship remains relatively less important than ticket sales, and in any event, it too is responsive to spectator attendance and broadcast audiences. Given the continued importance of spectators for revenue production in the NHL, it is critical to gain insight into the reasons why spectators choose to attend NHL games. Understanding spectator behaviour is necessary to increase our knowledge regarding why people attend sport events in the first place, and uncovering the profile of spectators and the different reasons why different kinds of people attend games can be a valuable tool for spectator recruitment as well as retention. When spectators' expectations regarding games are met, there is a greater chance they will attend games. Using 2 research to develop profiles of spectator interests, likes and dislikes can help managers in the NHL to identify core customer groups and target markets, analyse market trends and the market environment, develop marketing strategies, attract corporate sponsors, and set long-term marketing objectives and plans. This dissertation examines the literature on spectator profiling and evaluates current survey methodologies being used to profile hockey spectators and their preferences regarding NHL hockey game attendance. To evaluate current methodology, a questionnaire was developed using variables identified in prior research relating to spectator interests and opinions and game attendance. By integrating these variables, the questionnaire served as a platform to holistically assess the explanatory power and the reliability and validity of attendance factors currently described in the literature, as well as the assumptions that have been made about spectator interests and game attendance. The dissertation is organized in five chapters. Following this introduction, Chapter 2 reviews prior research on spectator profiling and relevant theoretical frameworks in the literature. Chapter 3 sets out the methodology for the dissertation research project. Chapter 4 reports the research findings from the dissertation research project, and Chapter 5 discusses the implications and limitations of the findings as well as areas for further research. 3 Chapter 2. Review of Literature This chapter is organized in three parts. First I briefly discuss three theories of consumer behaviour that have been used to explain consumer decision-making: the Fishbein model, the theory of reasoned action, and means-ends theory. This is intended to provide a conceptual framework and background for understanding factors that influence consumer decisions to attend NHL games. In the second part of the chapter, I discuss prior research on variables relating to game attendance in sport, including some multi-sport studies that examined game attendance across a range of different sports that included hockey, and several studies that looked exclusively at hockey. Third, I identify some additional variables that have not been well covered in prior research. Theories of Consumer Behaviour Fishbein Model and the Theory of Reasoned Action The Fishbein model and the theory of reasoned action (the extended Fishbein model) provide a theoretical basis for understanding and analyzing consumer behaviour, and can be used to help explain consumer behaviour in sport. The Fishbein model which is one of the most influential multi-attribute attitude models focuses on measuring a consumer's attitude toward a product. (1) an individual holds many beliefs about any given object, i.e. many different characteristics, attributes, values, goals, and objects are positively or negatively associated with a given object; (2) associated with each of these 'related objects' is a mediating evaluative response, i.e. an attitude; (3) these evaluative responses summate; (4) through the mediation process, the summated evaluative response is associated with the attitude object, and thus (5) on future occasions the attitude object will elicit this summated evaluative response, i.e. this attitude (Fishbein, 1963, p.233) 4 The Fishbein model measures attitude along three dimensions: 1) an individual's beliefs about the "attitude object" (i.e., those beliefs about the object that are considered during evaluation), 2) a measure of each belief and its evaluative aspect (i.e., the probability that a particular object has an important attribute), and 3) a measure of his attitude toward the object (Fishbein, 1963). According to Solomon (1996), the predictability of the original Fishbein model, which focused on measuring a consumer's attitude towards a product, was improved by extending the number of factors in the model relating to consumer decision-making. The revised framework, known as the theory of reasoned action, includes all variables of the Fishbein model plus three new elements that concern the relationship of intention versus behaviour, social pressure, and attitude toward purchasing (Ajzen & Fishbein, 1975). The three elements are explained below. The theory of reasoned action offers a useful framework for conceptualising the reasons why people choose to attend sports events, and provides a holistic and comprehensive perspective on factors affecting attendance, particularly concerning the individual. Intentions versus behaviour. A number of factors might affect actual purchase behaviour, even though a consumer is willing to purchase a product (Ajzen & Fishbein, 1975). For instance, let us consider an individual who has an intention to attend a hockey game at a local ice arena. Prior to actually attending the game, a number of things could interfere with the person's intention to buy a ticket and go to the game. If, for instance, the tickets were already sold out or there were no means for getting to the arena, then the person logically would 5 choose not to go. Unexpected circumstances such as these can affect the actual buying behaviour of the consumer. The theory of reasoned action aims to measure behavioural intentions, recognizing that uncontrollable factors can inhibit behaviour and behaviour prediction. Identifying uncontrollable factors experienced by a consumer and working to solve potential problems can help to keep current customers and attract new customers. Social pressure. Based on the theory of reasoned action, other people may influence a particular individual's behaviour, which means that behaviour can occur based on social interactions (Ajzen & Fishbein, 1975). A new element in the theory of reasoned action, the subjective norm (SN), covers the socializing effects of other people on an individual's norms. According to Ajzen & Fishbein (1975), SN includes two other factors: the intensity of a normative belief that others believe an action should be taken or not taken and the motivation to comply with that belief. For instance, a person's actions to purchase a game ticket could be influenced by a normative belief such as a sense of community relating to supporting a local sport franchise or a sense of being committed to a peer group. Attitude toward buying. The third new element of the theory of reasoned action is attitude toward buying. An individual's positive or negative feeling associated with performing a specific behaviour. According to Ajzen and Fishbein (1975), in general, an individual will hold a favourable attitude toward a given behaviour if he/she believes that the performance of the behaviour will lead to mostly positive outcomes; on the other hand, if the individual 6 believes that mostly negative outcomes will result from the behaviour, he/she will hold a negative attitude toward it. Even if a group of people have a positive attitude toward the Vancouver Canucks, this information does not necessarily allow marketers to anticipate whether or not they will attend games. One way to predict their attendance at hockey games would be to conduct an interview to investigate their preferences in game attendance. While a person might have a positive attitude toward the Canucks, his/her attitude toward the act of buying a ticket might be negative due to other issues such as parking, the high price of tickets, and the location of the stadium. Uncovering the profile of extant spectators is very important, as they are the consumers who have already demonstrated a positive attitude towards the act of buying (Aact) with respect to the sport event. This is the population studied in this dissertation. Means-End Theory Means-end theory covers another range of factors that relate to why people attend professional sports events. Means-end theory focuses on the cognitive linkage between the relatively concrete aspects of products (means), the more abstract consequences these attributes provide for consumers, and the highly abstract dimension of personal values (the ends) (Klenosky, Gengler, & Mulvey, 1993). Based on means-end theory, Klenosky et al. (1993) uncovered four different domains of personal values relating to choices of skiing destinations including belonging, fun and excitement, safety, and achievement. Value is broadly defined as features of the world toward which people are oriented and as features of people that govern their orientation toward the world (Smith, 7 1991). More specifically, value is seen as an enduring belief that a specific mode of conduct or end-state of existence is personally or socially preferable to its opposite (Rokeach, 1973). According to Staub (1978), a personal norm is a person's expectation that he or she will engage in certain kinds of behaviour, presumably derived from personal values and beliefs. In other words, people's desirable outcomes from the same behaviours may be different because of their different personal values. In a marketing context, a person's set of values plays a very important role in his or her consumption activities, since many products and services are purchased because they will help us to attain a value-related goal (Solomon, 1996). For example, the reasons why people attend Canucks' games would be expected to vary according to their personal values. One group of spectators might seek game performance. Another group might go to games based on price point (i.e., ticket discount). Or yet another might value both of these dimensions of the game as an exciting and affordable social event. In this dissertation, I propose to test factors that influence game attendance. My research is not designed or intended to test theories of purchasing behaviour - (see Godin & Shephard, 1985; Dzewaltowski, 1989; Greenockle et al, 1990 for studies that tested Fishbein's theory) - but rather to use these theories as a conceptual framework to help identify" and evaluate variables that affect game attendance. Game Attendance Variables Identified In Prior Research Since 1970, a number of studies have examined variables that affect spectator decisions to attend games. Several different sports have been studied, including 8 baseball, (Baade & Tiehen, 1990; Carter, 1996; Marcum & Greenstein,1981, 1985), football (Carter, 1996; Hansen & Gauthier, 1989; Whitney, 1988; Kahle et al., 1996), hockey (Carter, 1996; Hansen & Gauthier, 1989; Jones, 1984; Smith, 1995; Zhang, 1998; Zhang et al., 1997a, 1997b, 1998, 1999, 2001), soccer (Bird, 1982; Hansen & Gauthier, 1989), golf (Hansen & Gauthier, 1993; Lascu, 1995), and basketball (Carter, 1996; Hansen & Gauthier, 1989; Whitney, 1988; Zak, Huang, & Siegfried, 1979; Zhang etal., 1995). Overall in these studies, five grouped variables have been found to affect spectator decisions to attend games (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989; Jones, 1984; Schofield, 1983; Smith, 1995; Whitney, 1988; Zhang et al., 1995; 1997a, 1998, 1999). The first variable is game performance which includes athletes' individual skills (Hansen & Gauthier, 1989), home and opposing teams' records (Hansen & Gauthier, 1989; Jones, 1984; Zhang et al., 1995, 1997a), league standing (Bird, 1982; Hansen & Gauthier, 1989; Zhang et al., 1995, 1997a), tough/aggressive play (Jones, 1984; Zhang et al. 1997a, 1999), player appeal/fighting (Zhang et al., 1997a, 1999) and superstars (Hansen & Gauthier, 1989; Johns, 1984; Zhang et al., 1995). Game performance variables have generally been positively related to game attendance and enjoyment (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989; Jones, 1984; Schofield, 1983; Zhang etal., 1995, 1997a, 1999). The second grouped variable encompasses promotional dimensions of attendance. It includes ticket price (Bird, 1982; Carter, 1996; Hansen & Gauthier, 1989; Zhang et al., 1995, 1997a), promotions including advertising, publicity, special events, 9 and direct mail and notification (Hansen & Gauthier, 1989; Zhang et al., 1995, 1997a; 1999), entertainment/music (Kahle et al., 1996; Zhang et al., 1999), television effects and competition with other sport events (Hill et al., 1982; Zhang et al., 1997b), and team history in a community (Hansen & Gauthier, 1989). While promotions such as advertising, special events and income have been positively related to game attendance, ticket price, television effects, and competition with other sport events have been negatively related to game attendance and enjoyment (Bird, 1982; Hansen & Gauthier, 1989; Hill et al., 1982; Zhang et al., 1995, 1997a, 1997b, 1999). Spectators' socio-demographic profile is a third grouped variable affecting the decision to attend games. This includes age, gender, ethnicity, occupation, education, and size of community. According to Greenstein and Marcum (1981) and Jones (1984), the population size of a community is positively related to game attendance. With respect to age, Zhang and his colleagues (1995) found that younger spectators account for a bigger proportion of the audience for NBA games. The fourth grouped variable is facility convenience, which has included accessibility to facility (Carter, 1996; Greenstein & Marcum, 1985; Hansen & Gauthier, 1989) and stadium quality (Greenstein & Marcum, 1981; Hansen & Gauthier, 1989; Zhang et al., 1995, 1997a). This variable has been found to be positively related to game attendance (Hansen & Gauthier, 1989; Zhang et al., 1995, 1997a). The last variable is schedule convenience, which has typically included the schedule of games and its suitability for spectators (Hansen & Gauthier, 1989; Zhang, 1998; Zhang et al., 1995). This variable has been found to be positively related to game 10 attendance (Greenstein & Marcum, 1981, 1985; Hansen & Gauthier, 1989; Zhang, 1998; Zhang et al., 1995). However, one study done by Zhang et al. (1997a) revealed that game convenience was not related to game attendance levels in minor league hockey. In this study done on minor league hockey, Zhang et al. (1997a) developed what they called the Scale for Attendance Decision (SAD) using five factors: Game Attributes, (speed of game, tough/aggressive play, player appeal, and love of hockey), Home Team (home team quality, star player(s), win/loss records, history, and league standing), Economic Considerations (ticket price, good seats, promotion, and direct mail and notice), Opposing Team and Game Convenience. Spectator demographic variables (including age, gender, and ethnicity) and spectator attendance frequencies were investigated. According to Zhang et al. (1997a), the number of games attended during the season is positively related to Game Attributes, Home Team, and Economic Considerations. However, Game Convenience and Opposing Team were not found to be related to game attendance. Even though previous studies have provided relevant information for developing marketing strategies based on spectator expectations and needs, there are weaknesses in these studies. First of all, in most cases the sample size has been too small to adequately represent the population being studied. For example, the sample sizes in previous studies (i.e., Carter (1996), N = 251; Hansen & Gauthier (1989), N= 54; Smith (1995), N = 31; Zhang et al. (1997a), #= 181, (2001), N= 257) on hockey were too small to have population-level significance. Although Jones' study (1984) contained a large data set (annual observations for all teams in the six team league that operated from 1946 to 11 1976, and game observations for all 18 teams and 720-games during the 1977-1978 NHL season) and provided important variables affecting professional hockey game attendance (i.e., win/loss record of home and visiting team, game schedule, superstars, game style (fighting), game played or team located in a Canadian city, market competition for other sports, population size, income, and ethnicity), the study is now dated, having been done in 1984 on data that is now quite old (1946-1978). It is not known whether Jones' findings are still applicable to the rapidly changing current market. Zhang (1998) and Zhang et al. (1997b, 1999) had a large sample size (N= 2,225) and had meaningful findings, however, they focused on minor league hockey games and it is unclear how relevant these findings are to major league hockey games (NHL). Second, it is critical to fully assess the validity of any questionnaires used in survey research, yet this has rarely been done in previous research. For instance, Hansen and Gauthier (1989) and Smith (1995) did not provide any evidence of having conducted validity tests on their respective questionnaires. Zhang et al. (1997a) were concerned about the validity of their instrument, but evaluated validity only on a limited basis with respect to content and internal consistency. Thirdly, confirmatory factor analysis is the recommended statistical method for uncovering factors in survey research, yet an exploratory factor analysis was used instead in most previous studies of hockey discussed above (Carter, 1996; Hansen & Gauthier, 1989; Zhang et al., 1997a, 1998). Due to the limitations of exploratory factor analysis, variables containing different meanings could have been loaded into the same construct. For instance, there are seven items including evening games, unobstructed view, 12 weekend games, offensive of home team, fourth quarter of season, third quarter of season, and years franchise in arena listed under the 'scheduling of games' factor in the Hansen and Gauthier (1989) study. However, only four (evening games, weekend games, fourth quarter of season, and third quarter of season) of the seven items were actually related to the 'scheduling of games' factor. If items containing different meanings are inadvertently placed in the same construct, the construct ends up being less reliable for measuring objective values which affects the validity of the whole study. Fourthly, although previous research including the studies by Carter (1996), Smith (1995), Hansen and Gauthier (1989), and Zhang et al. (1997) provided valuable insights into spectators' characteristics based on their demographic information, there are very few studies that have analyzed spectators' profiles based on their psychographic information. Uncovering additional information about spectators based on their lifestyles, interests and opinions and linking this to socio-demographic information could enhance understanding of the kinds of people who are sports fans and why and how they attend games. Lastly, although several studies (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989; Zhang et al. 1997a) assessed variables affecting hockey game attendance, it is unclear to what extent these findings can be applied to the case of NHL spectators. Zhang et al. (1997a) insist that because Greenstein and Marcum (1985) and Hansen and Gauthier (1989) studied multiple sports, they missed some of the unique characteristics of hockey (e.g., aggressive play and speed of game) and as a result their findings are not sufficient in themselves to explain hockey game attendance. As well, 13 findings of the studies of Zhang (1998) and Zhang et al (1997a, 1999) may not applicable to the N H L in that their studies were focused on minor league hockey. To understand N H L attendance, it is necessary to test the variables used in these previous studies and to identify key variables that affect game attendance in the N H L , particularly those that lead current spectators to continue attending games and those that attract new spectators. Additional Key Variables Building on the work of Zhang et al. (1997a) and previous studies (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989), three additional variables need to be considered with respect to N H L spectator attendance. Advertising Media Although researchers like Zhang and his colleagues used advertising as one of the game promotion variables, it was too general to be discriminating. According to Solomon (1996), a marketer might use different advertising media depending on levels of involvement in the product, in this case, games. Involvement refers to the level of perceived personal importance and/or interest evoked by a stimulus (or stimuli) within a specific situation. When consumers are intent on doing what they can to satisfy a need, they will be motivated to pay attention to and process any information felt to be relevant to achieving their goals. Consumption at the low end of involvement is characterized by inertia, where decisions are made out of habit because the consumer lacks the motivation to consider alternatives. At the high end of involvement, we can expect to find the type of passionate intensity reserved for people and objects that carry great meaning to the individual. (Solomon, 1996, p.146-147) Television is generally considered a low-involvement medium because it requires a passive viewer who exerts relatively little control over content. In contrast, print media demand high involvement. A marketer might use television as an advertising tool for 14 low-involvement spectators such as single-game ticket purchasers, and print media such as magazines and newspapers as media to attract high-involvement spectators such as season ticket holders or ten-game ticket holders. Another variable that should be included is spectators' dwelling information. Dwelling Information Geographic segmentation is useful when true differences in preference and purchase patterns for a product emerge along regional lines (Beckman, Kurtz, & Boone, 1992). For example, a marketer can use different strategies (e.g., different resource allocations for advertising) from region to region, if there are differences between regions and if the number of spectators varies along regional lines (Beckman, Kurtz, & Boone, 1992). Psychographic Information Although the lifestyle concept was developed nearly forty years ago by researchers in marketing (Laser, 1963; Levy, 1963; Moore, 1963; Plummer, 1971), application of the concept to sports marketing is still relatively new, and has real potential for enhancing understanding of sports spectating and for enabling more effective segmentation of sport spectators. The delayed adoption of lifestyle marketing is not entirely unique to sports marketing. According to Beckman, Kurtz and Boone (1992), socio-demographic segmentation remains a primary basis for grouping customers into target markets in many industries, and as a consequence, lifestyle segmentation is as yet an under-utilized resource. It is important to distinguish these two forms of segmentation. Demographic segmentation consists of dividing the market into groups on 15 the basis of socio-demographic variables. For conducting marketing research, the demographic profile is probably the most familiar and uses accessible information. Demographic information such as age, income, education, sex, occupation, religion, race, nationality, and household size are the most popular bases for distinguishing customer groups and developing appropriate marketing mixes in marketing research. According to Wells & Tigert (1971), psychographic research (lifestyle research) focuses on consumers' activities, interests, prejudices, and opinions, and its major aim is to draw human portraits of consumers. Lifestyle refers to a person's pattern of living as expressed in his or her major activities, interest, and opinions (AIO). Activities refer to work, hobbies, shopping, sports, and social events. Interests refer to food, fashion, family, and recreation. Opinions concern people themselves, social issues, business, and products. Beckman, Kurtz, and Boone (1992) defined psychographics as the use of psychological attributes, lifestyles, and attitudes to determine the behavioural profiles of different consumers. In a recent study related to NHL spectators, Carter (1996) segmented Toronto professional sports consumers into three groups based on their psychographic information. There were two main purposes in Carter's study. The first was to determine if naturally occurring psychographic segments were present and the second was to segment professional sport consumers based on their psychographic information including needs, wants, attitudes, interests, opinions, behaviours and demographics. For the purpose of his research, Carter collected and examined 251 questionnaires completed 16 by present and potential consumers of four different professional sports (football, baseball, basketball and hockey) from the Toronto area. With respect to market segmentation, Carter (1996) found that there were three different segments (Strongly Committed, Weakly Committed and Non-Professional Sports Fans) among Toronto professional sports consumers. Carter performed three separate exploratory factor analyses to identify thirteen factors with a total of 45 different sub-items. Based on the thirteen identified factors, the three different segments were derived using cluster analysis. Chi-square analyses were used to reveal the relationships of the three segments to their game attendance level. According to the results for the case of the Toronto Maple Leafs, the Strongly Committed group was more likely to attend professional sporting events while the Non-Professional Sports Fan group was less likely to attend sporting events. Even though Carter's study (1996) provided very important information for understanding professional sports market segmentation based on psychographics, there were a number of limitations to his study. Firstly, the sample size was too small (N = 251) to evaluate four different professional sport consumers. Secondly, due to the small sample size, Carter ended up performing three separate exploratory factor analyses which overall resulted in 13 factors. While these results were useful, the factor loadings probably would have been different if he had conducted a single factor analysis with all 45 items from the questionnaire and a larger sample size. As a consequence of using exploratory factor analysis (instead of confirmatory factor analysis), variables containing different meanings were loaded into the same construct. Items that had weak loading or 17 were inconsistent with the other items in the same factors were all excluded in his study. Again, it is unclear how these would have stood up in a larger, single-sport study. Lastly, Carter evaluated the validity of his questionnaire on only a very limited basis in terms of content validity and so other conditions of validity were overlooked. As noted by Beckman, Kurtz, and Boone (1992), demographic variables tend to be used more frequently in market segmentation than psychographic variables. There are several reasons for this. First, they are easy to identify and measure. Compared to psychographics, demographic information is easier to collect and measure because demographic questions are more straightforward to answer than are the interests and opinion questions required for psychographic information. As well, marketers can conduct research drawing on vast quantities of secondary data, such as census data, which can assist them in segmenting potential markets on a demographic basis. Marketing researchers collect socio-demographic information as a matter of routine and marketers feel comfortable using this information. Second, demographic variables are associated with the sale of many products and services. Third, they are typically referred to in describing the audiences of advertising media, so that media buyers and others can easily pinpoint the desired target market (Beckman, Kurtz, & Boone, 1992). Socio-demographic information is unquestionably a useful basis on which to segment consumers. Wells (1988) insisted that demographic information has so much influence on so many kinds of consumer behaviour that users of a product or a brand, viewers of a television program, or readers of a magazine are virtually certain to differ from the rest of the population on one or more of the common demographic dimensions. 18 Gender is an obvious variable for segmenting many markets, since many products are marketed as being gender-specific. Age is another important demographic variable to segment markets. Generally speaking, people's spending patterns tend to vary according to age. For example, suppose there is a company that is producing inline skates and snowboards and the company needs to decide its target market. In this case, even though the older and senior middle-aged adults have relatively more income or money than younger adults have, they are not the company's target group because their products generally do not appeal to this age group. Despite the importance of demographic factors, however, Solomon (1996) has noted that people tend to identify themselves with different lifestyle groups based on activities they prefer (psychographics), in addition to having different age-appropriate interests and spending patterns (demographics). Consumers often choose products, services, and activities over others because they are associated with a certain lifestyle or a lifestyle group (Beckman, Kurtz, & Boone, 1992). For this reason, lifestyle-marketing strategies attempt to position a product by fitting it into an existing pattern of consumption. Because a goal of lifestyle marketing is to allow consumers to pursue their chosen ways to enjoy their lives and express their social identities, a key aspect of this strategy is to focus on product usage in desirable social settings (Solomon, 1996, p. 579- 580). Lifestyle marketing does not focus its analysis on the characteristics of consumers as such, but rather it focuses on understanding consumers based on their patterns of behaviour. For example, Beckman, Kurtz, and Boone (1992) explain that while demographics allow us to describe who buys, psychographics allow us to understand why they buy. The obvious conclusion is to use both methods to profile 19 consumers so that we can understand who buys a product as well as why they buy it. This is the approach taken in this dissertation. As can be seen from the discussion above, even though the theory of reasoned action and means-end theory are not formally evaluated in this dissertation, the two theories provide an important basis for understanding factors that influence ticket purchase decisions as well as game attendance. Their role in the dissertation is to help clarify the varieties of factors that influence game attendance and to provide an overarching perspective (broader than in most previous studies) on how these factors relate to one another. The research in this dissertation is focused on uncovering, inter alia, factors affecting ticket purchases and spectators' attitudes towards ticket purchases. In this respect, "attitude toward buying", which is one of three core concepts of the theory of reasoned action (the other two: "social pressure" and "intentions versus behaviour"), plays an important, albeit implicit, role to help evaluate game attribute items in the dissertation research. By comparison, because means-ends theory focuses on the cognitive linkages among products, consequences, and personal values, its main contribution is for understanding how spectators' attitudes and opinions relate to their behaviours and for taking account of the emotive, affective side of spectator attendance at games. The theory of reasoned action is used mainly in this dissertation as a framework for isolating cognitive factors and variables in previous research that can be seen to influence the purchase decision in buying tickets and attending games. Means- ends theory is used to help identify lifestyle and affective variables. Peoples' activity patterns and choices are rooted in personal values even as much as are their attitudes and 20 opinions. Although both the theory of reasoned action and means-end theory are effective tools to help reveal construct domains, this dissertation research was not intended to evaluate these two theories but to use them as a basis for developing and evaluating an instrument for testing extant understandings of factors affecting game attendance and for uncovering spectators' demographic and psychographic profiles. Having reviewed previous research in the areas of game attendance and spectator profiling, I turn in the next chapter to how I designed a questionnaire to 1) evaluate the efficacy of present methods of assessing game attendance factors and 2) describe the spectator profiles of NHL game attendees at Vancouver Canucks games in Vancouver, Canada. 21 Chapter 3. Methodology The main purpose of this dissertation research was to investigate the consumer profiles of NHL spectators and to identify spectator profile factors relating to game attendance in the NHL. The research had two inter-related objectives. One was to investigate the consumer profiles of sport spectators associated with a specific NHL franchise, the Vancouver Canucks, and to relate these to game attendance. The second objective was to develop, test and evaluate a questionnaire designed to incorporate latest thinking on variables that affect professional sports game attendance. The study was intended to make a novel contribution to the literature as little research had previously focused on profiling NHL fans as such, and no research had been undertaken to holistically test the different attendance-related variables identified in prior research using confirmatory factor analysis. The procedures followed in the evaluation of this questionnaire are discussed under the following headings: (a) subjects, (b) overview of instrument design and validity, (c) defining constructs and selecting items, and (d) data collection and data analysis. Subjects For the purpose of this study, three different groups of subjects were required for the three different components of the study (a pilot test, main tests, and a reliability test). For the main tests, actual Vancouver Canucks spectators who were attending a live game formed the target population. As well, all subjects needed to be 19 years of age or older for the main test to comply with UBC ethics standards and to represent the core target group for Canucks spectators. For the pilot and reliability tests, students from the 22 University of British Columbia were targeted on a one-off basis so that students in the pilot test would not participate in the reliability test. Overview of Instrument Design and Validity The instrument was designed to evaluate the most promising and theoretically compelling variables identified in Chapter 2, and has four discrete parts (see Table 1). Table 1. Parts and Components Part Component Part 1 Reasons for Game Attendance (Number of Items = 33) Overall performance, uniqueness of hockey, facility convenience, game schedule, game promotion, advertising, sense of community Part 2 Sports Interests and Activities (NI = 20) Ticket information, frequency of game attendance, ticket purchasing pattern, activities (work, hobbies, sports, and social events), interests (family, media, and recreation) Part 3 Personal Background (NI = 9) Sex, age, education, occupation, income, ethnicity, and dwelling information Part 4 Opinions About Sports (NI = 33) Opinions (about themselves, social issues, politics, economics, education, culture, future, business, and products) under 10 constructs: self-participation in sports, cynicism, children oriented, perceived time cost, sports benefits, resources, community pride, attitudes about professional athletes, government role, and attitude about amateur sports Note. The complete questionnaire is presented in Appendix A. The first part of the questionnaire consists of questions to assess the reasons why NHL spectators attend games. It contains 33 items under seven constructs (Overall Performance, Uniqueness of Hockey, Facility Convenience, Game Schedule, Game Promotion, Advertising, and Sense of Community). The second part is related to 23 psychographic variables including spectators' sport interests and activities, and consists of 20 items. The third part is about spectators' personal backgrounds based on nine demographic characteristics including gender, age, education, occupation, income, ethnicity, and dwelling information. In the final part of the instrument, further psychographic variables relating to opinions about sports are examined using 33 items under 10 constructs: Self-participation in Sports, Cynicism, Children Oriented, Perceived Time cost, Sports Benefits, Resources, Community Pride, Attitudes about Professional athletes, Government Role, and Attitude about Amateur sports. As mentioned in the previous chapter, because psychographic information is an effective tool for understanding consumers, it is important to investigate spectators' psychographic profiles in addition to their demographic profiles in order to better understand spectator behaviour and potentially enhance marketing strategies in the National Hockey League. A key contribution of this research is that it fully assesses the validity of the questionnaire, including constructs and variables used in previous studies in which validity was not conclusively evaluated. The term 'validity' is used differently by different authors and in different contexts. According to Pedhazur and Schmelkin (1991), the meaning of validity in a measurement context differs from that in research design contexts. As well, within each of these contexts, one encounters different definitions of validity and distinctions among different types or kinds of validity. The American Psychological Association (AERA/APA/NCME, 1985, cited by Johnson and Plake, 1998) offers a characterization of validity for educational and 24 psychological measures, that incorporates the appropriateness, meaningfulness, and usefulness of the specific inferences made from test scores. It will be noted that this characterization of validity refers not to a measure itself but to inferences made on the basis of scores obtained on that measure. Messick (1989) characterized validity in a similar manner, making validity a feature of the meaning of the scores rather than a condition of the test or assessment as such. In other words, he emphasized the meanings, interpretations, and applications of the test scores, rather than test validity itself. Messick (1995) insists that validity should be a unified concept to address both score meaning and social values in test interpretation and test use. This comprehensive concept of validity made an important shift from content, criteria, and consequences into a construct framework to investigate rational hypotheses about score meanings and theoretically consistent relationships. Content, substantive, structural, generalizability, external, and consequential construct validity are six aspects of construct validity incorporated in the new unified concept of validity presented by Messick (1995). Figure 1. Steps in Questionnaire Validity Test (Schutz, 1998) [Define construct & Create and/or select items] I Review by experts for content validity, Revise i Pilot on small sample for clarity and ambiguities I Pilot on large sample for internal consistency, Revise I Pilot on large sample for convergent and discriminant validity, Revise i Pilot on large samples of divergent groups for concurrent validity 2 5 In this study, construct validity is examined as part of evaluating the questionnaire for the end purpose of assessing the consumer profiles of Vancouver Canucks' spectators. For testing an inventory in this study, there are six steps based on Schutz's inventory development procedure (1998) (Figure 1). Defining Constructs and Selecting Items As noted in the previous chapter, theory of reasoned action and means-ends theory were used respectively in this dissertation research to help conceptualize and identify cognitive and affective factors affecting N H L game attendance. The goal was to develop an inventory that could be used to profde N H L spectators and evaluate current understandings of factors in spectator attendance. A number of factors had been identified in previous research. These are critically reviewed below and the final selection of items is described. Game Attendance In previous studies (Baade & Tiehen, 1990; Bird, 1982; Carter, 1996; Dick & Sack, 2003; Hansen & Gauthier, 1989; Jones, 1984; Kahle et al., 1996; Marcum & Greenstein, 1981, 1985; Mawson & Coan, 1994; Schofield, 1983; Whitney, 1988; Zak, Huang, & Siegfried, 1979; Zhang et al., 1995, 1996, 1997a, 1997b, 1998, 1999, 2001), a number of factors were found to affect spectator attendance as described in the literature review section. Among the previous studies, Carter (1996), Hansen and Gauthier (1989), Jones (1984), Smith (1995), Schofield (1983), Zhang (1998), and Zhang et al. (1996, 1997a, 1999) have examined hockey attendance and focused on identifying variables which can affect the attendance level of spectators. Carter identified (1996) 10 26 variables (education, income, sports knowledge, participation in team sports, amount of television sports watched, too much time to get there, want to see specific teams, busy doing other things, ticket price, and a good opportunity to spend time with kids) that affect spectator attendance. Schofield (1983) examined 17 studies, and developed four categories (economic, demographic, game attractiveness, and general preferences) to explain spectator attendance factors. Jones (1984) insisted that winning related variables (home & opposing team) and market characteristics (scheduling, population size, minority composition, income, market competition, arena capacity, and Canadian tradition) were uncovered to affect attendance levels. Smith (1995) focused his study on both major and minor league hockey surveying public relations directors, and revealed eight attendance related variables: fast, action-packed, education, competitors/rivalries, fight potential, giveaways, gimmicks, tie-ins, and individual players. Zhang et al. (1997a, 1999) identified 13 variables under three factors (game attribute, home team, and economic consideration) and nine game enjoyment variables affecting the spectator decision to attend minor league hockey games. As well, Hansen and Gauthier (1989) uncovered 29 (out of 40) separate variables that influenced attendance at professional Table 2. Constructs and Items in Hockey-Related Studies Carter (1996) Schofield (1983) • education Economics • household income •ticket price • sports knowledge vthe existence of substitute • participation in team sports forms of entertainment • amount of television sports watched Demographics • too much time to get there • population size • only want to go to certain games to see specific • ethnic mix teams Game attractiveness • busy doing other things * team success A tickets too expensive * star players • offers a good opportunity to spend time with kids * closeness of competition 27 •special events General preferences • new stadium V weekend game vthe end of season Jones (1984) Smith (1995) * home and opposing team ©fast, action-packed V scheduling • education • population size * competitors/rivalries • income ©fight potential • minority composition • giveaways • market competition • gimmicks • arena capacity • tie-ins * Canadian tradition * individual players Zhang et. al (1997a, 1999) Hansen & Gauthier (1989) 1997a Scheduling of games Game attributes ©offensive output of home games © speed of game • unobstructed view © tough/aggressive play ¥ evening games © player appeal V weekend games * love of hockey V fourth quarter of season Home team V third quarter of season * win/loss records •years franchise in area * league standing Convenience of fans * star players • easy access to facility * home team quality • availability of parking Economic consideration • seating capacity of facility • ticket price • cleanliness of facility • good seats vTV coverage of another major sport during home * promotion game #>direct mail & notification • ethnic mix of population 1999 • average income of population * music Performance quality of the team © speed of game * record breaking performance of visitor athletes ©player appeal * record breaking performance of home athletes * entertainment * rivalry of team and opponent * people watching ©defensive output of home team ©tough/aggressive play ©player fighting * promotions * socializing with friends •special event occasions Team as a contender * team's place in division standings * team's place in league standings * team's involvement in 1st place race This study * Overall performance © Uniqueness of hockey • Facility convenience V Schedule convenience + team's involvement in playoff race * closeness of competition during season * record of visiting team • record of home team Price • price of season ticket • price of ticket to home game • Game promotion #> Advertising * Sense of community 28 sport events. Twenty-eight of the 29 variables loaded on the five constructs: scheduling, convenience, quality of performance, team as a contender, and price. Table 2 lists the constructs, and specific variables that may measure these constructs, for all of these studies. Although the names of the constructs from the five hockey-related studies were somewhat different, the implied meanings of the constructs were similar throughout the five studies and relate to the spectators' overall "attitude towards buying" tickets as well as to their personal values and affect. Based on the findings of these studies and in conjunction with studies in other professional sporting events (e.g., baseball, football, basketball), these following seven constructs are deemed to be the most important factors affecting hockey attendance: Overall Performance, Uniqueness of Hockey, Facility Convenience, Schedule Convenience, Game Promotion, Advertising, and Sense of Community (Figure 2). The seven constructs and their associated items, along with the rationale for their inclusion, are discussed in the following section. Table 3 provides a complete listing of the seven constructs and 33 items. Overall performance. The first construct, overall performance, encompasses eight items related to the general quality of games. Factors in this construct are related to performance quality (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989; Zhang et al., 1995, 1997a; 1999), star players/superstars (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989; Smith, 1995; Zhang, et al., 1995, 1997a), the team as a contender (Bird, 1982; Hansen & 29 Table 3. Constructs and Items for Part 1 Constructs Items Overall Performance (0. P.) (Number of Items = 8) • home team win/loss records • visiting teams win/loss records • home team superstars • visiting team superstars • record breaking performance of visitor's athletes • record breaking performance of home athletes • rivalry of team and opponent • closeness of competition (home vs. visiting teams) Uniqueness of Hockey (U. H.) (NI = 3) • offensive output of home games • defensive output of home team • frequent fights and penalties Facility Convenience (F. C.) (NI = 6) • unobstructed view • comfortable seat • easy accessibility to facility • availability of parking • seating capacity of facility • cleanliness of facility Game Schedule (G S.) (NI = 2) • game time (afternoon, evening) • game day (weekday, weekend) Game Promotion (G P.) (NI = 5) • giveaway/prize • ticket discount • special event occasions • price of season ticket • price of game ticket Advertising (ADVT) (NI = 6) • direct mail & notification • advertising through TV. • advertising through newspapers • advertising through magazines • advertising through radio • advertising through internet (websites) Sense of Community (COMM) (NI = 3) • behaviour of fans during games • accompanied by friends • excitement Gauthier, 1989; Zhang et al, 1995, 1997a.), and the closeness of competition during season (Hansen & Gauthier, 1989). This construct includes the following items: home- team win/loss record, visiting-team win/loss record, home-team superstars, visiting-team superstars, home-team record breaking performance, visiting-team record breaking performance, rivalry of team and opponent, and closeness of competition during season (home and visiting teams). Uniqueness of hockey. The second construct of the hockey attendance component of the questionnaire includes variables related to the unique characteristics of professional hockey games. The three items belonging to this construct are offensive output (speed of game, and minutes of power play), defensive skills, and frequent fights and penalties (Smith, 1995; Zhang etal., 1997a, 1999). Facility convenience. The third construct, defined as facility convenience, contains six items focusing on the convenience of fans. Items in this construct are: unobstructed view (Hansen & Gauthier, 1989; Zhang et al., 1995), good seat (Zhang, 1995, 1997a), easy accessibility to facility (Greenstein & Marcum, 1981; Hansen & Gauthier, 1989), seating capacity of facility (Jones, 1984; Greenstein & Marcum, 1985), availability of parking (Hansen & Gauthier, 1989; Zhang et al., 1995), and cleanliness of facility (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989). Schedule convenience. The fourth construct represents variables related to game schedule. Two items in this construct are related to scheduling of game times (afternoon and evening) and game day (weekday and weekend) (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989; Schofield, 1983; Zhang, 1998; Zhang et. al, 1995, 1997a). These variables have been shown to influence spectator attendance. 31 Game promotion. The fifth construct represents items related to game promotion. The five items in this construct are giveaway/prize (Smith, 1995; Zhang et al., 1995), special events (Hansen & Gauthier, 1989; Schofield, 1983), price of season ticket, price of home game ticket (Bird, 1982; Carter, 1996; Hansen & Gauthier; Schofield, 1983; Zhang et al., 1995, 1997a), and ticket discount (Zhang et al., 1995), all of which were verified to affect spectator attendance. Advertising. The sixth construct encompasses six items including direct mail and notification (Zhang et al., 1995, 1997a), advertising through T.V., radio, newspaper, magazines, and internet (Mawson & Coan, 1994). The items in this construct are intended to test spectators' game involvement based on advertising media affecting their game attendance. Sense of community. The final construct consists of three items including behaviour of fans during games, being accompanied by friends, and excitement. According to Kahle et al. (1996) some spectators have a strong emotional attachment to and excitement in their teams and this attachment possibly affects their team's performance. Zhang et al. (1999) identified two emotional items (socializing with friends and people watching) that affected spectators' game enjoyment of minor league hockey games. Means-ends theory supports including these items within a single construct because they reflect an end value of sociality that is informed by communal identity. The items demonstrate a concern for community that is both serious (supporting the team) and playful (socializing with 32 friends and people watching). Game attendance is a means for engaging in these activities and fulfilling this value. Sports Interests and Activities In addition to the seven constructs in part 1, part 2 consists of 20 items related to the spectators' lifestyles including sports interests and activities, hobbies, social events, fan history, and spending patterns for their activities (Table 4). According to Beckman, Kurtz, and Boone (1992), although demographic segmentation traditionally has been the primary basis for grouping customers and industries into segments to serve as target markets, psychographic segmentation focusing on lifestyles of potential consumers may provide more relevant and important information for market segmentation purposes. Carter (1996) insisted that the marketer could appeal to each segment by knowing the characteristics of different psychographic segments and he identified three different professional sport markets using spectators' lifestyle information. In addition to spectators' sports interests and activity patterns, items such as their ticket choices, amount of money spent for tickets, and frequency of attending games are included to investigate spectators' ticket purchasing characteristics along with their game attendance levels. The interests and activities questionnaire items were all designed to be open-ended, except for ticket choices, timing of purchasing tickets, ticket purchasers, best promotional medium, and colleagues for attending games, which were multiple choice items. Personal Background Demographics are statistics that measure observable aspects of a population 33 Table 4. Questionnaire Items for Part 2 and 3 Parts Items Part 2 Sports Interests and Activities (Number of Items = 20) • number of games attended in the past 12 months • amount of money spent for game tickets • number of games attending in the next 12 months • ticket choices (season, 11 pack, single etc..) • timing of purchasing tickets • ticket purchaser • best promotional medium • colleagues for attending games • fan history • memberships of sports clubs • intention to participate in a new physical activity • favourite hobbies • favourite sports activities • amount of money for sports activities • favourite outings other than hockey games • amount of money for outings • favourite web sites • frequencies to visit and amount of time spent for web sites • favourite sports when young • favourite sports of kids Part 3 Personal Background (NI = 9) • age • gender • postal code • first language • ethnic background • business, industry or service to work in • occupation • income • level of education (Solomon, 1996). In previous studies, eight demographic characteristics including age/ gender (Greenstein & Marcum, 1985), population size (Greenstein & Marcum, 1985; Jones, 1984; Schofield, 1983), social status (Greenstein & Marcum, 1985), education (Cater, 1996; Smith, 1995; Zhang et al., 1995), occupation (Zhang et al., 1995), income 34 (Jones. 1984; Hansen & Gauthier, 1989), and ethnicity (Hansen & Gauthier, 1989; Jones, 1984; Schofield, 1983; Zhang et al., 1995) have been revealed to affect spectators' game attendance levels. Part 3 of the questionnaire contains nine items to uncover spectators' demographic information including age, gender, education, occupation, income, ethnicity, and dwelling information (Table 4). Opinions about Sports In this part, thirty-three psychographic items were included. All thirty-three variables were taken from Carter's multi professional sport study done in 1996. Carter identified 14 factors based on 45 variables associated with issues including the role of the family, participating versus spectating, sports avidness, views on types of sports or athletes, sensation seeking, personality, societal treatment of athletes, sense of financial security adding value to the quality of life, and the role of government. The fourteen identified factors are as follows: self-participation in sports, degree of excitement, skill versus genetics, emotion arousing, cynicism, inconclusive, children oriented, perceived time cost sports benefits, resources, community pride, attitudes about professional athletes, government role, and attitude about amateur sports . Among the forty-five items under 14 constructs in total, thirty three variables relevant to hockey were selected to test the relationship between spectators' psychographic information and the frequency of game attendance in this study. The thirty-three items are categorized into ten constructs and the names of constructs are as follows: self- participation in sports, cynicism, children-oriented, sports benefits, resources, community pride, attitudes about professional athletes, government role, attitude about 35 amateur sports, and perceived time cost. Table 5 provides a complete listing of the ten constructs and 33 items. Table 5. Constructs and Items for Part 4 Constructs Items Self-Participation in sports (S.P.) (Number of Items = 6) • Given the choice, participating in sports is more enjoyable than watching. • It is always good to try new recreational activities. • It is difficult to understand how people could just lay around when there are so many recreational activities to do. • Individual sports are superior to team sports like soccer because you don't have to rely on others • Attending a sporting event with a friend is far more enjoyable than attending alone. • It is good to see what sports reporters say about the team before deciding whether or not to buy a ticket. Cynicism (C.C.) (NI = 3) • Buying professional sport memorabilia is a waste of money. • Buying tickets from scalpers is not worth the extra cost, even for excellent seats. • Belonging to professional sports fan clubs is a waste of time. Children Oriented ( C O . ) (NI = 2) • Parents should always try to attend their children/s sport contests. • It is important to participate in regular physical activity with your child. Sports Benefits (S.B.) (NI = 4) • Professional sports provides an escape from my home and family pressures. • Participation in organized sports is not one of the best ways for kids to spend their free time. • Taking children to professional hockey events is a great way to spend time with them. • For children, the positive aspects of participating in sports for outweigh the negative ones. Resources (R.S.) (NI = 2) • For recreation, time is a bigger problem than money. • Our income is high enough to satisfy nearly all our desires for sporting event tickets. Community Pride (CP.) (NI = 5) • It would be difficult to live in a city that does not have one or more professional sports teams nearby. • When my home team wins, it really makes me feel good. • The success of a professional sports team brings pride to the city. • Professional athletes make good role models. • Being able to cheer for my home professional sports 36 franchise adds to my sense of community. Attitude about Pro-athlete (A.P.) (NI = 3) • Most professional athletes show a genuine caring for the city in which they play. • Professional athletes are only in it for the money. • Professional athletes tend to be whine and complain too much. Role of Government (R.G) (NI = 3) • Tax dollars should not be spent on building stadiums and areas for professional sport franchises. • Government should not assist financially troubled professional sports organizations. • Government should offer tax incentives to keep a team in a city or to encourage another team to relocate there. Attitude about Amateur sports (A.A.) (NI = 4) • The enthusiasm of the amateur athlete makes amateur sports more enjoyable to watch than professional sports. • Amateur sport athletes/teams deserve more funding. • Attending an amateur sporting event is not as exhilarating as attending a professional sporting event. • Amateur sporting events are quite often far superior in entertainment value than their professional counterparts. Perceived Time cost (RT.) (NI= 1) • Professional sporting events take too much time to watch. Review by Experts for Content Validity The content aspect of construct validity includes evidence of content relevance, representativeness, and technical quality (Messick, 1989). According to Messick (1995), a key issue for the content aspect of construct validity is the specification of the boundaries of the constructs to be assessed. In this study, the relevant constructs are the seven components of game attendance, and the ten components of opinion about sports. For content validity, four experts (professors and marketers whose specialties are in sport marketing, statistics, and marketing) reviewed the contents of the items hypothesized to measure the content topics. The experts were requested to indicate if an item did not adequately reflect the construct associated with it, and if the item was 37 ambiguous or unclear. For part 1, game attendance, 18 items rated as having low relevance to the constructs were excluded. These included team's involvement in 1st place race, team's place in division standings, and timing (in the season) of the game. Also, items showing similarity with other items were excluded. In part 2, the formats of the items related to favourite hobbies, activities, and outings were changed from multiple choice questions to open-ended questions. For part 3, personal background, the item about marital status was excluded because it was too personal to answer. In parts 2, 3, and 4, the layout and wording of some items underwent minor revisions. It was anticipated that these changes would result in a questionnaire that would take 20 minutes to complete (compared to the 30 minutes requested for the original questionnaire). Pilot on Small Sample for Clarity and Ambiguities Once content validity was established, a pilot study was conducted on a small sample for improving clarity and eliminating ambiguities. The subjects for this pilot test did not have to be spectators. Seventeen students from the University of British Columbia were used. Subjects were asked to review items and to rate the degree of difficulty to answer each item. Ratings were done using the following 5-point scale: 1 (Very Difficult), 2 (Difficult), 3 (Fair), 4 (Easy), and 5 (Very Easy), and mean scores were calculated for each item. The results indicated that none of the items' mean scores were less than 4 (expected mean score). However, about 70 percent of the subjects pointed out that the questionnaire was too long. Based on the results of the pilot test and comments from experts regarding content validity, three different forms of the 38 questionnaire were developed, namely Form A, B, and C (see Table 6). Table 6. Questionnaire Forms and Components Forms Components Form A Part 1: Game Attendance Part 2: Sports Interest and Activities Part 3: Personal Background Form B Part 2: Sports Interest and Activities Part 3: Personal Background Part 4: Opinions about Sports Form C Part 1: Game Attendance Part 2: Sports Interest and Activities Part 3: Personal Background Part 4: Opinions about Sports Part 1 (33 items) and part 4 (33 items) formulated the majority of the questionnaire's 95 items, and the logical choice was to separate them. Additionally, these two sections required two separate, unrelated confirmatory factor analyses and there was less necessity for uncovering the connection between them than among the other parts. Consequently Form A consisted of parts 1, 2, and 3, and Form B of parts 2, 3, and 4. These two forms were deemed to be short enough to be completed by spectators before a live hockey game. Form C included all items from parts 1, 2, 3, and 4 and was used with a university sample (UBC students recruited on campus). Test on Large Sample for Internal Consistency and Internal Structural Validity Two separate confirmatory factor analyses (CFAs) were envisioned for testing the internal consistency and internal structural validity of part 1 (game attendance-related variables), and part 4 (opinions about sports). According to Smith, Schutz, Smoll, and Ptacek, (1995), confirmatory factor analysis should be performed because this allows tests to evaluate the degree to which the structural characteristics of an instrument 39 confirm a hypothesized underlying model. Given that the two scales used in this study had been constructed on sound theoretical grounds with well-defined latent constructs, CFA rather than the more commonly utilized exploratory factor analysis (EFA) is appropriate here. EFA is a non-theory based data-driven procedure which may lead to factor structures unique to a specific sample, and with little or no theoretical justification on generalizability. Although it may be useful in the early stages of scale development in some situations, the constructs of the questionnaires used in this study were sufficiently defined a priori, thus negating the need for EFA. Several models can be tested based on inspection of the inter-factor correlations, modification indices, and normalized residuals. For evaluations of sequential models, several different indicators including the chi- square difference, Q test, root mean square residual (RMSR), root mean square error of approximation (RMSEA), the parsimony goodness-of-fit index (PGFI), and comparative fit index (CFI) can be used to assist in selecting the best model (Smith et al., 1995). A RMSEA of .05 or less indicates that the model based on the sample data represents a close fit to the population, and a value less than .08 indicates a reasonable fit (Joreskog & Sorbom, 1993; cited by Smith et al., 1995). For PGFI, the larger the PGFI, the more parsimonious the model. CFI values of .90 and larger indicates an adequate fit of the model to the data (Smith et al., 1995). Based on the inter-factor correlations, modification indices, and normalized residuals a better fitting model could be uncovered by revisions to the scale such as deleting items and combining subscales into a new one, dividing one into two factors, or deleting some subscales which showed low factor loadings, large residuals, and redundancy. 40 For internal consistency, Cronbach's alphas are used as an indicator. Internal consistency reliability is concerned with the homogeneity of the items comprising a scale (Pedhazur & Schmelkin, 1991). High inter-item correlations suggest that the items are all measuring the same thing. Generally, alphas higher than 0.7 are acceptable (Nunnaly, 1978), and alphas higher than .80 are considered adequate (Schutz, 1995), but alpha is also a function of the number of items in the scale, such that lower values are considered acceptable for scales with few (2-4) items. Test on Large Samples for Convergent and Discriminant Validity Convergent validity refers to a convergence among different methods designed to measure the same construct (Pedhazur & Schmelkin, 1991). Ideally, the correlations between the new inventory and other scales used to measure the same construct should be high. Discriminant validity refers to the distinctiveness of constructs demonstrated by the divergence of methods to measure different constructs (Pedhazur & Schmelkin, 1991). Ideally, the correlations between the new inventory and other scales to measure different constructs should not be high. For testing convergent and discriminant validities of the scale which assesses personal values toward professional sports, the inventory of Zhang's study (1999) was used as his scale included both similar and dissimilar constructs in part 1 of this study. Reliability refers to the degree to which measures are free from random error and therefore yield consistent results (Zikmund, 1997). In other words, reliability is the stability of a test score, that is, the ability of a test to provide a consistent measurement on repeated occasions. For testing reliability of part 1 and 4 of Form A, B, and C, 28 41 subjects were used in a test-retest design. To avoid carry-over effects, the interval between the two administrations was two weeks. Bivariate correlation coefficient were calculated for each construct between two tests and used as a reliability index. Test on a Large Sample for Criterion-Related Validity A criterion is any variable one wishes to explain or predict by resorting to information from other variables, and the correlation coefficient between a predictor and a criterion is referred to as a validity coefficient (Pedhazur & Schmelkin, 1991). Criterion validity has two subdivisions depending on when the criterion data is collected, concurrent validity and predictive validity. One way to test criterion-related validity of an inventory which assesses variables affecting spectator attendance of professional sport is to analyze correlation coefficients between predictors and criterion information such as people's ticket preferences (season ticket holder, game pack holder, and single game holder) or the frequency (how many times last season, this season, and next season) of watching games as spectators. In this study, the frequency of game attendance was used as a criterion variable, and the seven constructs in part 1 and ten constructs in part 4 used as the predictors. Data Collection and Data Analysis Collection of Data A group-administered data collection method was used. For the selection of the subjects, spectators who were attending Canucks' games and UBC students who had attended Canucks' games at least once in the past were contacted. The following statement was used to introduce the survey and was printed on the questionnaire: 42 Dear Madame/Sir: The purpose of the survey is to describe the activities and interests of hockey fans and poll opinions about why people attend NHL games. You have been randomly selected as a Canucks spectator. If you are an adult (19 years of age or older) and willing to participate in the survey, you will need to fill out as much of the questionnaire as you can. Written permission was obtained from the Vancouver Canucks to conduct in- stadium data collection at General Motors Place prior to the start of live games. Surveys were conducted prior to four weekday games during the 2001-02 season. Approximately 15 students assisted with the data collection for each game. Distributing and collecting questionnaires were done during the 50 minutes before each game. There are 56 different seating sections in the Canuck's ice rink, but access to about 10 seating sections was restricted by the Canucks organization for various reasons including security problems and conducting a survey of their own. Each time, 15 sections were randomly selected and an average of 350 questionnaires was distributed. With respect to data collection for Form C, two hundred and fifty questionnaires were distributed to UBC students at the Student Union Building for four days during the hockey season in 2002. Data Analysis Responses to the questions were entered into a data file using the Statistical Package for Social Sciences (SPSS for Windows). The data were divided into three different data sets based on the questionnaire form completed (A, B, or C), and then analyzed in two studies, based on the purpose of the data analysis. Table 7 outlines these divisions. Confirmatory factor analyses, cluster analysis, discriminant function analysis, and decision tree analysis were conducted using LISREL 8.3, SPSS" 10.0 for Windows, 43 and SPSSX Answer Tree 3.1. All confirmatory factor analyses (CFAs) were conducted with LISREL. Table 7. Data Sets and Component Data Set Respondents Form Parts 1 Ice Rink Spectators A 1,2,3 2 Ice Rink Spectators B 2,3,4 3 UBC Students C 1,2,3,4 2, 3 (Common components) Study 1 Consisted of Data Sets 1 and 3 Study 2 Consisted of Data Sets 2 and 3 Study 1. Study 1, based on data sets 1 and 3, had as its purpose the investigation of reasons for game attendance and relevant predictors. Data analyses consisted of a confirmatory factor analysis for testing the validity of part 1 of the questionnaire (modelled in Figure 2), and a decision tree analysis for classifying spectators based on all items (part 1, 2, and 3). Study 2. In study 2, using data sets 2 and 3, a confirmatory factor analysis for evaluation of part 4 was performed. The purpose of study 2 was to uncover the connections between spectators' opinions and game attendance, as well as to identify the potential predictors for game attendance levels. In addition to a confirmatory factor analysis, cluster analysis for classifying spectators based on identified factor scores, principle component analysis for revealing directions of clusters, and chi-square analyses for uncovering spectators' profiles based on relationship between clusters and the remaining items (part 2 and 3) were conducted. 44 45 Chapter 4. Results Descriptive Statistics Response Rate One thousand four hundred questionnaires in total were distributed to spectators in the arena, and 1,172 (Form A: 560 and Form B: 612) questionnaires were returned, for a return rate of 83.7 percent. Two hundred and sixty-five questionnaires were excluded from analysis due to excessive missing data (more than 10 items missing). In the end, 907 (425 Form As and 482 Form Bs) questionnaires were used for further data analyses. Two hundred and fifty Form C questionnaires in total were passed out to UBC students during four days of data collection at the Student Union Building and 215 questionnaires were collected for a return rate of 86 percent. Twelve questionnaires were excluded because of more than ten unanswered items and 203 questionnaires were put into a data file. In total, 1,110 (Form A: 425, Form B: 482, & Form C: 203) questionnaires were available for the analyses of this study (Table 8). Table 8. Data Sets and Components Data Set Respondents N Form Part & Components 1 Ice Rink Spectators 425 A 1,2,3 2 Ice Rink Spectators 482 B 2,3,4 3 UBC Students 203 C 1,2,3,4 Total 1,110 2, 3 (Common components) Study 1 Consisted of Data Sets 1 and 3 (N = 628) Study 2 Consisted of Data Sets 2 and 3 (N = 685) Demographics Spectators' demographic information is presented in Table 9. The age of the subjects ranged from 18 to 85 years (M= 32.42; SD = 11.94). Although the target 46 subjects of this study were people who were 19 years or older, some university students were 18 years old at the time of data collection. The majority consisted of subjects in their twenties and thirties (59.3 %), with 718 male (68.1 %) and 337 female subjects (31.9 %) participating in this study. Table 9. Descriptive Statistics for Demographic Variables Variable N % Valid Cumulative % Age Less than 21 178 16.0 16.9 21 - 2 5 199 17.9 35.8 2 6 - 3 0 178 16.0 52.7 31 - 4 0 248 22.3 76.2 41 - 5 0 153 13.8 90.7 More than 50 99 8.8 100.0 Missing 55 5.0 Total 1,110 100.0 Gender Male 718 64.7 68.1 Female 337 30.4 100 Missing 55 5.0 Total 1,110 100 Residence Vancouver 252 22.7 22.7 Burnaby 60 5.4 28.1 Richmond 60 5.4 33.5 West Vancouver 28 2.5 36.0 North Vancouver 60 5.4 41.4 Coquitlam 32 2.9 44.3 Port Coquitlam 18 1.6 45.9 New Westminster 13 1.2 47.1 Surrey 76 6.8 54.0 Langley 30 2.7 56.7 Delta 53 4.8 61.5 Maple Ridge 22 2.0 63.5 Abbotsford 26 2.3 65.8 Port Moody 8 0.7 66.5 White Rock 9 0.8 67.3 Others 134 12.1 79.7 Missing 225 20.3 100.0 Total 1,110 100.0 Ethnicity Caucasian 860 77.5 82.2 African Canadian 6 0.5 82.8 Asian 166 15.0 98.7 Native 14 1.3 100 Missing 64 5.8 Total 1,110 100 Occupation Management 167 15.0 16.1 Social Science etc. 69 6.2 22.8 Business 133 12.0 35.7 Sales & Services 141 12.7 49.3 47 Applied Sciences 67 6.0 55.7 Manufacturing 16 1.4 57.3 Retired 27 2.4 59.9 Trades & Transportation 51 4.6 64.8 Primary Industry 11 1.0 65.9 Student 272 24.5 92.2 Unemployed 26 2.3 94.7 Art, Culture, Sport 31 2.8 97.7 Health related 24 2.2 100.0 Missing 75 6.8 Total 1,110 100.0 Income Less than C$ 10,000 84 7.6 8.6 C$ 10,000- 19,999 57 5.1 14.5 C$ 20,000-29,999 74 6.7 22.1 C$ 30,000 - 39,999 77 6.9 30.0 C$ 40,000 - 59,999 181 16.3 48.6 C$ 60,000 - 79,999 166 15.0 65.7 C$ 80,000 - 99,999 118 10.6 77.8 More than C$ 100,000 216 19.5 100.0 Missing 137 12.3 Total 1,110 100.0 Education Some high school 32 2.9 3.1 High school diploma 218 19.6 24.3 Some college 295 26.6 53.0 Two-year college diploma 125 11.3 65.1 Bachelor's degree 235 21.2 87.9 Some graduate school 49 4.4 92.7 Master's degree 59 5.3 98.4 Doctoral degree 16 1.4 100.0 Missing 81 7.3 Total 1,110 100.0 With respect to dwelling area, 252 spectators (22.7 %) replied that they lived in Vancouver and the rest as follows: 6.8 % in Surrey, 5.4 % each in Burnaby, Richmond, and North Vancouver, 4.8 % in Delta, 12.1 % in areas including Whistler, Mission, Victoria, Nanaimo, and other B.C. areas. With respect to ethnicity, 77.5 %, 15.0 %, 1.3 %, and 0.5 % answered that they were Caucasians, Asians, Natives, and African Canadians (self-identified) respectively. Management (15.0 %), business and finance (12.0 %), and sales and service (12.7 %) related occupations were the three most common occupations of subjects in this study. About 50 percent (n = 521) of subjects replied that their annual household income exceeded C$ 40,000 and 19.5 % earned more than C$ 100,000 annually. Table 10 presents the frequency for spectators' game attendance. Table 10. Descriptive Statistics for Attendance Level Variable mean median SD N % Valid Cumulative % Last 12 month (total) 5.0 2.0 7.5 1,110 100.0 100.0 0 — — 50 4.5 4.5 1 ~ — 336 30.3 34.8 2 ~ ~ ~ 197 17.7 52.5 3 - 4 — ~ ~ 205 18.5 71.0 5 - 9 — — ~ 204 18.4 89.4 More than 10 — ~ — 118 10.6 100.0 Frequency of Next 12 month (total) 5.7 3.0 8.1 1,021 92.0 100.0 game 0 — — — 114 10.3 11.2 attendance 1 — — — 150 13.5 25.9 2 ~ — — 218 19.6 47.2 3 - 4 — — — 183 16.5 65.1 5 - 9 ~ — ~ 235 21.2 88.1 More than 10 ~ — ~ 121 10.9 100.0 Money spent Total (C$) 367.43 60 1,143.29 1,110 100.0 100.0 for ticket Missing — — ~ 35 3.2 3.2 (last 12 0 — — — 431 38.8 42.0 months) 1 - 5 0 — ~ - 86 7.7 49.7 51 - 100 — ~ — 142 12.8 62.5 101 -200 — ~ — 156 14.1 76.6 201 -500 — ~ — 121 10.9 87.5 More than 500 - - - 139 12.5 100.0 Ticket Season ticket - - 113 10.2 --information Half season ~ - -- 23 2.1 --Twenty two game -- - -- 13 1.2 -Fifteen game - -- - 16 1.4 Eleven game - -- - 58 5.2 ~ Group rate - -- - 103 9.2 ~ Single ticket - ~ - 838 75.5 --0 — — — 103 9.3 9.4 1-5 — — — 142 12.8 22.5 6 - 8 — — — 113 10.2 32.8 9 - 1 0 — — - 147 13.2 46.3 Years of being 11 - 15 — - - 150 13.5 60.0 a fan 16-20 ~ - ~ 193 17.4 77.7 More than 20 ~ — ~ 243 21.9 100.0 Missing ~ - ~ 19 1.7 Total 13.9 13.0 9.6 1,110 100.0 Internet Yes ~ — ~ 703 63.3 63.3 No -- -- - 408 36:8 100.0 49 Spectators attended five games (SD = 7.5, Median = 2.0) on average in the last 12 months and planned to attend 5.7 games (SD = 8.1, Median =3.0) on average in the next 12 months. They spent C$ 367.43 (SD = 1,143.29, Median = C$ 60.00) on average for their season's game tickets and 431 (38.8 %) subjects reported that they spent C$ 0.00 for their game tickets, which implied they obtained free tickets (e.g., tickets from companies, gifts, school). Spectators in this study had been fans for 13.9 years (SD = 9.6, Median = 13.0) on average and 10.2 % were season ticket holders. 50 Study 1 In Study 1, the hypothesized seven-factor structure containing reasons for game attendance was evaluated using confirmatory factor analyses (CFAs). All confirmatory factor analyses in this study were conducted with LISREL 8.3 (Joreskog & Sdrbom, 1999), and estimation was done using maximum likelihood procedures with a covariance matrix as data input. Psychometric Evaluation of Reasons for Game Attendance Confirmatory factor analysis. The results of the confirmatory factor analysis for part 1, game attendance, are presented in Table 11. Four different models were evaluated with data from part 1 of data sets 1 and 3 (N = 628). In the original model, all 32 items and seven factors were evaluated. Before evaluating the original model, descriptive statistics of items including mean, standard deviation, skewness, and kurtosis were examined. One item, 'unobstructed view from seats' under the factor 'facility convenience' was not included in the original model due to a high kurtosis score (3.8). The parsimony goodness-of-fit index (PGFI), comparative fit index (CFI), and root mean square error of approximation (RMSEA) showed that the data did not conform to the original model of 32 items and seven factors (Table 11). After examination of the residuals, modification indices, and standardized solution, nine items were excluded from the next model due to large residuals, large modification indices, or low factor loading scores. Two of three items showing low loading scores (.14 and .31) under the factor, Sense of Community, were removed from the next model and thus just one item remained for this factor. The next model with 23 items and seven 51 factors was significantly better than the first model, xltff = 1*003, dfdjff = 236, p < .001, but the goodness of fit indices suggested that it was still not an adequate model of the data structure. Table 11. Results of the Confirmatory Factor Analyses for Part 1 Model (item, construct) x\ df, P Q PGFI CFI RMSEA Original (32, 7) 1,911, 448, <.001 4.3 .71 .83 .074 Revision 1 (23, 7) 908, 212, <.001 4.3 .68 .90 .072 Revision 2 (21, 7) 495, 171, <.001 2.9 .68 .95 .056 Final form (19, 7) 339, 134, <.001 2.5 .67 .96 .049 Based on the result of the second model, two items with large residuals and low loading scores were removed from the model and the third model (revision 2 with 21 items and seven factors) revealed a significant improvement from the previous model Xdifr = 413, dfm =5\,p< .001. Although revision 2 had a relatively stronger dimensional structure than the previous models, the RMSEA (.056) indicated that it did not represent a close fit but only a reasonable fit. Therefore, by removing two additional items (one item at a time) with loadings less than .60 (.56 and .58), two additional models were tested. The final model contained 19 items and seven factors. A number of model fit indices including chi-square statistics (Q = 2.5), SRMR (.043), CFI (.96), and RMSEA (.049) revealed a good fit of the final model. Among the seven factors, each of three factors (Overall Performance, Facility Convenience, and Game Promotion) had three items; each of two factors (Uniqueness of Hockey and Schedule Convenience) had two items; Advertising had five items; and Excitement had one item. Three factors 52 (Uniqueness of Hockey, Schedule Convenience, and Excitement) contained less than three items, which could possibly misrepresent the meaning of latent variables and provoke the low reliabilities of factors (Marsh et al., 1998). Even if it could be a weakness of this study in that three factors had less than three items, items of these three factors are assumed to adequately represent the content of latent variables due to high factor loading scores (.75 to .83). Table 12. Factors, Items, Means, and Loading scores of the Final Model Factor Item M Loading score Overall Performance The Canucks' win/loss record 3.3 .63 (O.P.) The Canucks' superstars 3.4 .71 Record breaking performance of the Canucks 3.6 .67 Uniqueness of Hockey (U.H.) Offensive output of the Canucks (speed of game, power play) 4.0 .75 Defensive output of the Canucks 3.5 .77 Facility Convenience Comfortable seats 4.1 .82 (F.C.) Easy accessibility to facility 3.8 .77 Cleanliness of facility 3.8 .75 Schedule Convenience Game time (time of day) 3.6 .83 (S.C.) Game day (day of week) 3.4 .80 Ticket discount 4.0 .70 Game Promotion Price of season ticket 3.8 .62 (G.P.) Price of game ticket 4.3 .87 Advertising through TV. 2.5 .88 Advertising (ADVT) Advertising through newspapers 2.4 .92 Advertising through magazines 2.1 .85 Advertising through radio 2.5 .83 Advertising through internet (websites) 2.2 .75 Excitement (EXCIT) Excitement 4.4 1.0 The 19 items from the final model are presented with means and factor loading scores in Table 12. All 19 loadings exceeded .60 and the lowest loading score was .63. The labels of factors, except for factor seven, remained the same based on the item content. The last factor in Table 12 had only one item, which was 'excitement' and the 53 factor's label was changed from Sense of Community to Excitement. Internal consistency and test-retest reliabilities. In Table 13, mean scores, standard deviations, and internal consistency statistics for the final model were presented. In addition to descriptive statistics, the two-week test-retest reliability coefficients (correlations) are provided. For the test-retest reliability, the sample consisted of 28 male and female adults who attended at least one Canucks' hockey game in the last year. Table 13. Descriptive Statistics, Internal Consistency and Test-Retest Reliabilities Factor M SD Cronbach's a Test-retest r O.P. 9.96 2.72 .71 .81 U . H . 7.57 1.73 .72 .83 F.C. 11.74 2.62 .82 .80 S.C. 6.92 2.04 .80 .90 G.P. 11.95 2.92 .75 .82 A D V T . 11.75 4.88 .93 .90 E X C I T . 4.37 .78 - .82 For the internal consistency of each factor, Cronbach's a was applied, which ranged from .71 to .93. As foreseen by Schutz (1995), the alpha coefficients were influenced by the number of items in the scale and three factors which contained 2 or 3 items showed low alpha coefficients (.71, .72, .75). The Overall Performance factor had the lowest alpha coefficient, .71 with three sub-items. Although the Schedule Convenience factor had only two items, its coefficient exceeded .80 because of a high inter-item correlation (.67). The Advertising factor showed not only the highest alpha coefficient, .93 but also high inter-item correlations (.64 to .82) with five sub-items, which supported that alphas were influenced by the number of highly correlated items. 54 The test-retest reliabilities were very high, ranging from .80 to .92. Convergent and discriminant validity. Using data set 3 including 197 subjects (6 list-wise deletions), convergent and discriminant validities of the final 19-item model were evaluated with Zhang's Spectator Enjoyment Scale (1999). Zhang et al. identified 9 variables affecting spectator enjoyment of hockey games: music, socializing with friends, speed of the game, player appeal, promotions, entertainment, people watching, tough/aggressive play, and player fighting. With respect to convergent validity, correlations between similar latent constructs should be higher than others. Table 14 presents correlations between the constructs. For instance, the factor U.H. (Uniqueness of Hockey) showed a stronger relationship with the four latent constructs player appeal, speed of game, aggressive play and player fighting, as represented by the relatively higher correlations (.21, .21, .29 and .20 respectively). Both the Game Promotion and Advertising factors had higher correlations (.29 and .28) with Zhang's promotion's factor than with any other factors. With respect to discriminant validity, correlations between conceptually different latent constructs should not be high. While the U.H. factor had relatively strong relationships with similar latent constructs such as speed of game or tough/aggressive play, this factor had low correlations with dissimilar constructs including music (.01), socializing with friends (.06), promotions (.06), entertainment (.07), and people watching (.06). This pattern, showing consistency in representing correlations between similar 55 latent constructs and correlations between dissimilar constructs, presented evidence for convergent and discriminant validities of the final model, which presumed that similar items of both scales were measuring the same values. Table 14. Correlations between Latent Constructs with Zhang's Scale Factor Zhang's Scale OP. C.H. FC. S.C. G.P. ADVT EXCIT Music .01 .01 .19 .00 .07 .18 .17 Socializing with Friends -.03 .06 .08 .04 .09 -.06 .16 Speed of Game .18 .21 .09 -.17 .10 -.03 .14 Player Appeal .29 .21 .00 -.02 .20 .29 .20 Promotions .13 .06 .22 -.02 .29 .28 .12 Entertainment .20 .07 -.01 -.06 .07 .12 .23 People Watching .12 .06 .19 .25 .18 .17 .26 Tough/Aggressive Play .16 .29 -.00 -.10 .08 -.10 .14 Player Fighting .10 .20 -.01 -.09 -.00 -.07 .22 Criterion validity. In order to assess criterion validity, the seven factors were identified as predictor variables and the three attendance measures were defined as criterion variables, namely, frequency of game attendance in the past 12 months, frequency of game attendance in the next 12 months, and spectators' ticket information. Multiple regressions were conducted for each of the three attendance measures, the results indicating that a linear combination of the seven factors representing reasons for game attendance did not explain much variance in any of the attendance measures (R2 = .028 for number of games attended in the past 12 months, R2 = .019 for number of games attended in the next 12 months, R2= .026 for spectators' ticket information). In order to more closely examine the nature and strength of the relationship between the predictors 56 and outcome measures, zero-order correlation coefficients were computed (see Table 15). The correlation coefficient indicates that there are no significant relationships between the seven factors purporting to reflect reasons for game attendance, and the three behavioural measures of game attendance. In other words, the seven factors identified as important game attributes for spectators' attending games and their game attendance levels (both current and intentional levels) were independent of one another, which meant that the final model did not predict well spectators' game attendance levels or their ticket preferences. Table 15. Correlation Coefficients for Criterion Validity Criterion \ Factor O.P. U.H. FC S.C CP. ADVT EXCIT # of games attended in last 12 months .07 .13 .03 -.03 -.15 -.01 .02 # of games attended in next 12 months .03 .10 .02 -.04 -.16 -.01 .02 Ticket information .07 .14 .11 -.02 -.10 .08 .01 Due to the poor predictive probability of constructs identified from the confirmatory factor analysis, further statistical analyses using these seven factor scores may not be appropriate. Instead, in order to classify spectators and attempt to identify common characteristics of attendees, a decision tree analysis, a heuristic data analysis method, was conducted. Predicting Game Attendance Using Decision Tree Analysis According to Hair et al., (1998), decision trees are a sequential partitioning of the dataset in order to maximize differences on a dependent variable. In other words, decision tree analysis is the process of extracting information from the dataset. In this study, a decision tree analysis using Exhaustive CHAID was performed in order to 57 classify spectators based on all variables in parts 1, 2, and 3 of the questionnaire. CHAID (Chi Squared Automatic Interaction Detector) is an exploratory method for classifying categorical data. The purpose of the procedure is to split a set of objects in a way, that the subgroups differ significantly with respect to a designated criterion. The criterion matches the dependent variable, while the remaining attributes represent their predictors in the model. The segments derived by CHAID are mutually exclusive and exhaustive which means, that the segments do not overlap and each object of the sample is contained in exactly one segment. Therefore the application of the method approves the classification of new objects by knowing the categories of the predictors (Magidson, 1993; cited from Lehmann & Eherler, p. 1). A revision to the basic CHAID algorithm presented above, Exhaustive CHAID, was proposed by Biggs et al. (1991). In this procedure the best split emerges by merging similar pairs continuously until only a single pair of categories remains. Table 16. Variables for the Initial Decision Tree Analysis Variable Items Sub-items Labels Measurement Dependent Frequency of game — N U M G A M Continuous variable (Target) attendance in last 12 months Part 1 33 game attribute - Continuous items Part 2 Money spent for ticket — G A M E M O N Continuous Ticket information Season ticket SEA TIC Nominal Eleven pack E L E T I C Nominal Fifteen pack FIF TIC Nominal Independent Twenty two pack TWENJTIC Nominal Variables Group rate ticket GRP TIC Nominal (Predictors) Half season ticket H A L F TIC Nominal Single game SIGL TIC Nominal Who buy tickets I do I DO W B Y Nominal Spouse SPOU W B Y Nominal Friend FRID W B Y Nominal Other family member FA M W B Y Nominal My company C O M P W B Y Nominal Best promotional T V . T V M E D Nominal medium Radio R A D I O M E D Nominal Newspaper N E W P M E D Nominal Magazine M A G A M E D Nominal Website WEB M E D Nominal 58 Independent Variables (Predictors) Game attending Alone A C A L Nominal colleague Spouse AC_SPOUS Nominal Children A C CHILD Nominal Family A C F A M Nominal Relatives A C R E L A T Nominal Co-worker A C C W O R K Nominal Neighbours A C NEIGH Nominal Friends A C FRIEN Nominal Years of being a fan — Y R S FAN Continuous Membership of club Soccer club C L U B SOC Nominal Hockey club C L U B HOC Nominal Golf club C L U B GOL Nominal Fitness club C L U B FIT Nominal Hobby Golf H GOLF Nominal Hockey H H O C K E Y Nominal Art H ART Nominal Fitness H FITNES Nominal Amateur sports H A M A S P Nominal Reading H READIN Nominal Ski & Snowboard H SKI BOA Nominal Favourite sports Golf A C T GOLF Nominal activity Baseball A C T BASE Nominal Hockey A C T HOCK Nominal Ski & snowboarding A C T SKIB Nominal Basketball A C T B A S K Nominal Soccer A C T SOCC Nominal Volleyball A C T V O L L Nominal Favourite outings Movie OUT MOVI Nominal Football OUT FOOT Nominal Concert OUT CONC Nominal Amateur sports OUT A M S P Nominal Dinner OUT DINN Nominal Drinking OUT DRIN Nominal Favourite website Major media W E B M E D I Nominal Sport media WEB SPME Nominal Computer related WEB C O M Nominal Hockey related WEB HOCK Nominal Favourite sports as a Baseball YSP B A S E Nominal child Hockey YSP HOCK Nominal Basketball YSP B A S K Nominal Football YSP FOOT Nominal Soccer YSP SOCC Nominal Volleyball YSP V O L L Nominal Part 3 Age - A G E Continuous Gender — GENDER Nominal Dwelling information — RESIDENT Nominal Job — JOB Nominal Household income — INCOME Continuous Level of education - E D U L E V E L Nominal 59 Three decision tree analyses were conducted as part of study 1. In the first analysis, data set 1 (N = 628) was used and the target (dependent) variable and predictive (independent) variables were configured as presented in Table 16. The frequency of game attendance in the past 12 months was used as a dependent variable due to the fewer number of missing cases (1 case) than those of intentional future game attendance (49 missing cases). In terms of the measurement scales of the variables, the dependent variable was a continuous and the independent variables consisted of both continuous and nominal variables (see Table 16). The tree diagram based on data set 1 is shown in Figure 3. The mean frequency of game attendance over the last 12 months, the dependent variable, had a mean of 4.8 games (SD = 7.2) at node 0 (the total sample). Eleven terminal nodes (2, 4, 6, 7, 8, 9, 10, 11, 12, 14, & 15) in total were identified by five predictors; SEATIC (season ticket holder or not), WEBHOCK (hockey related websites as favourites), G A M E M O N (money spent for game tickets), JOB, and GENDER with the mean attendance in the terminal nodes ranging from 16.4 to 0.9 games. SEATIC was the best predictor, and there was a large difference in mean scores (16.4 versus 3.6 games) between season ticket and non-season ticket holders. With respect to WEBHOCK, people (node 4) who considered hockey related websites as their favourite web-sites represented higher game attendance frequencies (6.9 games) than the opposite group (node 3, 3.1 games). G A M E M O N was an important interactive predictor with WEBHOCK. Specifically, of the people who did not identify hockey websites as their favourite, those who spent more than C$ 250 on 60 Figure 3. Tree Diagram of the First Analysis Mean 4.8312 Std. Dev. 7,1915 % 100.(0 Predtted 4.8312 SEAJ1C Adj. P-vakje-0.0000, F=231.1123, dF=l,626 Nodel Node2 Mean 3.6292 Mean 16.4237 Std.Dev. 4.51D7 Std. Dev. 14.4714 n 569 n 59 % 90.61 % 9.39 Predicted 3.6292 Predicted 16,4237 1 bl wEBjea Mj, P-value-0.0000, F-57.5680, J-1,567 Mean e3 3.0558 Std. Dev. 3.7448 n 484 % 77.07 Predicted 3.0558 GAMEMON Adi. P-vake-0.0000. F-24.4097. dl =5,478 <50jS°] Node6 Mxie7 Mean 1.8163 Mean 2,3913 Std.Dev. 1.2192 Std.Dev. 1.8071 n 49 n 46 % 7.80 % 7.32 Predicted 1.6163 Predicted 2,3913 (90,150] Node 8 Mean 2.6B52 Std.Dev. 1.8806 n 61 % 9.71 Predicted 2.8852 >250l<TBSsing> Mean Std. Dev. Predrcted 6.7123 5.6874 73 11.62 6.7123 Adj, P-vakje-O.OuTO, F-15,3303, df-2,206 Node 11 Node 12 Node 13 Mean 2.1525 Mean 3.8389 Msan 1.1810 Std.Dev. 1.7500 Std. Dev. 4.3371 Std.Dev. 1.8069 n 59 n 45 n 105 % 9.39 % 7.17 % 16.72 Predrcted 2.1525 Predicted 3,6689 Predicted 1.1810 1 tl Adj. P-vak»=0.EMi5, F=4.2627, df=l,103 1 male 1 1 female l Node 14 Node 15 Mean 0.8644 Mean 1.5870 Std.Dev. 1.0900 Std. Dev. 2.3905 n 59 n 46 % 9.39 % 7.32 Predkted 0.8644 PrerJcted 1.5870 tickets (node 10) had higher game attendance frequencies (6.7 games) than those spending less money (nodes 5, 6, 7, 8, & 9). That is, people who spent more money for game tickets tended to attend more games than others who spent less money for tickets. Although two other predictors, JOB and GENDER, were identified as significant predictors in the initial analysis, all nodes (11, 13, 14, and 15) except for node 12 61 classified by these two predictors represented lower game attending frequencies than other terminal nodes (2, 4, 7, 8, 9, and 10). Individuals from node 12 whose job specialties were in business & finance, natural & applied science, unique to processing, manufacturing & utilities, trades & transport, art, culture, recreation & sport, and health represented higher game attendance levels than individuals from node 11 (management, social science, education, government service & religion, sales & services, and unique to primary industry) and node 13 (retired, student, and currently unemployed). With respect to gender, female respondents from node 13 represented higher game attendance frequencies (1.59 games) than males (0.86 game). None of the 33 game-attribute items was identified as an important predictor in the initial analysis, which supports no significant relations between the frequency of game attendance and game attribute items of the final form of the questionnaire. With respect to a sample size for CHAID, Sonquist et al. (1964) recommended a minimum sample size of 1,000. In terms of the ratio between the number of observations and predictor variables, MacLanchlan and Johansson (1981) suggested 33 times as many observations as predictors. In addition, Perreault and Barrksdale, (1971) insisted that a minimum group size of 40 would be reasonable. Given these recommendations, increasing the sample size and reducing the number of predictor variables would be desirable for a reliable decision tree analyses. Since none of 33 items from Game Attendance (Part 1) was identified as an important predictor by the initial analysis, those items were excluded for further analyses, and thus data sets 1, 2 and 3 (which all contained components 2 and 3) could be combined to form a data set (# 4) 62 with a large #(1,110) suitable for further analyses. Table 17. Variables for the Second Decision Tree Analysis Variables Items Sub-items Measurement Dependent variable (Target) Frequency of game attendance in last 12 months — Continuous Ticket information Season ticket Eleven pack Single game Nominal Nominal Nominal Who buys ticket? I do Nominal Best promotional medium T.V. Radio Newspaper Website Nominal Nominal Nominal Nominal Game attending colleague Spouse Co-worker Friends Nominal Nominal Nominal Years of being a fan - Continuous Membership of club Soccer Hockey Golf Nominal Nominal Nominal Independent Variables (Predictors) Hobby Golf Hockey Reading Nominal Nominal Nominal Favourite sports activity Golf Hockey Soccer Nominal Nominal Nominal Favourite outings Movie Concert Dinner Nominal Nominal Nominal Favourite website Major media Sport media Computer related Hockey related Nominal Nominal Nominal Nominal Favourite sports as a child Baseball Hockey Basketball Soccer Nominal Nominal Nominal Nominal Age - Continuous Gender - Nominal Dwelling information — Nominal Ethnicity - Nominal Job — Nominal Household income - Continuous Level of education ~ Nominal In addition to removing the 33 game attribute items, variables containing a 63 number of missing values (ticket purchasing time, frequency of future game attendance, and children's favourite sports) or showing a low frequency (HALFTIC, n = 23, and FIFTIC, n = 16) were removed from further analyses. The thirty-nine variables for the second decisions analysis are presented in Table 17. The second decision tree analysis was conducted and the full tree diagram is presented in Figure 4. Nine predictive variables were identified; SEATIC (season ticket holder), AC_CWORK (accompanied by co-workers), WEB_HOCK (hockey related websites as favourite websites), ACFRIEN (accompanied by friends), SIGLTIC (single game ticket holder), CLUB HOC (membership of hockey club), I D O W B Y (I am the ticket purchaser), JOB (occupations), and AC_SPOUS (accompanied by spouse). The mean for the dependent variable, number of games attended in the past 12 months, was 4.99 (SD = 7.49) and SEATIC was identified as the most important predictor for classifying frequencies of game attendance in the second analysis. Season ticket holders attended more games than non-season ticket holders (16.9 versus 3.6 games), as expected. Although SEATIC was identified as the best predictor in the second analysis, it can be excluded from predictive variables for further tree analyses. According to one of pre-defined stopping rules for CHAID classification, the minimum number of cases in a parent node must be more than 100 and the minimum number of cases in a terminal node must be more than 40 (Perreault & Barrksdale, 1971). Due to this stopping rule, no further branching was done on node 1, and therefore no further information on the 113 season ticket holders was revealed. For the next tree analysis, the three ticket information variables were eliminated. 64 Figure 4. Tree Diagram of the Second Analysis with Data set 4 NodeS Mean 2,7721 SEA TIC Adi. P-vdue=0.0000, F=451.3H6, cf= .1108 V « B _ H O C K Adj. P-vatue-0.0000, F-59.4694, df= Node 3 Mean 3.2088 Std. Dev. 4.0916 n 862 % 77.66 Predicted 3.2088 Node 4 AC_CWORK Adj. P-vakje-0.0O3O, F-68.0666, df-1,860 I SIGL_TIC Adi. P-vakje-O.OOOO. F-33669S, df - Node 7 Mean 4,2743 Std, Dev. 4.3839 n 113 % 10.18 Predated 4,2743 I DO_WBY p-value-0.0047, F-8,3126, df—i, 111 ! AC_SPOUS Adi- P-vdue-0.0033. F-9.0213. df-1,114 I ] aUB_HOC Adj. P-vakje-O.OQOOj F-26.3561, df-1,631 _ L Node 1 Mean 5.7442 Std. Dev. 5.1459 n 43 % 3.87 Predicted 5.7442 7;11;2;10;6 I Mean 1.8369 Std. Dev. 1.7593 n 239 % 21.53 Predicted 1.8368 3.3714 3.5921 AC_SPOU5 Adj. P-vakB -0.0H3.F-6.0894. df-1.237 Node 13 Mean 2.3251 Std. Dev. 22079 n 569 % 51.26 Predicted 2.3251 Adj. P-value-0.0008, F-13.5093, df-2,566 3.0789 2.9544 AC_FRIEN Adj. P-value-O.0341. F-t.6025, df-1.112 12j8;l;3;9.<rrissing> Ncde 17 Mean 2.4676 5td. Dev. 2.0683 n 216 % 19.46 Predicted 2.4676 Node 20 Mean 2.4630 Std, Dev. 2.3689 n 54 % 4.86 Predicted 2.4630 Node 21 Mean 3,6333 Std, Dev. 3.3190 n 60 % 5.41 Predicted 3.6333 After removing all three ticket information variables, SEATIC, ELETIC, and SIGL TIC, the final decision tree analysis with 36 predictive variables was conducted. 65 The full tree diagram is presented in Figure 5. Figure 5. Tree Diagram of the Final Model with Data set 4 NodeO Mean 4.9901 I S i Dev. 7.+B57 | 1110 100.00 Predcted 4.9901 j T""" DHOW *di.P-value-0.onoo, W4.3647, rj-2,1107 L Mem 3,6364 Std.Dev, 5.2005 n 473 % 42,61 Predcted 3.6364 1 (5,7L<rrisarj> Noda2 Mean 5,0499 Std. Ow. 7.751! n 421 % 37.93 Predtted 5.0499 WB HOCK Act. P-valuN).0t00, M8.5585, rj-l,471 Mean 7,8380 Std.Dev, 10,0045 n 216 % 19.46 Prodded 7.8380 WEBJWX Adi. Naue=O.O001,M57529, tj=l,4l9 ~~1 B YS> BASE « 4 P - v a > j e * 0 1 X , F = S . 2 7 7 6 , d r . | , 2 1 4 Node5 rean 76154 Sld.Dev. 9.1048 % 586 Predtted 71154 Adt. Pvakt-OOOTO. F-14.6S79.df-2.40S 1 7;10;9 1 1 4;12;8;ll;5,<mS5rig> 1 l;3;2|13;6 Node 10 Node 11 Node 12 Mean 1.9306 Bean 4.3478 Mean 2.7540 Std Dev.' 2.2360 Std.Dev. 5.6331 StiCev. 2.4485 n 144 n 138 n 126 % 12.9? % 12.43 % 11.35 Predtted 1.9306 Predcted 4.3478 Predcted 2.7540 ACCWOftr. Adt Naiue-Ofl305, F-12,5103, dr-1,368 YSPHOCX Adl, hdur-0.0156", F-S.9931, rJ-1,141 I ' 1 Node 13 Nat 14 Mean 3.9969 Mean 7.8163 Std. Dev. 6.7033 Std. Dev. 8.9714 n 321 n 49 % 28.92 % 4.41 Predcted 3.9969 Predtted 7.8163 nunc Adi, rVvdue=0.01B2, F.10.1028, df=l,142 I t * 15 Node 16 Mean 9.0000 Mean 5.4211 Std.Dev. 11.1964 Std.Dev. 6.4784 n 51 n 95 % 4,59 % 8,56 Predtted 9.0000 Predcted 5.4211 fflSFAN Adi. P-vjlue=0.0000, (==14,7711. df=2,318 1 1 1 1 3;4;2 1 1 <=15J<rresrig> (15,28] 1 >28 1 Node 17 tat 18 Node 19 Node 20 Node 21 Meal 2.3200 Mam 1,0455 Mem 2.5888 Mem 5.3333 Mean 8.1250 Std. Dei. 2.59)3 Std.Dev. 0.8880 Std.Dev. 4.0303 Std.Dev, 7.2117 Std Dev. 12.0431 n 100 n 44 n 197 n 84 n 40 % 9.01 % 3.96 % 17.75 * 7.57 % 3,(0 Predcted 2,3200 Predtted 1.0455 Predcted 2.5888 freckted 5.3333 Predtted 8.1250 AC SPOUS Adi. P-valiHI.0055. F-7S977, df-1,195 Node 22 Node 23 Hem 2.0000 Mem 3.6571 Std.Dev. 1.7275 Std.Dev. 6.2366 n 127 n 70 14 11.44 % 6,31 Predtted 2.0003 Predtted 3.6571 ~~1 IV MED Act P-valr»40048" M.2522. dM,125 Node 24 Node 25 Mem 2.4821 Mean 1.6197 Std.Dev. 2,0626 StiOev. 1,3024 n 56 n 71 % 5,05 % 5.40 Predtted 2,4821 Predtted 1.6197 66 The full tree diagram was classified by 10 predictors (INCOME, WEB_HOCK, YSP_BASE, JOB, AC_CWORK, ETHNIC, T V M E D , YRS_FAN, and YSPHOCK), and yielded 25 nodes, 15 of which were terminal nodes. A brief description of each terminal node is presented in Appendix B. Node 9 represented 70 spectators whose mean frequency of game attendance was 10.3 games (SD = 12.2). In general, the traits and behaviours of those individuals with the highest attendance frequencies were: high income, identifying hockey-related websites as their favourite websites, played baseball or hockey when they were young, preferred being accompanied at games by co-workers, and had been hockey fans for a long time. A gain summary of the tree analysis is presented in Table 18. Eight segments (nodes) had frequencies greater than the mean frequency of game attendance (5.0). The most important segment was node 9 with an index of 205.8 %, which implied that targeting an individual of node 9 would be anticipated to increase the frequency of game Table 18. Gain Chart Defined by 9 Predictors Node n M Node (%) Cumulative (%) Index (%) 9 70 10.27 6.3 6.3 205.8 1 51 9.02 4.6 10.9 180.8 15 51 9.00 4.6 15.5 180.4 21 40 8.13 3.6 19.1 162.8 14 49 7.82 4.4 23.5 156.6 5 65 7.62 5.9 29.4 152.6 16 95 5.42 8.6 37.9 108.6 20 84 5.33 7.6 45.5 106.9 11 138 4.35 12.4 57.9 87.1 23 70 3.66 6.3 64.2 73.3 12 126 2.75 11.4 75.6 55.2 24 56 2.48 5.0 80.6 49.7 17 100 2.32 9.0 89.6 46.5 25 71 1.62 6.4 96.0 32.5 18 44 1.05 4.0 100.0 21.0 67 attendance 2.06 times over a random selection. If a franchise targeted people belonging to the top six segments (node 9, 7, 15, 21, 14, & 5), the frequency of game attendance would be expected to increase by a range of 152.6 to 205.8 percent. That is, people who had one or more of the following set of characteristics: (a) node 9: high incomes (greater than C$ 100,000) and played baseball when young; (b) node 11: middle to high incomes (C$ 60,000 to 99,999) and hockey websites as their favourite internet sites; (c) node 15: high incomes and played hockey when young; (d) node 21: the greatest fan history (more than 28 years) and middle to high incomes; (e) node 14: middle to high incomes and accompanied by their co-workers; (d) node 5: low to middle incomes (less than C$ 60,000) and hockey websites as their favourite internet sites. Comparing Low, Middle, and High Game Attendance Groups Activity pattern comparisons among the three attendance groups. Descriptive statistics are presented in Table 19 to examine differences in spectators' characteristics based upon their game attendance frequencies. Since there was no specific definition available to divide spectators into low, middle and high attendance groups based strictly on their game attendance frequencies, three groups of similar size were formed to force the sample into three populations. The three categories were as follows: Low frequency group (0 to 1 game), Middle frequency group (2 to 4 games), and High frequency group (more than 4 games). As indicated previously, fan history had a positive relationship with game attendance. The Low frequency Group represented the shortest fan history with a mean of 11.2 years and the High Group had the longest fan history mean (16.8 years). While 68 172 (53.4 %) individuals from the High Group had been Canucks' fans for more than 16 years, only 44 (13.7 %) had fan histories of less than 6 years. On the other hand, 225 (58.2 %) people in the Low Group belonged to a short fan history category of 0 to 10 years. Table 19. Group Descriptive Statistics of Activity Pattern Group Low # Group Middle # Group High # Group (n-=386) (n= 402) (n= 322) Predictors 0 -1 2 -4 > 4 Mean Mean Mean Game attendance 0.87 2.7 12.8 Chi-square or F" P Years of being a fan 11.2 14.2 16.8 F= 30.681 .001 n % n % n % Season 10 2.6 13 3.2 90 28.9 156.730 .001 Half season 2 0.5 4 1.0 17 5.2 23.217 .001 11 pack 5 1.3 14 3.5 39 12.1 45.340 .001 15 pack 1 0.3 2 0.5 13 4.0 21.592 .001 Group 26 6.7 35 8.7 42 13.0 7.742 .021 Single 331 85.8 345 85.8 163 40.6 155.514 .001 Best medium T.V. 185 47.9 202 50.2 119 37.0 14.043 .001 Newspaper 102 26.4 110 27.4 113 35.1 7.488 .024 Game colleague Co-worker 32 8.3 47 11.7 72 22.4 31.531 .001 Friends 208 53.9 252 62.7 181 56.2 6.690 .035 Sport club Hockey 26 6.7 51 12.7 65 20.2 28.476 .001 Hobby Golf 37 9.6 33 8.2 47 14.6 8.380 .016 Hockey 56 14.5 77 19.2 93 28.9 22.935 .001 Favourite activity Golf 53 13.7 59 14.7 66 20.5 6.834 .033 Hockey 84 21.8 129 32.1 130 40.4 28.895 .001 Outing Concert 97 25.1 100 24.9 109 33.9 8.975 .011 Favourite website Major media 40 10.4 55 13.7 21 6.5 9.799 .007 Sport media 19 4.9 35 8.7 34 10.6 8.165 .017 Hockey related 27 7.0 47 11.7 78 24.2 46.229 .001 Play as a child Hockey 109 28.2 141 35.1 127 39.4 10.170 .006 Note. Chi-square orF" : Reported test statistic is %2, unless otherwise noted. The complete descriptive statistics table for activity pattern is presented in Appendix C. 69 With respect to a promotional medium, TV was considered as the best medium by all three groups. About 48 and 50 percent of observations in the Low and Middle Groups believed TV was the best promotional medium. For the case of the High Group, TV (37.0%) was the most preferred medium and newspapers (35.1 %) the second most preferred. In terms of attending games with colleagues, the three groups presented similar patterns regarding friends but differed significantly regarding co-workers with 22.4 % of the High Group attending games with co-workers versus 8.4% and 11.4% of the Low and Medium Groups, respectively. While 65 individuals (20.2 %) in the High Group replied that they were members of hockey clubs, 6.7 and 12.7 percent individuals in the Low and Middle Groups were members of hockey clubs. In all three groups, hockey was considered as the most favourite activity and golf was the second. For the case of favourite outing, differences between groups were uncovered concerning the item, concert. Individuals in the High Group reported greater interest (33.9 %) in concerts than individuals in the other two groups (25.1 and 24.9 %). The use of hockey related websites was identified in the tree analyses as one of the most important variables to predict game attendance. In terms of the use of hockey websites, the difference between the High Group and the Low and Middle Groups was clear. While 27 (7 %) and 47 (11.7 %) of survey respondents in the Low and Middle Groups replied that they were hockey website users, 78 individuals (24.2 %) in the High Group reported being hockey website users. With respect to preferred sports played when young, all three groups were the 70 same in that hockey had the highest frequencies. Unlike other sports including soccer, basketball, football, and volleyball, it was interesting that playing both hockey and baseball reflected a linear relationship with game attendance frequencies between the Low, Middle, and High Groups (hockey: 28.2 %, 35.1%, 39.4 % and baseball: 21.5 %, 25.6 %, 28.3 %). As can be seen in Table 19, the differences in relative frequencies of all these reported activities were significantly (p < .05) different among the three attendance groups, as indicated by a chi-square test. However, given the large number of significant tests, an alpha of .01 was used to partially protect against Type I errors (a full Bonferroni correction, .05/20 = .0025, would have led to overly conservative significance tests). Six predictors including group ticket holder, newspaper as the best medium, accompanied by friends, golf as a hobby, golf as a favourite activity, and sport media as a favourite websites were not significant between groups after this correction. Demographic characteristics comparisons among three groups. Individuals in the Low Group were the youngest with an average age of 31.3 years, and people in the High Group were the oldest with a mean of 34.7 years. In terms of residence area, while the majority of the Low and Middle Groups were living in Area 1 (Vancouver), the residence areas of the High Group were distributed through Area 1, 3 (Richmond, Delta, &White Rock), 4 (North & West Vancouver), and 6 (Surrey & Langley). With respect to job categories, the Low and Middle Groups exhibited similar patterns. Compared to the other two groups, the portion (19.6 %) occupied by management related jobs was larger than those of the Low (13.5 %) and Middle (12.9 %) 71 groups. With respect to income, the Middle group had a slightly higher household income than the Low Group overall. The highest income level was represented by people of the High Group with 101 (31.4 %) individuals whose household income exceeded C$ 100,000. In terms of gender, the High Group represented a higher portion (71.1 %) of male spectators than the other two groups (60.9, 63.2) (Table 20). Table 20. Demographic Profiles Group Low # Group Middle # Group High # Group Predictors (n= 386) (n= 402) (n= 322) 0 -1 2 -4 > 4 Mean Mean Mean Age 31.3 31.7 34.7 n % n % n % Chi-square p Age 27.342 .002 21 > 78 20.2 71 17.7 29 9.0 21 - 2 5 77 19.9 70 17.4 52 16.1 2 5 - 3 0 53 13.7 70 17.4 55 17.1 3 1 - 4 0 86 22.3 85 21.1 77 23.9 4 0 - 5 0 43 11.1 53 13.2 57 17.7 50 < 30 7.8 32 8.0 36 11.2 Missing 19 4.9 21 5.2 16 5.0 Gender 9.634 .008 Male 235 60.9 254 63.2 229 71.1 Female 132 34.2 128 31.8 77 23.9 Missing 19 4.9 20 5.0 16 5.0 Residence 62.332 .001 Area 1 86 22.3 110 27.4 56 17.4 Area 2 20 5.2 31 7.7 22 6.8 Area 3 43 11.1 34 8.5 44 13.7 Area 4 18 4.7 26 6.5 44 13.7 Area 5 35 9.1 32 8.0 25 7.8 Area 6 24 6.2 41 10.2 41 12.7 Area 7 71 18.4 41 10.2 24 7.5 Missing 89 23.1 87 21.6 66 20.5 Ethnicity 12.293 .056 Caucasian 285 73.8 310 77.1 265 82.3 African Canadian 4 1.0 1 0.2 1 0.3 Asian 71 18.4 60 14.9 35 10.9 Natives 7 1.8 4 1.0 3 0.9 Job 61.692 .001 Managerial 52 13.5 52 12.9 63 19.6 Social science 23 6.0 25 6.2 21 6.5 Business / finance 46 11.9 44 10.9 43 13.4 Sales & service 36 9.3 59 14.7 46 14.3 Applied science 16 4.1 25 6.2 26 8.1 72 Manufacturing 8 2.1 6 1.5 2 0.6 Retired 13 3.4 6 1.5 8 2.5 Trade/transportation 16 4.1 21 5.2 14 4.3 Primary industry 2 0.5 4 1.0 5 1.6 Student 128 33.2 103 25.6 41 12.7 Unemployed 10 2.6 9 2.2 7 2.2 Recreation/sport 8 2.1 12 3.0 11 3.4 Health related 7 1.8 5 1.2 12 3.7 Household income 57.330 .001 <C$ 10,000 42 10.9 30 7.5 12 3.7 10,000-19,999 23 6.0 24 6.0 10 3.1 20,000-29,999 28 7.3 30 7.5 16 5.0 30,000-39,999 24 6.2 34 8.5 19 5.9 40,000 - 59,999 66 17.1 72 17.9 43 13.4 60,000 - 79,999 61 15.8 63 15.7 42 13.0 80,000 - 99,999 40 10.4 40 10.0 38 11.8 >C$ 100,000 53 13.7 62 15.4 101 31.4 Missing 49 12.7 47 11.7 41 12.7 73 Study 2 In Study 2, the hypothesized ten-factor structure containing opinions about sports was evaluated using confirmatory factor analyses (CFAs). Using a covariance matrix as data input, all confirmatory factor analyses in Study 2 were conducted with LISPvEL 8.3 (Jdreskog & Sorbom, 1999) and estimation was done using maximum likelihood procedures. Psychometric Evaluation of Opinions about Sports Confirmatory factor analysis. The results of the confirmatory factor analysis for part 4, opinions about sports, are presented in Table 21. Three different models were evaluated with data set 1 (N = 685) which included questionnaire Form B and C. In the original model, all 33 items and 10 factors were evaluated. The parsimony goodness-of-fit index (PGFI), comparative fit index (CFI), and root mean square error of approximation (RMSEA) showed that the data did not conform well to the hypothesized model. However, a high inter-factor correlation of .89 between the Child-oriented and Sport benefit factors indicated that a better structure would be established by combining these two constructs. After an examination of the residuals, modification indices, and the standardized solution, nine items were excluded from the next model due to large residuals, large modification indices, or low factor loading scores. The next model with 24 items and 9 factors showed a significant improvement in structure with = 7^4, dfdlff = 234,;? < .001, to the original model. Although revision 1 resulted in a stronger dimensional structure than the 74 original model, the RMSEA (.051) and CFI (.88) suggested that it was still not a good fit. Therefore, by reducing one item in each model, two additional models were tested. Two items under Community Pride with large residuals and low loading scores were removed and the final model with 22 items and 9 factors revealed a significant improvement from the previous model (Xdijr = 156, dfdiff = 43,/? < .001). A number of model fit indices, SRMR (.048), CFI (.90), and RMSEA (.047), indicate that this final model represented a good fit to the data. Table 21. Results of the Confirmatory Factor Analyses for Opinions about Sports Model (item, construct) x\ df, P Q PGFI CFI RMSEA Original (33, 10) 1,344, 451, <.001 2.9 .71 .76 .058 Revision 1 (24, 9) 580, 217, <.001 2.7 .67 .88 .051 Final form (22, 9) 424, 174, <.001 2.4 .65 .90 .047 The 22 items from the final model are presented along with means and factor loading scores in Table 22. All 22 loadings except for three (.41, .43, .35) exceeded .50. Table 22. Factors, Items, Means, and Loading scores of the Final Model Factor Item M Loading score Given the choice, participating in sports is more enjoyable than watching. 4.1 .41 Self-Participation It is always good to try new recreational activities. 4.3 .80 (S.P.) It is difficult to understand how people could just lay around when there are so many recreational activities to do. 3.7 .43 Cynicism (C.C.) Buying professional sport memorabilia is a waste of money. 2.9 .75 Belonging to professional sports fan clubs is a waste of time. 3.0 .50 Parents should always try to attend their 4.5 .61 Sports Benefits (S.B.) children's sport contests. Taking children to professional hockey events is a great way to spend time with them. 4.4 .57 75 Sports Benefits (S.B.) For children, the positive aspects of participating in sports far outweigh the negative ones. 4.2 .50 It is important to participate in regular physical activity with your child. 4.5 .67 Resources (R.S.) For recreation, time is a bigger problem than money. 3.2 2.5 .74 .35 Our income is high enough to satisfy nearly all our desires for sporting event tickets. Community Pride (CP) When my home team wins, it really makes me feel good. 4.2 4.4 3.8 .61 .72 .60 The success of a professional sports team brings pride to the city. Being able to cheer for my home professional sports franchise adds to my sense of community. Attitude to Pro- athlete (A.R) Professional athletes are only in it for the money. 2.9 3.3 .83 .59 Professional athletes tend to whine and complain too much. Role of Government (R.G.) Tax dollars should not be spent on building stadiums and areas for professional sport franchises. 2.7 2.9 3.5 .59 .70 -.63 Government should not assist financially troubled professional sports organizations. Government should offer tax incentives to keep a team in a city or to encourage another team to relocate there. Attitude to Amateur sports (A.A.) The enthusiasm of the amateur athlete makes amateur sports more enjoyable to watch than professional sports. 3.4 3.0 .72 .52 Amateur sporting events are quite often far superior in entertainment value than their professional counterparts. Perceived Time cost (P.T.) Professional sporting events take too much time to watch. 2.0 1.0 Internal consistency and test-retest reliabilities. In Table 23, mean scores, standard deviations, and internal consistency statistics for the final model are presented. In addition to descriptive statistics, two-week test- retest reliability coefficients (correlations) are provided. For two-week test-retest reliability coefficients, 28 male and female adults who attended at least one Canucks' hockey game in the past were contacted. For the internal consistency of each factor, 76 Cronbach's a was computed, and ranged from .41 (Resources) to .68 (Community Pride). A low inter-item correlation (.26) between two items confirmed that one of two items on the Resource factor represented the weakest loading score in the confirmatory factor analysis result. Table 23. Descriptive Statistics, Internal Consistency and Test-Retest Reliabilities Factor M SD Cronbach's a Test-retest r S.P. 12.1 2.2 .53 .80 CC. 6.0 1.8 .55 .87 S.B. 17.7 2.3 .67 .81 R.S. 5.7 2.0 .41 .85 CP. 12.5 2.0 .68 .83 A.P. 6.2 1.9 .65 .79 R.G. 8.1 3.0 .67 .84 A A 6.4 1.7 .54 .80 RT. 2.0 1.1 - .82 The inter-factor correlations based on the raw score item totals are presented in Table 24, and the relatively low correlations (.002 to .43) indicated that there was no multicollinearity among factors. Table 24. Inter-Factor Correlations Factor SP CC SB RS CP AP RG AA PT Self Participation — Cynicism -.07 — Sport Benefit .30 -.15 — Resource .03 .11 .04 — Community Pride .20 -.26 .43 .04 — Attitude about Pro-athlete .10 .28 -.07 .10 -.11 -- Role of Government -.05 .28 -.19 -.00 -.34 .24 — Attitude about Amateur sports .17 .09 .10 .07 .00 .24 .19 — Perceived Time cost -.06 .32 -.19 .09 -.32 .23 .23 .08 - 77 Cluster Analysis of Opinions about Sports Using raw factor scores based on identified factors from the confirmatory factor analysis, cluster analyses were conducted to uncover distinctive segments related to observations' opinions about sports. There is no standard to determine the number of clusters (Hair et. al, 1998), and thus both the hierarchical and K-Means methods were examined to determine the appropriate numbers of clusters. According to Hair et al. (1998), determining the number of clusters, profiling the cluster centres, and identifying outliers should be established by using a hierarchical technique, and then a non- hierarchical method with the cluster centres obtained by the hierarchical method should be conducted to form clusters. Using a hierarchical method (Ward's), three different cluster analyses (two to four clusters) were conducted to determine the number of clusters based on the raw factor score totals. The number of clusters from each analysis and the number of cases per cluster are presented in Table 25. Table 25. The Number of Clusters and Cases per Cluster Cases per cluster 1 2 3 4 2 282 403 # of clusters 3 282 244 159 4 282 112 132 159 With respect to the cluster profiles, the cluster centroids of cluster #1 on three different cluster solutions were identical, which indicated that the first cases of all three cluster solutions (n = 282) consisted of the same individuals throughout the two to four cluster solutions. According to Hair et al. (1998), a cluster centroid represents the average value of the objects contained in the cluster on all the variables in the cluster 78 variate. Based on the cluster centroids, cluster 1 and 2 from the two-cluster solution indicated two clusters were almost opposite images of each other. Although this two- cluster solution would suggest two different segments and was easy to characterize, it was possible to miss some important information underlying cluster # 2 in that it contained relatively a large number of observations (n = 403). In a cluster analysis, a large number of observations in a cluster implies that it would be increasing the average distance within each group, which suggests a less homogeneous group membership (Hair et al). The difference between the three-cluster and four-cluster solution was cluster #2 on the three cluster solution was divided into two separate clusters (#2 and #3) on the four cluster solution (Table 25). Inspection of the cluster centroids (reported in Table 26) of the four cluster solution revealed that some distinctive features of cluster #2 from the three cluster solution were distributed into clusters #2 and #3 in the four cluster solution. For instance, the S.P. centroid (0.62) of the 3 cluster solution was transformed into cluster #2 (0.24) and cluster #3 (0.93) in the 4 cluster solution. While cluster #2 represented the highest centroid on S.P. among the three clusters, the centroid of cluster #3 (0.93) still showed the highest score among the four clusters on the four cluster solution, but the centroid of cluster #2 (0.24) resulted in an ambiguous place among clusters. Due to an ambiguity of clusters and a difficult profiling for the four cluster solution, the three cluster solution was examined. Compared to the two cluster solution, the three cluster solution represented an improved distinctive group. Compared to the four cluster solution, the three cluster solution represented less ambiguity in terms of 79 profiling each cluster. Table 26. Cluster Centroids from the Hierarchical Cluster Analysis Cluster SP CC SB RS CP AP RG AA PT Two-cluster solution 1 0.37 -0.56 1.03 0.65 1.16 -0.52 -2.24 -0.30 -0.36 2 -0.26 0.39 -0.72 -0.45 -0.81 0.37 1.56 0.20 0.25 Three-cluster solution 1 0.37 -0.56 1.03 0.65 1.16 -0.52 -2.24 -0.29 -0.36 2 0.62 0.61 0.38 -0.51 -0.54 0.48 2.62 0.58 0.27 3 -1.61 0.04 -2.42 -0.37 -1.22 0.20 -0.07 -0.37 0.22 Four-cluster solution J 0.37 -0.56 1.03 0.65 1.16 -0.52 -2.24 -0.29 -0.36 2 0.24 1.37 0.24 0.74 -1.17 1.37 3.99 1.05 0.70 3 0.93 -0.03 0.50 -1.58 -0.00 -0.29 1.46 0.17 -0.10 4 -1.61 0.04 -2.42 -0.37 -1.22 0.19 -0.07 -0.37 0.22 Once the three-cluster solution was determined as the final cluster solution, K- Means (a non-hierarchical cluster analysis method) with three clusters was conducted and the numbers of cases per cluster are presented in Table 27. The cross-tabs which measured how much overlap existed between the K-Means and Ward's methods indicated 83 percent similarity, which represented consistency in clustering between the two different methods. Table 27. Overlap between K-Means and Ward's Methods Ward's K-Means Cluster 1 Cluster 2 Cluster 3 Total Cluster 1 242 35.33 % 35 5.11 % 14 2.04 % 291 42.48 % Cluster 2 21 3.07% 195 28.47 % 17 2.48 % 233 34.01 % Cluster 3 19 2.77 % 14 2.04 % 128 18.69% 161 23.50 % Total 282 41.17% 244 35.62 % 159 23.21 % 685 100% 80 Multivariate analysis of variance (MANOVA) compared the means of the nine factors among the three clusters to determine whether their differences were statistically significant. The significant multivariate F {Pilai F = 105.67, df= 18,/? < .001) indicated that the sectors of nine means of Opinion about Sports differs among the three clusters. Table 28. Cluster Centroids, MANOVA Table, & Profiles of the Three-Cluster Solution Cluster SP CC SB RS CP AP RG AA PT Cluster 1 0.58 -0.62 1.15 0.18 1.10 -0.55 -2.37 -0.15 -0.39 Cluster 2 0.55 0.59 0.35 -0.07 -0.38 0.63 2.94 0.59 0.31 Cluster 3 -1.85 0.26 -2.60 -0.22 -1.44 0.09 0.03 -0. 56 0.26 S.S. df M.S. F Sig. Self participation in sport 719.105 2 359.553 96.734 .001 Cynicism 201.614 2 100.807 33.661 .001 Sports Benefits 1502.210 2 751.105 254.411 .001 Resource 18.428 2 9.214 2.212 .110 Community Pride 717.096 2 358.548 121.984 .001 Attitude to Pro-athletes 182.087 2 91.043 28.708 .001 Role of Government 3642.828 2 1821.414 550.139 .001 Attitude to Amateur sports 135.473 2 67.737 25.954 .001 Perceived Time cost 75.721 2 37.861 33.011 .001 Cluster 1 Cluster 2 Cluster 3 Self participation in sport Active Active Inactive Cynicism L o w High Neutral Sports Benefits Positive Neutral Negative Resource High Neutral L o w Community Pride Positive L o w negative Negative Attitude to Pro-athletes Positive Negative L o w negative Role of Government Positive Negative L o w positive Attitude to Amateur sports Less prefer Prefer Least prefer Perceived Time cost L o w High L o w high Figure 6, a K-Means bi-plot visually shows the separation of the three clusters. Follow- up ANOVAs indicated that three clusters were significantly different at the .001 level for all nine factors except for the factor 'Resource' (Table 28). In addition, Scheffe post-hoc tests were conducted to inspect cluster differences for the eight factor score means. The 81 results of the post-hoc tests were as follows: cluster 1 and 2 were different from cluster 3 in the factor 'Self-Participation in sport; cluster 2 was not different from cluster 3 in the factors 'Cynicism' and 'Perceived Time cost'; cluster 1, 2, and 3 were different from one another in the factors 'Sports Benefits', 'Community Pride', 'Attitude to Pro-athletes', 'Role of Government', and 'Attitude to Amateur sports'. Figure 6. K-Means Method Clusters Plotted on the Principal Components Variables and observations (axes F1 and F2: 43 %) 0 . 8 ; r - 0 . 6 ; u - 0 . 8 - 0 . 6 - 0 . 4 - 0 . 2 0 0 . 2 0 . 4 0 . 6 0 . 8 — a x i s F1 ( 2 6 % ) — > As can be seen from Table 28 and Figure 7, individuals in cluster 1 tended to have positive aspects to Self-Participation in sports, Sports Benefits, Community Pride, Cynicism about professional sports, Attitude toward Pro-athletes, Perceived Time cost, and Role of Government (note that the latter four variables are reverse scored, thus negative centroids reflect positive attitudes). In terms of entertainment value, people in 82 cluster 1 preferred professional sports to amateur sports. In addition, they had higher values on Resource and their values for Perceived Time cost were low. They tended to be very faithful to sports in general, especially to professional sports. Figure 7. Cluster Profiles for the Three-Cluster Solution Centroid 20.0 Cluster 1 — • — Cluster 3 18.0 16.0 / \ \ : \ \ 14.0 ' f—T—;—ft /'' - \ A \ 1; A 1 \ 12.0 X li v. / ' v. V 7 / 10.0 8.0 6.0 .) -—- \ 4.0 V- 2.0 0.0 S.P. C C . S.B. R.S. C P . A.P. R.G. A.A. P.T. Factor Individuals in cluster 2 represented negative feelings towards professional sports and athletes. With respect to Sports Benefits and Community Pride, they had less positive attitudes than people in cluster 1, but still less negative than people in cluster 3. They expressed little concern over the extent to which government was supporting local professional franchises, however their attitudes towards amateur sports were very positive. In terms of Perceived Time cost, they had the highest value among people in three clusters. Although they tended to have negative attitudes towards professional sports in general, they were the active individuals in participating in sports. 83 Individuals in cluster 3 consisted of people who had negative attitudes towards sports in general (low S.P., S.B., CP., and A.A.), but somewhat positive attitudes towards professional sports. Their low values on Sports Benefits and Community Pride indicated that they probably tended to consider professional sports to be just entertaining events. In addition, they exhibited the least interest in amateur sports, the lowest resources and interests in participating in sports among people in three clusters. Follow-up discriminant analysis of the three-cluster solution revealed that about 95.2 percent of cluster memberships were correctly classified, which suggested that the 9 factors identified from the confirmatory factor analysis were effective tools to predict cluster memberships based on their opinions about sports (Table 29). Table 29. Prediction Result from the Discriminant Analysis for Cluster Membership Predicted cluster 1 2 3 Total Actual cluster Cluster 1 278 9 4 291 Row % 95.5 % 3.1 % 1.4% 100.0% Cluster 2 7 221 5 233 Row % 3.0% 94.8 % 2.2 % 100.0% Cluster 3 4 3 154 161 Row % 2.5 % 1.8% 95.7 % 100.0% Total 289 233 163 686 Cluster membership and the frequency of game attendance. In the previous section, three distinctive spectator groups (clusters 1, 2, & 3) were identified based on their reported opinions about sports. Using a chi-square, the relationships between the three identified clusters and the frequency of game attendance was examined and the results are presented in Table 30. To conduct a chi-square test, the 84 continuous variable of spectators' frequencies of game attendance (range 0 to 45) was transformed into a categorical variable. The frequencies of game attendance were categorized into three groups as in study 1 (i.e., Low Group: 0-1 games, Medium Group: 2-4 games, High Group: more than 4 games), however, the number of cases for each category were not as symmetrical: 271 (39.5 %), 247 (36.1 %), and 167 (24.4 %), respectively. Table 30. Cluster Membership and the Game Attendance Frequency 0-1 2-4 >4 Observed total Cluster 1 Observed n (%) Row % Expected n 96(14.0) 33.0 115 107(15.6) 37.0 105 88(13.0) 30.0 72 291 (42.5) 100.0 Cluster 2 Observed n (%) Row % Expected n 95 (13.9) 41.0 92 87(12.7) 37.0 84 51 (7.5) 22.0 56 233 (34.1) 100.0 Cluster 3 Observed n (%) Row % Expected n 80(11.7) 50.0 64 53 (7.7) 32.0 58 28 (4.1) 18.0 40 161 (23.5) 100.0 Expected total 271 247 167 685 The results of chi-square test indicated that the relationship between spectators' frequencies of game attendance and 3 clusters was significant at alpha .01 with the chi- square observed value of 16.8 (df= 4, p = .003). As can be seen from the row percent values in Table 30, spectators of Cluster 1 exhibited the highest game attendance and the spectators of Cluster 3 represented the lowest game attendance. Cluster 1 contained 88 (expected n = 72) observations in the high frequency category and only 96 (expected n = 115) observations in the low frequency category. The results indicated that spectators in cluster 1 were more likely to be frequent spectators, and individuals in cluster 3 less 85 likely. Thirty percent in cluster 1 attended more than four games, compared to 18 percent in cluster 3. For the case of cluster 2, the frequency in all three cells was very close to the expected values. Profiling the cluster solution on additional variables. In the previous section, 3 distinctive clusters based on their reported opinions about sports were identified by a cluster analysis. The next step was to test the capability of implementation of the cluster measures identified by seeing if each cluster could be linked back to observable activity patterns, ticket information, or demographic variables in a meaningful way. Sports interests and activity profiles. The chi-square tests were conducted to reveal spectators' profiles along with their cluster membership. Additional variables for the chi-square tests consisted of items encompassing ticket information, ticket-purchasing patterns, best promotional medium, hobby, favourite activity, game attending colleague, membership in a club, fan history, and demographic information (items under part 2 and 3). The results of the chi-square tests and cluster descriptive statistics are presented in Tables 31 and 32. Profiles of cluster 1 contained the following characteristics (as supported by p value of less than .01): (a) the highest mean value for game attendance of 5.3 games (b) the longest fan history of 14.3 years, on average (c) tend to attend hockey games with their families (d) the highest percentage of playing hockey as their hobbies (24.7 %) 86 (e) the highest percentage of internet users (70.1 %) (f) the highest percentage of the use of hockey related websites (18.7 %). Table 31. Profiles of the Three Clusters on Sports Interests & Activities Cluster Variables Cluster 1 (n = 291) Cluster 2 (n = 233) Cluster 3 (n = 161) Mean Mean Mean Chi-square or F" P Game attendance 5.3 3.5 3.8 F=5.04 .007 Years of being a fan 14.3 13.7 10.6 F=8.25 .001 n % n % n % Ticket Purchaser Company 23 7.9 39 16.7 12 7.5 12.931 .002 Game colleague Child Family Friends 31 101 189 10.7 34.7 64.9 35 53 129 15.0 22.7 55.4 10 39 106 6.2 24.2 65.8 7.592 10.772 6.425 .022 .005 .040 Member of sport club Yes No 153 138 52.6 47.4 94 139 40.3 59.7 43 118 26.7 73.3 28.988 .001 Sport club Hockey 49 16.8 29 12.4 13 8.1 7.126 .028 Hobby Hockey 72 24.7 24 10.3 28 17.4 18.275 .001 Favourite activity Basketball 30 10.3 15 6.4 24 14.9 7.570 .023 Use of internet Yes No 204 87 70.1 29.9 134 99 57.5 42.5 110 51 68.3 31.7 9.436 .009 Favourite website Sport media Hockey related 33 55 11.3 18.9 10 18 4.3 7.7 13 17 8.1 10.6 8.566 15.386 .014 .001 Play as a child Hockey Soccer 112 100 38.5 34.4 66 65 28.3 27.9 50 64 31.1 39.8 6.488 6.210 .039 .045 Note. Chi-square orF" : Reported test statistic is X » unless otherwise noted. The complete table for the three clusters on sports interests and activities is presented in Appendix D. Profdes of cluster 2 contained the following characteristics: (a) the highest percentage (16.7 %) of tickets bought by companies (b) the lowest percentage (57.5 %) of the use of internet (c) the oldest group with mean value of 31.0 years old. 87 Profiles of cluster 3 contained the following characteristics: (a) the shortest fan history with the mean value of 10.6 years (b) the lowest percentage (26.7 %) of members of sport clubs (c) the highest percentage of Asians (30.4 %). Partial Bonferroni corrections were applied for a priori comparisons which resulted in an alpha of .01. Six predictors were not significant between groups, including accompanied by child, accompanied by friends, member of a hockey club, basketball as a favourite activity, played hockey, and soccer when young. Demographic profiles. Table 32 presents the demographic statistics for each of the three clusters. Age and ethnicity were the only variable that exhibited between cluster differences. Cluster 2 individuals were slightly older (by about 3 years) than those in clusters 1 and 3, and cluster 3 was comprised of more Asians and fewer Caucasians that the other 2 clusters. Table 32. Demographic Profiles of the Three Clusters Cluster Cluster 1 Cluster 2 Cluster 3 Variables (n = 291) (n = 233) (n = 161) Mean Mean Mean Age 27.7 31.0 27.8 n % n % n % Chi-square p Age category 28.440 .005 21 > 71 24.4 41 17.6 41 25.5 21 - 2 5 65 22.3 33 14.2 36 22.4 2 5 - 3 0 47 16.2 30 12.9 23 14.3 31 - 4 0 52 17.9 43 18.5 28 17.4 4 0 - 5 0 23 7.9 41 17.6 15 9.3 50 < 19 6.5 25 10.7 10 6.2 Missing 14 4.8 20 8.6 8 5.0 Gender 4.261 .372 Male 190 65.3 155 66.5 103 64.0 Female 87 29.9 59 25.3 50 31.1 Missing 14 4.8 19 8.2 8 5.0 Residence 12.728 .548 Area 1 70 24.1 50 21.5 48 29.8 88 Area 2 20 6.9 14 6.0 12 7.5 Area 3 43 14.8 21 9.0 16 9.9 Area 4 21 7.2 18 7.7 14 8.7 Area 5 21 7.2 14 6.0 10 6.2 Area 6 26 8.9 27 11.6 13 8.1 Area 7 33 11.3 28 12.0 16 9.9 Missing 57 19.6 61 26.2 32 19.9 Ethnicity 28.198 .001 Caucasian 224 77.0 182 78.1 101 62.7 African Canadian 2 0.7 2 0.9 1 0.6 Asian 46 15.8 29 12.4 49 30.4 Natives 4 1.4 0 0.00 1 0.6 15 5.2 20 8.6 9 5.6 Job 31.485 .211 Managerial 37 6.5 37 15.9 20 6.8 Social science 13 12.7 17 7.3 10 12.4 Business & finance 25 4.5 28 12.0 19 6.2 Sales & service 32 8.6 20 8.6 10 11.8 Applied science 14 11.0 10 4.3 5 6.2 Manufacturing 5 4.8 2 0.9 1 3.1 Retired 5 1.7 5 2.1 2 0.6 Trade & transportation 9 1.7 14 6.0 7 1.2 Primary industry 3 3.1 2 0.9 68 4.3 Student 110 37.8 59 25.3 4 42.2 Unemployed 5 1.7 6 2.6 3 2.5 Art recreation &sport 11 3.8 5 2.1 1 1.9 Health related 3 1.0 7 3.0 1 0.6 Missing 19 6.5 21 9.0 11 6.8 Household income 18.856 .276 <C$ 10,000 31 10.7 17 7.3 19 11.8 10,000-19,999 16 5.5 7 3.0 17 10.6 20,000 - 29,999 20 6.9 12 5.2 10 6.2 30,000 - 39,999 19 6.5 17 7.3 7 4.3 40,000 - 59,999 44 15.1 33 14.2 23 14.3 60,000 - 79,999 44 15.1 36 15.5 20 12.4 80,000 - 99,999 34 11.7 29 12.4 14 6.7 >C$ 100,000 52 17.9 48 20.6 27 16.8 Missing 31 10.7 34 14.6 24 14.9 Education level 20.825 .185 Some high school 8 2.7 3 1.3 4 2.5 High school degree 60 20.6 48 20.6 33 20.5 Some college 93 32.0 56 24.0 32 19.9 Two-year college diploma 29 10.0 24 10.3 19 11.8 Bachelor's degree 52 17.9 49 21.0 39 24.2 Some graduate school 12 4.1 17 7.3 5 3.1 Master's degree 12 4.1 13 5.6 16 9.9 Doctoral degree 4 1.4 3 1.3 2 1.2 Missing 21 7.2 20 8.6 11 6.8 89 Predicting Game Attendance using Decision Tree Analysis The purpose of the cluster analysis in the previous section was to identify segments underlying spectators' reported opinions about sports. Three distinctive clusters were identified and distinctive characteristics of each cluster were presented on additional variables. A decision tree analysis is similar to a cluster analysis in that both methods minimize differences within nodes or clusters, and maximize differences between nodes or clusters on a given dependent variable. However, the two methods use totally different approaches to achieve this purpose. In this study, the cluster analysis began with identifying clusters based on reported opinions about sports and predicted game attendance behaviours, and then examined the relationships between clusters and reported game attendance. On the other hand, a decision tree analysis begins with the reported game attendance behaviours (frequency of game attendance), and maximizes differences over branched nodes. Comparing the results of these two different methods of analysis has the potential to provide a richer and deeper understanding of how professional sports spectators differ and to facilitate segmenting spectators and predicting game attendance. Accordingly, a decision tree analysis was conducted with the 39 predictive variables used in the final tree analysis of Study 1 and the nine factor raw score totals identified by the confirmatory factor analysis in Study 2. This analysis was identical to the final decision tree analysis in Study 1 except that the nine opinion factor scores were added and the data were taken from data sets 2 and 3 only. The tree diagram is presented in Figure 8. 90 The full tree diagram was classified by six predictors including WEBHOCK, JOB, I DO_WBY, SPBENEF (Sport Benefit), RESICAT, and COMPRID (Community Pride). There were nine terminal nodes with three of them (nodes 2, 7, and 12) representing groups of spectators who reported attending more than 4.3 games (the mean for the total sample). Node 2 represented 90 individuals whose favourite internet websites were hockey related sites, with a mean frequency of game attendance of 8.4 games. Node 7 represented 52 individuals with jobs in sales and service who reported that they were the actual ticket purchasers. Their frequency of game attendance was 5.2 games. Node 12 consisted of 64 individuals who had jobs in management, business, trades, and the health sector or were unemployed. Their residence areas were Vancouver, Richmond, New Westminster, North Vancouver, and West Vancouver and their game attending frequency was 9.8 which was the highest of all nodes. The individuals reporting the lowest attendance (1.6 games), represented by node 9 (with a further, but trivial, breakdown to nodes 13 and 14), were characterized by a mixture of job categories (manufacturing, retired, student), were non-users of hockey websites, and had lower than average scores on Sports Benefits. The most important segment was node 12 with an index of 227.5 %, which implied that targeting an individual in node 12 would be expected to increase the frequency of game attendance 2.27 times over a random selection. If a franchise targeted people belonging to the top three segments (nodes 2, 7, & 12), the frequency of game attendance would be expected to increase by a range of 121.2 to 227.5 percent. 91 Figure 8. Tree Diagram on Activity Pattern and Opinions about Sports I DO WBY Ad). P-value=0,0477, F=4.0191, df=l,99 Node 9 Mean 3.3061 Std. Dev. 3.S67 n 49 % 7,15 Predicted 3.3061 NodeO Mian 4.3 Std. Dev. 6.9482 n 685 % 100.00 Predicted 4.332E WEB HOCK Adi. P-vakje=O.00O0, F̂ 7.3904, df-1,6 I Nodel Node2 Mean 3.7176 Mean 8,4000 Std. Dev. 6.0186 Std, Dev. 10.4889 n 595 n 90 % 86,86 % 13,14 Predicted 3,7176 Predated 8,4000 1 Z JOB Adj. P-vak]e=O.0O00, M5.8122, df-3,591 10;7;6 1 El SP BENEf Adj. P-value=0.0006, F= 17.4381, df=l,217 1 3;l;9;8;ll;13 1 1 5;2:12 Node5 Node 6 Mean 5,6650 Mean 3.2800 Std. Dev. 8.9467 Std, Dev. 3.2113 n 200 n 75 % 29.20 % 10,95 Predicted 5,6650 Predcted 3,2800 1 b RESI CAT Adj, P-vabe=0.0003, F-Z2.970I, df=l,198 1 B COM PRE Adj. P-vabe=0,0056,F=13,8370, df-1,174 J 1 <=11 1 1 >U 1 Node 13 Node 14 Mean 1.0313 Mean 1.8661 Std Dev. 1.1948 Std, Dev. 1,5509 n 64 n 112 % 9,34 % 16,35 Predicted 1,0313 Predated 1.8561 WEBHOCK was identified as the best predictor, which was consistent over both the cluster analysis and the decision tree analysis. Among the nine factors identified from the confirmatory factor analysis, Sport Benefit and Community Pride were two factors included in the tree analysis as predictors. Compared to the results of the first tree analysis of Study 1 where none of the 33 game attribute items were 92 significant, the seven identified factors from the confirmatory factor analysis in Study 2 had better predictability. 93 Chapter 5. Discussion and Implications The research for this dissertation had two inter-related objectives. One was to investigate the consumer profiles of sport spectators associated with a specific NHL franchise, the Vancouver Canucks, and to relate these to game attendance. The second objective was to develop, test and evaluate a questionnaire designed to incorporate recent research on variables that affect professional sports game attendance. These two objectives were operationalized in the questionnaire design and the research methodology of the study and were also reflected in the two studies that were reported in the previous chapter. Study 1 focused on evaluating game attribute items that had been identified as important variables affecting hockey game attendance in previous research, and also on revealing the interactions between identified game attribute items and lifestyle related variables (game attendance behaviours, sports interests, and activity patterns), and demographic information. Study 2 focused on investigating spectators' opinions and investigating the linkages between identified opinions and lifestyle related variables and demographic information. In this chapter I discuss the findings from these two studies in light of the objectives of the research and what was learned. The chapter is organized in the following sections: (a) examination of study 1, (b) examination of study 2, (c) short-term marketing implications, (d) long-term marketing implications, (e) limitations, and (f) future research. Examination of Study 1 In study 1, questionnaire validity for the 33 game attendance items was accomplished using a confirmatory factor analysis. The final model that resulted 94 contained 19 items and seven factors (Overall Performance, Uniqueness of Hockey, Facility Convenience, Schedule Convenience, Game Promotion, Advertising, and Excitement), and the model fit indices revealed a good fit to the data. Table 33 lists the mean scores of all 33 items for part 1 of the questionnaire. Table 33. Descriptive Statistics of 33 Items for Part 1 Items M S.D. 1. Behaviours of fans during games 3.44 1.16 2. Special events 2.97 1.22 3. Accompanied by friends 4.12 1.02 4. Excitement *4.34 .84 5. The Canucks' win/loss records *3.29 1.20 6. Visiting teams' win/loss records 2.76 1.17 7. The Canucks' superstars *3.41 1.07 8. Visiting teams' superstars 3.22 1.14 9. Record breaking performance of the Canucks *3.25 1.17 10. Record breaking performance of visitors 2.69 1.16 11. Rivalry of the Canucks and visiting teams 3.76 1.06 12. Frequent fights and penalties 3.12 1.28 13. Offensive output of the Canucks *4.02 .94 14. Defensive output of the Canucks *3.50 1.05 15. Closeness of competition 3.76 1.02 16. Unobstructed view from seats 4.41 .84 17. Comfortable seats *4.13 .93 18. Easy accessibility to facility *3.78 1.09 19. Availability of parking 3.38 1.33 20. Seating capacity of facility 3.09 1.27 21. Cleanliness of facility *3.82 1.02 22. Game time *3.56 1.08 23. Game day *3.36 1.14 24. Advertising through T.V. *2.46 1.14 25. Advertising through newspapers *2.39 1.08 26. Advertising through magazines *2.16 1.02 27. Advertising through radio •2.51 1.15 28. Advertising through internet *2.23 1.11 29. Direct mail & notification for ticket discount 2.20 1.10 30. Giveaway and prize 2.87 1.28 31. Ticket discount *3.94 1.14 32. Price of season ticket *3.73 1.41 33. Price of game ticket *4.27 .99 Total /V=628 Note. *: items in the final model of confirmatory factor analyses. The mean values of the 19 items were all of a magnitude of 3.0 or greater (on a 95 5-point Likert scale), except for the five items belonging to ADVT (advertising) whose values were 2.51 and less. The item, 'unobstructed view from the seats' represented the highest mean (4.41) among all 33 items. The high mean values suggest that all 33 items are important game attributes considered by spectators when they decide to attend hockey games, however, the low correlation coefficient between the criterion variables (game attendance frequencies in the last and next 12 months, and spectators' ticket information) and the seven predictive constructs indicated that the game attribute items did not distinguish spectators according to their game attendance frequencies. Thus the questionnaire may be of little use in segmenting spectators or predicting game attendance. Even if the predictability of the final model was lower than expected, it did not necessarily mean that the 14 excluded and 19 included items from the final model were not important factors in why spectators decide to attend hockey games. For instance, the mean value of the item, 'comfortable seats' was 4.13, which indicates most individuals in this study believed a comfortable seat was a very important consideration when they decided to a attend hockey game, whether they were frequent attendees or infrequent attendees. The low predictability for game attendance found in this research is not consistent with the findings from the previous studies reported in Chapters 2 and 3 that focused on minor league hockey games (Smith, 1995; Zhang et al., 1997a, 1999) or on other major professional sports (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989; Zhang et al, 1995). These studies showed statistically significant relationships between the frequency of game attendance and the identified reasons for game attendance. 96 Although profound and somewhat unexpected, the differences between the findings of the present study and earlier studies can at least partially be accounted for on the basis of three conditions which I discuss below: (a) different market characteristics or spectator characteristics between NHL and minor league hockey; (b) Canadians' unique attitudes towards hockey; (c) availability of alternative professional sporting events. First, there may well be differences between major league (NHL) and minor league hockey markets and spectators. This possibility was initially raised in Chapter 2 when I discussed the applicability of previous research on minor hockey to major league hockey. In the research done by Zhang et al. (1997a), three factors (game attributes, home team, and economic consideration) were identified as principally affecting game attendance levels of minor league hockey spectators. Zhang et al. used a multiple regression analysis, and found that a total of 28.8 percent of the variance was explained by these three factors (6 %, 14.8 %, 8 % respectively). By comparison, in this dissertation only 2.1 percent of total variance regarding the dependent variable "frequency of game attendance" was explained by two (uniqueness of hockey and game promotion) of the seven factors. The two factors (game attributes and economic consideration) from Zhang et al.'s study and the two factors (uniqueness of hockey and game promotion) from this dissertation, cover almost the same items; however, the total variance explained in the two studies (14 % versus 2.1 %) are significantly different, which implies that major and minor league markets may be quite different. Canadians' unique attitude towards hockey is a second consideration in explaining the findings from this study. In an early study focusing on the NHL, Jones 97 (1984) insisted that Canada was different in that spectators' game attendance levels were affected by the variable, 'game played or team located in a Canadian city'. Jones explained that winning-related variables were closely related to spectators' attendance levels, but they did not affect the game attendance levels of franchises located in Canadian cities. Carter (1996) reported a similar finding in his study dealing with 251 game attendees of the four Toronto professional sports franchises (CFL Argonauts, NBA Raptors, NHL Maple Leafs, MLB Blue Jays). Carter identified three distinctive clusters based on their psychographic information, and examined the linkages between cluster memberships and game attendance levels for the four professional sports. The three clusters demonstrated significant game attendance differences for three of the professional sports (Argonauts, Raptors, Blue Jays), but Carter found there was not much difference in game attendance frequencies between the three clusters for the Maple Leafs' games. Both Jones and Carter felt that their findings were related in part to Canadian culture and Canadians' unique attitudes towards hockey. The absence of a professional sport substitute may be a third factor. According to Zhang et al. (1997b), the presence of other professional sporting events, especially major league basketball and football games, affected hockey spectators' frequencies of game attendance because their seasons overlapped. Due to the absence of other substitute forms of professional sports in the Vancouver area, the spectators of the Vancouver Canucks did not have any other choices to attend other professional sporting events even when the performance of the local team was lower than what they would have liked. 98 Because of the poor predictive probability identified in study 1, a decision tree analysis was conducted in order to classify spectators and relate their characteristics to game attendance. In the final decision tree analysis in study 1,10 items (income, the use of hockey websites, played baseball when young, type of job, accompanied by co- workers, ethnic background, T.V. as the best promotional medium, fan history, and played hockey when they were young) were identified as the most important predictors that discriminated spectators based on their game attendance frequencies. The use of professional hockey-related websites was identified as one of the most important predictors for discriminating spectators according to their game attendance levels. Duncan & Campbell (1999) have noted that sport organizations could achieve many competitive advantages by identifying appropriate uses of the internet for their customers including improved customer service, improved advertising, creation of a distribution channel, better communications, low cost entertainment, and better information management. According to CyberAtlas (2003), 580 million people worldwide used the internet in 2002, and this number will reach 709.1 millions in 2004. In the case of Canadians, 16.84 million people (52.8 % of 31.9 million Canadians) used the internet in 2002 with internet use reported as highest in Alberta and British Columbia with a rate of 61 percent of the population. In this study, 703 (63.3 %) of 1,110 individuals were identified as internet users, and 152 (21.6 %) of the 703 individuals replied that professional hockey-related websites were their favourite sites. Use of the internet was significantly related to the frequency of game attendance. Internet users had higher game 99 attendance (mean score 5.5 games) than non-users (4.2 games). In addition, use of the internet also showed significant relationships with age, income, and education level. In this study, participants in the younger age groups and those with higher incomes and more education reflected higher internet use rates. With respect to participants' personal sports histories, two items, playing baseball or playing hockey when young, were identified as important predictors for discriminating between spectators based on their game attendance frequencies. Overall, 377 individuals (34.0 %) reported that they played hockey when they were young and the mean attendance frequency of this group was higher than those of other sport groups (soccer, 365; baseball, 277; basketball, 201). This finding is reminiscent of the niche analysis research done by Milne et al. (1996). They evaluated niche market overlap between four spectator sports (college basketball, pro basketball, golf, and tennis) and three participant sports (basketball, golf, and tennis) using both demographic and psychographic criteria. Whereas the demographic criteria (age, gender, education, income and race) showed significant overlap for sports participant and spectator niches (e.g., 84% of the golf participant niche overlapped the golf spectator niche, meaning golf spectators tended to be demographically similar to golf players), the psychographic criteria showed higher correlations among the participant sports and among the spectator sports than between (e.g., golf and tennis participants overlapped 74% psychographically and spectators 92%, but golf participants only overlapped 31 % psychographically with tennis spectators and 28% with golf spectators). This can partially be explained by the psychographic criteria used in their study which focused on personal motivation factors 100 for why people participate in activities (Personal Improvement, Sport Appreciation, and Social Facilitation) making similarities between sports participants more likely than between participants and spectators. Nevertheless, even if a direct comparison between their study and the research reported in this dissertation is not appropriate due to the different research methods and samples, a similar pattern between hockey spectators and participants is evident. The limited niche overlap between hockey spectators and participants in the dissertation findings indicates that hockey spectators did not need to be hockey participants. That only 343 (30.9 %) of 1,110 spectators reported they played hockey as a favourite activity and that only 377 spectators (34 %) replied that they played hockey when they were young, confirmed the lack of a strong niche overlap between hockey spectators and participants. On the other hand, some of the remaining 767 spectators (69.1 %) who did not list hockey as a favourite activity, reported playing other sports, and in this sense they are active sport participants. For example, among other sports activities currently pursued golf (n = 178; 16 %) was the second most reported activity. Ironically, in terms of game attendance levels for Canucks games, spectators who reported that they currently played golf had a marginally higher game attendance average (6.3 games) than spectators who reported playing hockey (6.0 games). With respect to participants' sports history profiles in their youth, hockey (n = 377, 34 %, m = 5.8) was the most and baseball (n = 277, 25 %, m = 5.7) the second most reported sport played by participants when they were young. While hockey is perhaps understandable, it is not clear why baseball ranked so highly. Perhaps this is 101 simply a factor of the popularity of baseball as an organized youth sport, however, it would be interesting to explore this association more closely. Compared to other participant sports including soccer (n = 365, m = 4.8), basketball (« = 201, m = 4.8) football (n = 122, m = 5.4), and volleyball (n = 102, m - 5.2) having a background of playing baseball correlated with a high Canucks game attendance frequency, second only to playing hockey. This study also found that fan history is positively related to frequency of game attendance. Spectators who attended more Canucks hockey games tended to have longer fan histories than those attending fewer games. This pattern was consistent from the youngest age group to the oldest group when all other item effects were controlled. With respect to promotional media, the ranking of newspapers as an information source was seen to vary according to participants' game attendance levels. Although TV was considered as the best medium among all spectators in this study, more of the spectators who attended a greater member of hockey games tended to believe newspapers were the best medium than those who did not. As indicated in the previous chapter, 48 and 50 percent of individuals in the Low and Middle game frequency Groups believed TV was the best promotional medium. In the case of the High game frequency Group, 37 percent of individuals identified TV and 35.1 percent identified newspapers. This pattern reinforces the notion that television is a low- involvement medium because it requires only a passive viewer who exerts relatively little control over content whereas print media requires higher involvement, i.e., highly involved fans tended to use print media. 102 Income was also an important predictor for discriminating between spectators based on their game attendance levels. Spectators who attended more hockey games tended to have higher income levels than those who did not. This result was in-keeping with previous studies (Greenstein & Marcum, 1985; Hansen & Gauthier, 1989; Zhang et al., 1995). With respect to game attendance colleagues, spectators who attended more games tended to go with their co-workers than those who attended fewer games. Examination of Study 2 In study 2, an evaluation of the questionnaire validity of 33 spectators' opinion items was accomplished using a confirmatory factor analysis. The final model contained 22 items and nine factors (Self participation in sport, Cynicism, Sports Benefits, Resource, Community Pride, Attitude towards Pro-athletes, Role of Government, Attitude towards Amateur sports, and Perceived Time cost). The model fit indices indicated a good fit for the final model. Based on the identified nine factor score totals, a cluster analysis was conducted to uncover segments underlying the survey participants' reported opinions about sports. The cluster analysis yielded three distinctive clusters with 291, 233, and 161 respondents in clusters 1, 2 and 3, respectively. Cluster 1 differed from cluster 2 along 5 dimensions including Cynicism, Attitude towards Pro-athletes, Role of Government, Attitude towards Amateur sports, and Perceived Time cost. While individuals in cluster 1 expressed low cynicism, positive attitudes towards professional athletes and towards the role of government in sports, less preference for amateur sports, and less perceived time cost, individuals in cluster 2 demonstrated high cynicism, negative feelings towards 103 professional athletes and the role of government in sports, high preference for amateur sports, and greater perceived time cost. Individuals in both clusters were active sport participants. Unlike the other two clusters, individuals in cluster 3 were not active in participant sports. In addition, individuals in cluster 3 demonstrated negative reactions to benefits given by sports and to the idea of community pride through sports, the lowest resources, and the lowest preference for amateur sports among individuals in all three clusters. In terms of values on other factors, cluster 3 lay between clusters 1 and 2 on Cynicism, Attitude towards Pro-athletes, Role of Government, and Perceived Time cost. Regarding a link between the clusters and the frequency of game attendance, individuals in cluster 1 were more likely to be frequent spectators. On the other hand, individuals in cluster 3 were less likely to be frequent spectators and cluster 2 was in between cluster 1 and 3. The three clusters were profded using additional variables including activity patterns, fan history, and demographic information, as noted below. Cluster 1 Compared to clusters 2 and 3, spectators in cluster 1 had the longest fan history and this was positively related to the frequency of game attendance. In terms of game colleagues, they tended to attend games with their family, and more spectators in cluster 1 tended to be members of sport clubs than in the other clusters. Among a variety of sport clubs, hockey clubs were a favourite for this group and their most popular activity was playing hockey. More spectators in cluster 1 tended to use the internet and this was 104 positively related to the frequency of game attendance. The differences between cluster 1 and the other two clusters on the use of the internet were in the use of sport media and hockey related websites. More spectators in cluster 1 tended to use sports-related media and to identify hockey websites as their favourite sites than the other two clusters. Among spectators in cluster 1, hockey was the favourite sport to play when they were young. Cluster 1 was also marginally the youngest group (average 27.7 years old) compared with the other two clusters (cluster 2, 31.0 years; cluster 3, 27.8 years). Although their income levels were not high, individuals in this cluster highly valued attending hockey games and considered that their income levels were high enough to justify the cost. Similarly, they felt that the time spent watching professional sports was justifiable. Cluster 2 Spectators in cluster 2 were the second most frequent game attendees compared to cluster 1. In terms of ticket sources and purchasing, more spectators received their tickets from companies than spectators in clusters 1 and 3, which indicates that there was a possibility for them to use hockey games for their business purposes. More spectators in cluster 2 tended to attend hockey with their kids or co-workers than spectators in the other clusters. Fewer people in this cluster used the internet compared to the other clusters, and they were the oldest with an average age of 31 years. Spectators in cluster 2 had relatively higher household income levels than spectators in the other two clusters. Although they had relatively higher income levels than people in cluster 1 and 3, their perceived time cost was the highest among the three clusters. Due to their 105 negative attitudes towards professional sports and athletes, they probably tended to attend hockey games to spend time with their kids or for business purposes. Cluster 3 Spectators in cluster 3 had the shortest fan histories and the lowest household income levels among the three clusters. Due to their very negative attitudes towards the benefits of sports, they tended to be very inactive in participating in sports, and they had slightly higher education levels than people in the other clusters. In terms of ethnic background, cluster 3 contained more Asians than clusters 1 and 2. Due to the lowest resource value (lower income and having less time to watch hockey games), spectators in cluster 3 tended to have the highest perceived time cost. In the decision tree analysis (Figure 8) in study 2, six predictors including the use of hockey related websites, job, ticket buyer, residence areas, sports benefit, and community pride were identified as significant predictors related to spectators' game attendance levels. The result of the tree analysis was not consistent with the result of the cluster profiling using chi-square analysis, however, due to the different algorithms in the two methods, this is not unexpected. Nevertheless, it is possible that some important interactions between items were missed in the decision tree analysis due to the small sample size (n = 685) and the stopping rules that were applied. As described earlier, one of the accepted stopping rules in decision tree analysis is that a parent node must contain a minimum of 100 cases and a branch node must contain a minimum of 40 cases. Using this rule, smaller segments are lost from analysis even though they might have descriptive and explanatory power. 106 Figure 9. Tree Diagram on Activity Pattern and Opinions about SportsOl ten (,H28 Si Da 6.9B % iD0.cn Pedrtal t.B281 'tai Iede2 ta inn Ban IflO si in, t.sm Sol Civ. 10.(889 n 595 n * i te.tt % lilt Mttd I.M tailed E.tOOO 1 tl 1 bl 1SP50CC Met Weio ta 34667 ta $933) adDev. 3.8411 Std Dev. 7.58M n II n 33 » US? % 2.92 Piloted 16657 neided iW COM HI AAf̂ â M27,F=!a.l838.dh=l,n 1 (=12 1 >12 1 'oil? •ta3 Ban 12222 Heai 43222 Site 1.8839 9J.Hn. unt- i 36 il 4S 1 US % i.S7 Ftete) 2.2222 Predcted 41222 M l â ,IO)l.H5,ll22,rW591 4)P'.i«jeJJH!F=3tES,d=!,!J 3;ttlii;i3 IM1 Met Men 13359 ta l»4 Std.Dev, 4,9(32 Scd.Dw. 1M48 0 1DI 1 21! % H.74 « 31.97 Fredcted 4333) Predcted 111? 1 tJ 1 H Hean m StiDev, 8,9*7 % 29,20 MM S.6650 fti^«w.qii,iNia; man »hlW«,W2.911111rf=l1 9l Well Ken 1.5625 StiDev. 1.4831 1 US * 2S.S9 PailsJ I E Hen 36912 Std.De>. 'Kl S MS Predkted 3,6912 HHCOB tthaWoi<,M53r2,rHm I ' 1 ttP^.[iaHI«,fNl,l3l ta 128X Stito. 1.2113 n 75 * MS PedcW 3.280Q run nan StiDev. 16.6318 u Hear 8 S.58S7 Std. to. 5.43S7 » « 1022 Predcted SSSSJ 001CCIC tt,P^,IT3iw,SB9),lt=l,7i T OJIOX M,haW,ttnS>llTOo'-l,6l J_ I I ite;: lt*2 Mean 1.4359 real 2.59)1 Stito. 1.4113 Std. Dev. 1.7614 n 1SS n 23 1 22.77 Vi 2.92 PteJtEd 1.43S9 Predcted 2.5503 Mite 23 'toil Men 3.1(65 Mean 6.8500 SiDev, 39112 Stito. 75343 ll 116 i 28 » 16)3 * 2.92 Ptedttaj 3.1(65 Predcted t.3500 PHU1CE «.haW.0u?ZHS.!3S31iRll4 -r—B 1HCCJSI ttPWaW.0442 W,52II,rf=l,!« I 1 Me 15 Heai 2J50D StiDev. 23563 1 52 « 759 PYedcted 2.75O0 loill Heai 4.4783 5* to. 44300 D 21 % 1.36 Predcted 4.4783 Weli Heai (.O ta 8«S! Stito. 4S512 Stitev, S265( r, (7 n 21 ". lit. » 336 Predcted (.0633 Petal SflS? ICMH toPvaWmMMziso Mode2S Nrjde26 Mean 2.0000 Mem 13(48 SUKV, um Stitev, 17552 i 23 n 29 % 3.36 * 4.23 ftedced 2.KCD Piedctej 334(8 loi?) tt*3! Men 1.0488 ta 1.9216 Std.Dev. 2.5194 Std Dev. 1.3512 n (1 n SI % 5.99 * 7(5 MM 3.D488 fete 1.9216 107 To check for the possibility of more detailed interactions among items, a further tree analysis was conducted after the main analysis reported in Chapter 4 using a less restrictive stopping rule with the software set for a minimum of 50 cases for a parent node and a minimum of 20 cases for a branch node. A diagram of the tree analysis is presented in Figure 9. Although the result is less reliable than that of the previous tree analysis, it is more revealing and corroborates the findings from the cluster analysis. Four predictors (WEB_HOCK, HHOCKEY, ACCWORK, & ACFRIEN) show significant differences among the three clusters, and four factors (Sports Benefit, Resources, Community Pride, and Perceived Time cost) are identified as significant predictors, thus supporting the effectiveness of the discriminant functions of the nine factors originally identified in study 2. In this dissertation, the main focus was on uncovering factors affecting ticket purchases. As described in Chapter 2, the two theories, the theory of reasoned action and means-end theory, provided an important basis for understanding hockey spectators, however, the purpose of this dissertation was not to evaluate these two theories, but to construct a questionnaire and conduct a survey to evaluate current research and thought concerning factors that influence game attendance. Using one (attitude toward buying) of three core concepts of the theory of reasoned action, game-attending spectators were surveyed and evaluated based on 33 game attribute items (part 1 of the questionnaire) to identify attraction factors and potential issues in game attendance. As explained above, the results show that all 33 game attribute items were important for the respondents in their decisions to attend games, but there was no statistically significant correlation 108 between the 33 items and game attendance frequencies. With respect to the concepts of means-end theory, spectators in this research appeared to attend hockey games as a means of obtaining pleasure and personal value from the product (the game itself) and its associated attributes and consequences (e.g., source of entertainment, social atmosphere, family fun, save money from game promotion, save time via the convenient game schedule and easy accessibility to facility). Although the opinions and attitudes findings support means-ends theory as a framework and reflect a variety of possible means-ends chains and end values (e.g., sense of belonging, sense of accomplishment, security, fun and enjoyment in life, excitement, self-respect), defining respondents' personal values in detail using means-end theory was beyond the scope of this study. In both study 1 and 2, meaningful predictive variables and distinctive clusters were identified that relate to the frequency of game attendance levels. Based on the predictive variables for game attendance, spectators' attitudes towards sports, and their lifestyle profiles found in these analyses, several short and long term marketing strategies can be identified that could help professional hockey clubs better identify and retain their core spectators (frequent attendees), increase game attendance levels of infrequent attendees, and attract new spectators. Short-term Marketing Implications To retain loyal spectators and to increase the game attendance levels of infrequent spectators, several short term marketing strategies can be undertaken focusing on the internet, ticket prices, and image modification through supporting the local community. 109 The Use of Internet In study 1 and 2, the use of internet and particularly hockey related websites and sports-related media sites were strongly related to the frequency of game attendance. Because British Columbia together with Alberta ha ve the highest internet use rates in Canada, the internet potentially is an important marketing channel for reaching fans. One strategy would be to promote the existing website of a franchise (such as the Canucks) in order to increase the number of regular users. A second would be to redesign the site to attract a broader user group. In either case, the site could be used to advertise ticket promotions and sell tickets, and to collect information about fans, spectators and site users. This information could then be used to improve communications and services by identifying what the highly committed fans already using the website want and need, and, where possible, providing these services through the website or improving extant services. According to CyberAtlas (2003), about 84 percent of internet users connect to email, and many use it daily. In this study, approximately 40 percent of the participants who used the internet identified an email related website (e.g., Yahoo, MSN, Hotmail) as a favourite website. This is potentially valuable information for a professional sports franchise or league. For example, some fans might welcome having an email account offered through the Canucks' website. Offering a franchise-branded service might increase the number of spectators visiting the site, the number of overall site visitors, and the length of time visitors spend on the site. It might also help convert some occasional spectators into fans. 110 Ticket Prices Ticket prices have long been an issue in professional sports markets and a number of spectators in this study reported that ticket prices and ticket discounts were very important criteria in their decision to attend hockey games. Ticket prices logically are more sensitive for spectators whose income levels are relatively low or spectators who have high perceived time costs. For the case of spectators in cluster 2 who had relatively higher income levels, but represented a higher perceived time cost, marketers could try to change their negative perceptions of the value of hockey games and tickets. As described earlier, spectators in cluster 2 tended to go to games with their co- workers and to obtain more game tickets from their companies than other people did. For this group of people, marketers need to improve the value of hockey games from a business standpoint and communicate this possibility to business spectators. One way to do this would be to provide examples of companies that successfully use hockey games for business. Considering spectators' reported activity patterns, marketers might consider promoting ticket packs jointly with other sports activities. For instance, 178 of the 1,110 survey participants identified golf as a favourite sport activity. These individuals had an average game attendance of 6.3 games. There would be a possibility of attracting more golfers who like to attend hockey games, if a joint ticket package (playing golf and watching hockey games) could be provided at a reasonable price. Perceptions of Pro-sports and Athletes One of the major differences between cluster 1 and the other two clusters lay in i l l their perceptions of professional sports and athletes. Spectators who had positive attitudes towards professional sports and athletes tended to have higher game attendance levels in this study, and vice versa, spectators who had negative attitudes towards professional sports and athletes attended fewer more games. This raises the possibility that some of the more negative spectators might attend more games if their perceptions could be changed to be more positive. Sports marketers are well aware that they need to build a positive public image of their athletes and their sport, however, this finding identifies a concrete relationship between negative perceptions and attendance rates among people who are already attending games. In response, marketers might consider a series of campaigns that build public good will and communicate good will via various promotional media. Television would be particularly important in this mix, because it was identified as the most preferred promotional medium among infrequent spectators. In addition, supporting local amateur sports could be helpful for building positive perceptions of professional athletes and the sports organization. If an NHL franchise decided to support a local amateur sports team or league, the sports fixture would not need to be hockey or hockey-related. As noted above in Chapter 4, spectators in this study included people who had a variety of sports backgrounds and activity patterns, including golf and tennis, and, in their youth at least, baseball. Long-term Marketing Implications Fan history was strongly connected to frequency of game attendance levels in both studies 1 and 2. One way to increase fan association with a professional hockey franchise over time would be to support local youth's sporting events. As just noted, the 112 youth's sports would not need to be hockey-related. Involvement with youth's sports would not only reach the youth themselves who are potential future spectators, but also their parents and the sports officials (coaches, leaders, members of the youth sports organization). Further to establishing long-term relationships with the young people who are the fans of the future, supporting youth sports might lead adults to develop more positive perceptions of professional sports, the franchise and the athletes which potentially could result in higher game attendance rates. Attracting spectators to the Canucks' website and providing email accounts to the public also has the potential to build long-term relations with fans which, if handled properly, could lead them to become loyal customers. Through the internet, fans can obtain their needed information and marketers can create a database of their customers via on-line surveys and from information collected directly from their site-use patterns. Based on a database of customers, marketers could track market trends and set both short and long-term marketing plans as appropriate. With attractive domain names such as canucks.com or gocanucks.com, Canucks email users, especially young users, could feel that they belong to a fan community whenever they use their emails and that they are also supporting their team. In addition, this might lead them to form positive perceptions about hockey and become members of a larger hockey community. According to Zhang et al. (1996), knowledge of hockey was a significant predictor of hockey game attendance. Using the internet, marketers can reach both current and potential spectators, and they can educate them and improve their knowledge and understanding of hockey, which may help infrequent spectators to 113 become frequent spectators. Exploiting local communications channels to foster strong associations between the franchise and the local community could reduce the number of people who hold negative attitudes towards community pride and increase the number of loyal fans. This study evaluated the validity of a spectator questionnaire containing variables identified in previous research as affecting spectators' game attendance. It also investigated the demographic and psychographic profiles of spectators of a major league hockey franchise, the Vancouver Canucks, and related these findings to game attendance. The study was intended to make a novel contribution to the literature and hopefully on a small scale it has done so. Little previous research had focused on profiling NHL fans, and none had been undertaken to holistically test the different attendance-related variables identified in prior research using confirmatory factor analysis. This step has now been taken, although the study had limitations and there is much more that can be done. Limitations Due to several limitations of this study, a number of important issues regarding professional sport markets were not examined. 1. Due to the large number of questionnaire items and limited time for data collection, the questionnaire was split into two different forms for the majority of respondents. This limited the examination of the relationships among some of the variables. 2. Questionnaire items concerning game attributes and opinions about sports were 114 derived from previous studies, and therefore the results of this dissertation research was constrained by these variables. 3. Because this study was conducted in a single market area (Vancouver, a western Canadian city) and the participants were all spectators at games of only one NHL franchise (Vancouver Canucks), the generalizability of the findings is limited. Future Research 1. Due to the differences between the findings of this study and previous research in the United States, future research should focus on differences between spectators in different regions and investigate and compare a variety of market environments in terms of cultural differences, minor or major league games, and market competition. In particular future research should be expanded to various regions of Canada and the United States to investigate the cultural differences between Canadian and American hockey spectators. With respect to the difference between minor and major league games, future study should investigate spectators of both minor and major league games to look for possible differences. In terms of market competition, future study should compare the profiles of spectators living in a more versus less completive market area to uncover if there are any differences between spectators from these two regions. 2. A future study should utilize a qualitative research design to further identify psychographic information and factors related to consumer buying behaviours. As described in the result chapter of this study, there were no significant relationships between the 33 game attribute items and spectators' game 115 attendance levels, which meant game attribute related items did not predict well spectators' game attendance levels. For instance, to identify spectators' psychographic informational dimensions, qualitative research including focus groups and comprehensive interviews with spectators should be conducted. According to Dick and Sack (2003), online ticket sales using the internet were starting to be viewed as an effective tool by NBA marketing directors in the middle 1990s, and have since become an integral part of the ticket sales process. The data they report demonstrate the importance of the use of internet in ticket sales, in that online ticket sales in the NBA rose 48 percent from the 1999-2000 season to 2000-2001. As indicated earlier, the use of hockey-related websites and sports-related websites were identified as important predictors affecting spectators' game attendance levels in this study. Duncan and Campbell (1999) outlined six fundamental functions for use of the internet that could help set a firm's marketing strategy for satisfying the wants and needs of consumers. The six functions for using the internet were improving customer service, improving advertising, creating a distribution channel, handling communications, providing low cost entertainment, and managing information systems. A future study should not only focus on marketing strategies using the internet for professional sport consumers, but also evaluate the effectiveness of those marketing strategies particularly with respect to increasing game attendance levels. 116 References Ajzen, I., & Fishbein, M. (1975). Belief, Attitude, intention, and behaviour: An introduction to theory and research. Reading, MA: Addison-Wesley. Baade, R.A., & Tiehen, L.J. (1990). An analysis of major league baseball attendance, 1969-1987. Journal of Sport and Social Issues. 14(1), 14-32. Badenhausen, K., & Nikolov, C. (1997). Sports Valuations: more than a game, Financial World, June, 17, 40-50. Badenhausen, K., & Stanfl, R. (2000). Team Valuations: name games, Forbes, 00156914, 12/11/2000, 166(15). Badenhausen, K., & Ozanian, M. K. (2002). Hockey Valuations: wily Coyotes, Forbes, 8647022, 12/23/2003, 170(13). Baumgartner, T. A., & Jackson, A. S. (1987). Measurement for evaluation in physical education and exercise science. 3rd edition. Dubuque, IA: Wm. C. Brown Publisher. Beckman, M. D., Kurtz, D. L., & Boone, L. E. (1992). Foundation of marketing, 5th Canadian edition. Toronto, ON: Holt, Rinehart and Winston of Canada, Limited. Biggs, D., de Ville, B. & Suen, E (1991). A Method of Choosing Multiway Partitions for Classification and Decision Trees. Journal of Applied Statistics, 18, 49-62. Bird, P. J. (1982). The demand for league football. Applied Economics, 14, 637. Canfora, J. L. (2001). NHL's state of flux: Teams signed superstars, salaries escalated and a potential work stoppage is looming. Washington Post, Saturday, September8, 2001, DO 1 Carter, R. H. J. (1996). A psychographic segmentation analysis of Toronto professional sports consumers. Master's Thesis, University of Guelph, 1996. Cortina, J. M. (1993). What is coefficient alpha? An examination of theory and application. Journal of Applied Psychology, 78 (1), 98-104. CyberAtlas. (2003). http://www. cyberatlas, com 117 Dick, R. & Sack, A. L. (2003). NBA Marketing Directors' Perceptions of Effective Marketing Techniques: A Longitudinal Perspective. International Sports Journal, 7 (1), 88-99. Dzewaltowski, D. A. (1989). Toward a model of exercise motivation. Journal of Sport & Exercise Psychology, 11, 251-269. Eherler, D. & Lehmann, T. Responder profiling with CHAID and dependency analysis. http://www. luc. ac. be/iteo/articles/lehmann.pdf. Fishbein, M. (1963). An investigation of the relationships between beliefs about an object and the attitude toward that object. Human Relations, 16, 233-240. Gauthier, R. & Hansen, H. (1993). Female spectators: Marketing implications for professional golf events. Sport Marketing Quarterly, 2(4), 21-28. Greenockle, K. M., Lee, A. A. & Lomax, R. (1990). The relationship between selected student characteristics and activity patterns in a required high school physical education class. Research Quarterly for Exercise and Sport, 61(1), 59-69. Godin, G. & Shephard, R. J. (1985). Psycho-social predictors of exercise intentions among spouses. Canadian Journal of Applied Sport and Science, 10(1), 36-43. Greenstein, T. N., & Marcum, J. (1981). Factors affecting attendance of major league baseball: I. Team performance. Review of Sport & Leisure, 6(2), 21-34. Greenstein, T. N., & Marcum, J. (1985). Factors affecting attendance of major league baseball: II. A Within-Season Analysis. Sociology of Sport Journal, 2, 314-322. Gullickson, H. (1950). Theory of mental tests. New York: John Wiley. Hair, J. f, Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate Data Analysis, 5th edition. Upper Saddle River, NJ: Prentice Hall. Hansen, H., & Gauthier, R. (1989). Factors affecting attendance at professional sport events. Journal of Sport Management, 3, 15-32 Hansen, H. & Gauthier, R. (1993). Spectators' views of LPGA golf events. Sport Marketing Quarterly, 2(1), 17-25. Hill, J. R., Madura, J., & Zuber, R. A. (1982). The short run demand for major league baseball. Atlantic Economic Journal, 10(2), 31. 118 Jones, J. C. H. (1984). Winners, losers and hosers: Demand and survival in the National Hockey League. Atlantic Economic Journal, 12(3), 54. Jones, R. A. (1991). Enhancing marketing decisions using conjoint analysis: an application in public leisure services. Society and Leisure, 14(1), 69-84. Johnson, J. L. & Plake, B. S. (1998). A historical comparison of validity standards and validity practices. Educational and Psychological Measurement, 58 (5), 736- 755. Joreskog, K. G., & Sorbom, D. (1999). LISREL (Version 8.3) [Computer software]. Lincolnwood, IL: Scientific Software International. Kahle, L. R., Kambara, K. M., & Rose, G. M. (1996). A functional model of fan attendance motivations for college football. Sport Marketing Quarterly, 5 (4), 51-60. Klenosky, D. B., Gengler, C .E., & Mulvey, M. S. (1993). Understanding the factors influencing ski destination choice: A Means-End analytic approach. Journal of Leisure Research, 25 (4), 362-379. Lascu, D. N., Giese, T. D., Toolan, C , Guehring, B., & Mercer, J. (1995). Sport involvement: a relevant individual difference factor in spectator sports. Sport Marketing Quarterly, 4 (4), 41-46. Laser, W. (1963). Life Style Concepts and Marketing, in Toward Scientific Marketing, ed. Stephen A. Greyser, Chicago: American Marketing Association, 130-139. LeBrun, P. (2003). NHL skating on thin financial ice: TV deal, CBA, skyrocketing salaries to blame for mess. The Halifax Herald, Saturday, January 11, 2003. Levy, S. (1963). Symbolism and Life Style, in Toward Scientific Marketing, ed. Stephen A. Greyser, Chicago: American Marketing Association, 140-150. Magidson, J. (1993). SPSS for Windows CHAID Release 6.0, Chicago, IL: SPSS Inc. Marcum, J. P., & Greenstein, T. N. (1985). Factors affecting attendance of major league baseball: II. Within-season analysis. Sociology of Sport Journal, 2, 314-322. Marsh, H. W., Hau, K. T., Balla, J. R., & Grayson, D. (1998). Is more ever too much? The number of indicators per factor in confirmatory factor analysis. Multivariate Behavioral Research, 33 (2), 181-220. Mawson, M. L. & Coan, M. S. (1994). Marketing techniques used by NBA franchises to promote home game attendance. Sport Marketing Quarterly, 4(1), 37-45. 119 Messick, S. (1989). Validity. In R. L. Linn (Ed.). Educational measurement (3 ed., pp. 13-103). New York: American Council on Education and Macmillan. Messick, S. (1995). Validity of psychological assessment. American Psychologist. 50. 741-749. Milne, G. R., Sutton, W. A., & Mcdonald, M. A. (1996). Niche analysis: a strategic measurement tool for sport managers. Sport Marketing Quarterly, 5(3), 15-22. Moore, D. G. (1963). Life Styles in Mobile Suburbia, in Toward Scientific Marketing, ed. Stephen A. Greyser, Chicago: American Marketing Association, 151-163. Mullin, B. J., Hardy, S., & Sutton, W. A. (1993). Sport marketing. Champaign, IL: Human Kinetics. Nunnaly, J. (1978). Psychometric theory. New York: McGraw-Hill. Pedhazur, E. J. & Schmelkin, L. P. (1991). Measurement, design, and analysis: an integrated approach. Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers. Phillips, R. D. (1992). Reinventing the wheel: on the wisdom of constructing new measures in psychosocial research. CHC, SUMMER, 21(3), 140-147. Plummer, J. T. (1971). Life Style Patterns and Commercial Bank Credit Card Usage. Journal of Marketing, April 1971, 35-41 Plummer, J. T. (1974). The concept and application of life style segmentation. Journal of Marketing, January 1974, 33-37 Rokeach, M. J. (1973). The Nature of Human Values. New York: Free Press. Schofield, J.A. (1983). Performance and attendance at professional team sports. Journal of Sport Behavior, 6(4), 196-206. Schutz, R. W. (1998). The big five in sport psychology questionnaires: valid, varified, varacious, versatile & visionary? or vacuous, vaporous, vulnerable, voluminous & vague?, AAASP 1998. Smith, R. E., Schutz, R. W., Smoll, F. L. & Ptacek, J. T. (1995). Development and validation of a multidimensional measure of sport-specific psychological skills: The Athletic Coping Skills Inventory-28. Journal of Sport & Exercise Psychology, 17,379-398. 120 Smith, S. M. (1995). Meltdown in marketing professional ice hockey: a survey exploring geographical difference in strategy. Sport Marketing Quarterly, 4(3), 17-23. Solomon, M. R. (1996). Consumer behavior 3rd edition. Englewood Cliffs, NJ: Prentice Hall. SPSS White Paper (2002). AnswerTree Algorithm Summary. SPSS Inc. Wells, W. (1988). Chapter 17, Psychographics: A critical review, Marketing classics edited by Enis, B. M. & Cox, K. K. Needham. MS: Allyn and Bacon, Inc. Wells, D. W. & Tigert, D. J. (1971). Activities, Interests and Opinions. Journal of Advertising Research, 11(4), 27-35 Whitney, J.D. (1988). Winning games versus winning championships: the economics of fan interest and team performance. Economic Inquiry, 26, 703-724. Zak, T.H., Huang, C.F., & Siegfried, J.J. (1979). Production efficiency: the case of professional basketball. Journal of Business, 52(3), 19-23. Zhang, J. J. (1998). Preference of spectators towards professional hockey game time. International Sports Journal, 2(1) Winter, 57-70. Zhang, J. J., Pease, D. G., Smith, D. W., Lee, J. T., Lam, E. T. C , & Jambor, E. A. (1997a). Factors affecting the decision making of spectators to attend minor league hockey games. International Sports Journal, 1(2) Summer, 39-49. Zhang, J. J., Pease, D. G., Smith, D. W. & Lam, E. T. C. (1998). Dimensions of spectator satisfaction toward support programs of profession al hockey games. International Sports Journal, 2(2) Summer, 2-17. Zhang, J. J., Pease, D. G., Lam, E. T. C , Bellerive, L. M., Pham, U. L., Williamson, D. P., Lee, J. T. & Wall, K. A. (2001). Sociomotivational factors affecting spectator attendance at minor league hockey games. Sport Marketing Quarterly, 10(1), 43-56. Zhang, J. J., Pham, U. L., Lam, E. T. C. & Kim, J. J. (1999). Variables affecting spectator enjoyment of minor league hockey games. 1999 NASSM Conference. Zhang, J. J., Smith, D. W., Pease, D. G., Hui, S. C , & Michaud, T. J. (1995). Variables affecting the spectator decision to attend NBA games. Sport Marketing Quarterly, 4(4), 29-39. Zhang, J. J., Smith, D. W., Pease, D. G., & Jambor, E. A. (1997b). Negative influence of 121 market competitors on the attendance of professional sport games: the case of a minor league hockey team. Sport Marketing Quarterly, 6(3), 31-40. Zhang, J. J., Smith, D. W., Pease, D. G., & Mahar, M. T. (1996). Spectator knowledge of hockey as a significant predictor of game attendance. Sport Marketing Quarterly, 3(3), 41-48. Zikmund, W. G. (1997). Exploring marketing research. 6th edition. Orlando, FL: Dryden Press. 122 Appendix A Questionnaire 123 The University of British Columbia U B C NHL SPECTATOR SURVEY This questionnaire is intended to investigate NHL spectators' profiles (activities, interests, and opinions). It takes about 20 minutes to complete. 124 Dear Madame/Sir: The purpose of the survey is to describe the activities and interests of hockey fans and poll opinions about why people attend NHL games. You have been randomly selected as a Canucks spectator. If you are an adult (19 years of age or older) and willing to participate in the survey, you will need to fill out as much of the questionnaire as you can. Please don't put your name on the questionnaire. In accordance with UBC policy, all paper materials and data files will be maintained anonymously in locked files for five years and then destroyed by shredding (paper) and erasing (data files). Participation in the survey is strictly voluntary, and you may withdraw at any time. If the questionnaire is completed and returned, it will be assumed that you have given consent to participate in the survey. If you have any questions about this study, please contact me at 822-9156 or Dr. Robert Sparks at 822-6515. If you have any concerns about your right or treatment as a research subject, you may contact Dr. Richard Spratley, Direct of the UBC Office of Research Services and Administration, at 822-8598. You may separate this letter from the questionnaire and keep it for your records if you wish. Thank you very much for your help! Sang Woo Bae, Graduate student Robert Sparks, Ph.D., Supervisor 604-822-9156 / sbaeCqjwIerchange.ubc.ca • 604-822-6515 / robert.sparks(ql ubc.cu 156 Auditorium Annex, 1924 West Mall 155A Auditorium Annex, 1924 West Mall Vancouver, BC V6T 1Z2 Vancouver, BC V6T122 125 Part I. Game Attendance A. How important are the following factors for your attending Canucks' games? Please check one box per item to indicate their importance. (See example.) Not important at all Very important 1 2 3 4 5 Example • • a El 1. Behaviour of fans during games a • • • • 2. Special events a • • • • 3. Accompanied by friends • a • • • 4. Excitement • • • a a 5. The Canucks' win/loss records • • • a • 6. Visiting teams' win/loss records a • • • a 7. The Canucks' superstars • • • • a 8. Visiting teams' superstars • a • • • 9. Record breaking performance of the Canucks • • • • a 10. Record breaking performance of visitors a • • • • 11. Rivalry of the Canucks and visiting teams • • • • • 12. Frequent fights and penalties • • • • a 13. Offensive output of the Canucks (speed of game, a a • a • minutes of power play, etc.) 14. Defensive output of the Canucks • • • • a 15. Closeness of competition (Canucks visiting teams) versus • • • • • 16. Unobstructed view from seats • • • • • 17. Comfortable seats • • • a • 18. Easy accessibility to facility • • • • a 19. Availability of parking a • • • • 20. Seating capacity of facility • • • • • 21. Cleanliness of facility • • • • • 22. Game time (time of day: afternoon/evening) a • • • • 23. Game day (day of week: weekday/weekend) • • • • a 24. Advertising through T.V • • • • • 25. Advertising through newspapers • • • • • 26. Advertising through magazines • • • • • 27. Advertising through radio • • • • • 28. Advertising through internet (websites) • a • • • 29. Direct mail & notification for game promotions • • • • a 30. Giveaway and prize • a • • • 31. Ticket discount • • • a • 32. Price of season ticket • • • • • 33. Price of game ticket a a • • • 126 Part II. Sports Interests and Activities. Please answer JSZ in the space provided. 34. In the past 12 months, how many Canucks' games did you attend? 35. In the past 12 months, approximately how much did you personally spend on tickets to see Canucks' games? $ 36. In the next 12 months, how many Canucks' games do you plan to attend? 37. What kind(s) of tickets have you bought (anticipate buying) this season? (Check more than one if applicable.) Season ticket • Game Pak • Half season ticket • 11 Game Pak • Group rated ticket • Single game ticket • 15 Game Pak a 38. If you are a single game ticket holder, when do you usually buy your ticket? (Check one.) Day of the game • Between one week to one month prior to the game • Week of the game • Greater than one month prior to the game • 39. Who usually buys your tickets? I do • Other family member • Spouse • My company • Friend • Other • Specify: T V • Newspaper • Radio • Magazine • Website • Other • Specify: 41. With whom do you usually attend Canucks' games? (Check more than one if applicable.) Alone • Co-workers • Spouse • Neighbours • Children • Friends • Family (parent, sister, brother) • Other • Relatives • (Specify other): 42. How many years have you been a fan of the Vancouver Canucks? 43. Are you currently a member of any recreation or sports clubs? Yes • No • If yes, which one(s): 127 44. In the next 12 months, do you intend to participate in a new physical activity (recreation or sport) that you have not previously tried? Yes • No • If yes, which one(s): 45. What are your two favourite hobbies or past-times? 1. 2. 46. What are your two favourite sports activities to 47. In ]ast 12 months, how much have you spent on participate in? these sports activities? 1. 1.$ 2. 2. $ 48. What are your two favourite outings other than attending Canucks' games? (E.g.. movies. CFL . amateur sports, art galleries, concerts.) 49. In ]ast 12 months, how much have you spent on each of these? 1. • 1.$ 2. • 2. $ 50. What are your two favourite web sites? 51. How many times do you visit these sites in an average week? How long (minutes) per visit? 1. —1 1. Visits/week= iMinutes/visit= 2. 2. Visits/week= lVIinutes/visit= —1 52. What two favourite sports did you play as a child? (E.g., hockey, football, basketball golf, soccer, tennis.) 53. If you have children (18 years and under), what are their two favourite sports (if any) to participate in? 128 Part III. Personal Background. Please answer jgX in the space provided. 54. Your age : I years 55. Gender: Male • Female • 56. Postal code: 57. What is your first language? (e.g. English, French, Chinese) 58. What is your ethnic background? 59. What business, industry or service do you work in? (e.g., corner store, travel agent company, secondary school, police, auto-body shop, unemployed etc.) Please be specific: I ~ 60. What is your occupation? (e.g., taxi driver, restaurant manager, tour guide, school teacher, auto mechanic, secretary, student, retired, unemployed) Please be specific: 61. What is your annual household income? (Please check one.) Less than C$ 10,000 • C$ 40,000 - 59,999 • C$ 10,000 - 19,999 • C$ 60,000 - 79,999 • C$ 20,000 - 29,999 • C$ 80,000 - 99,999 • C$30,000-39,999 • More than C$ 100,000 • 62. What is your highest level of education? (Please check one.) Some high school • Bachelor's degree • High school diploma • Some graduate school • Some college • Master's degree • Two-year college diploma • Doctoral degree • 129 Part IV. Opinions About Sports. Please check one box per item to indicate your level of agreement or disagreement with each of the following statements. (See example.) Completely Somewhat Neither Nor Somewhat Completely Disagree Disagree Agree Agree 1 2 3 4 5 Example • 0 • 63. Given the choice, participating in sports is more enjoyable than watching. • • • • • 64. It is always good to try new recreational activities • • • • • 65. It is difficult to understand how people could just lay around when there are so many recreational activities to do. • • • • • 66. Individual sports are far superior to team sports like soccer because you don't have to rely on others. • • • • • 67. Attending a sporting event with a friend is far more enjoyable than attending alone. • • • • • 68. It is good to see what sports reporters say about the team before deciding whether or not to buy a ticket. • • • • • 69. Professional sports provides an escape from my home and family pressures. • • • . • • 70. Participation in organized sports is not one of the best ways for kids to spend their free time. • • • • • 71. Buying professional sport memorabilia is a waste of money. • • • • • 72. Buying tickets from scalpers is not worth the extra cost, even for excellent seats. • • • • • 73. Professional sporting events take too much time to watch. • • • • • 74. For recreation, time is a bigger problem than money. • • • • • 75. Our income is high enough to satisfy nearly all our desires for sporting event tickets. • • • • • 76. Belonging to professional sports fan clubs is a waste of time. • • • • • 77. Parents should always try to attend their children's sport contests. • • • • • 78. Taking children to professional hockey events is a great way to spend time with them. • • • • • 79. For children, the positive aspects of participating in sports far outweigh the negative ones. • • • • • 80. It is important to participate in regular physical activity with your child. • • • • • 81. Tax dollars should not be spent on building stadiums and arenas for professional sport franchises. • • • • • 82. Most professional athletes show a genuine caring for the city in which they play. • • • • • 83. The enthusiasm of the amateur athlete makes amateur sports more enjoyable to watch than professional sports. • • • • • 130 Completely Disagree Somewhat Disagree Neither Nor Somewhat Agree Completely Agree 1 2 3 4 5 Example • • 0 • V 84. It would be difficult to live in a city that does not • a • • a have one or more professional sports teams nearby 85. Amateur sport athletes/teams deserve funding. more a • • • • 86. Professional athletes are only in it for the money • a • • • 87. Professional athletes tend to be whine and • • • • • complain too much. 88. When my home team wins, it really makes me feel good. • • • • • 89. Government should not assist financially troubled professional sports organizations. a • • • • 90. The success of a professional sports team brings pride to the city. • a • a • 91. Professional athletes make good role models. • • • • • 92. Government should offer tax incentives to keep a team in a city or to encourage another team to relocate there. • • • • • 93. Attending an amateur sporting event is not as exhilarating as attending a professional sporting event. • • • • • 94. Being able to cheer for my home professional sports franchise adds to my sense of community. • • • • • 95. Amateur sporting events are quite often far superior in entertainment value than their • • • • a professional counterparts. Thank you very much. 131 Appendix B Description of Terminal Nodes of the Final Tree Analysis in Study 1 132 There were 15 terminal nodes and 7 levels. (a) Node 0 notified that the sample of 1,110 individuals in total and the average frequency of game attendance was 4.9 games. (b) 8 terminal-nodes (node 5, 7, 9, 14, 15, 16, 20 & 21) exceeded the mean frequency of game attendance (5.0). (c) Node 5 represented 65 individuals whose house hold incomes were less than C$ 60,000. Hockey related web-sites were their favourite sites. They have the frequency of game attendance of 7.6 games. (d) Node 7 represented 51 individuals were internet users and their favourite websites were professional hockey related sites. They had middle to high household income ranged from C$ 60,000 to 99,999. The frequency of game attendance among these individuals was 9.0 games. (e) Node 9 represented 70 individuals and they had the highest household income of greater than C$ 100,000. Individuals belonging to this node played baseball when they were young. Their frequencies of game attendance levels were the highest with 10.3 games. (f) Node 11 represented 138 individuals belonging to low to middle house hold income groups. Their job specialties were in sale & services, applied science related, equipment operation related, recreation & sport related, or currently unemployed. Their game attendance frequencies were 4.3 games. (g) Node 12 represented 126 individuals whose household income level was as same as people in node 11. Their job specialties were in management, social science related, 133 business & finance, unique processing, or health related. The frequencies of game attendance were 2.8 games. (h) Node 14 contained 49 individuals with middle to high household income level ranged between C$ 60,000 and 99,999. They preferred being accompanied by their co- workers when they attended hockey games. The frequency of game attendance among these individuals was 7.8 games. (i) Node 15 represented 51 individuals showing the highest household income (more than C$ 100,000. They played hockey when they were young. Their game attending frequency was 9.0. (j) Node 16 represented people who had the highest house hold income with game frequencies of 5.4 games. (k) Node 17 represented 100 individuals whose ethnic backgrounds were in Caucasians. Their household income ranged low to middle and their job specialties in primary industry, student, or retired. The average number of attending hockey games of this group was 2.3. (1) Node 18 represented 44 individuals whose household income levels and job specialties were the same as individuals in node 17. The difference between node 17 and 18 was ethnicity. They were Asians, African Canadians, or Natives. The frequency of game attendance was 1.04. (m) Node 20 represented 84 individuals with middle to high household income level ranged between C$ 60,000 and 99,999. They had been fans of the Vancouver Canucks for more than 15, and equal to or less than 28 years. The average frequency of game 134 attendance among these individuals was 5.3 games. (n) Node 21 contained 40 individuals who had the greatest fan history, more than 28 years. They had middle to high household income level the same as individuals of node 20. The average frequency of game attendance was 8.1 games. (o) Node 23 represented 70 individuals who had middle to high income level. They had been fans of Canucks for less than 16 years and they tended to accompanied with their spouse in hockey games. The frequencies of game attendance were 3.7. (p) Node 24 represented 56 individuals whose household income ranged between C$ 60,000 and 99,999. Their fan history had been fans of Canucks for less than 16 years and believed that TV was the best medium to them for promoting hockey games. 135 Appendix C Group Descriptive Statistics of Activity Pattern In Study 1 136 Group Low # Group Middle # group High # group Predictors (n= 386) (n= 402) (n= 322) 0 - 1 2 -4 > 4 Mean Mean Mean Game attendance 0.87 2.7 12.8 Chi-square or F" P Years of being a fan 11.2 14.2 16.8 F = 30.681 .001 n % n % n % Years of being a fan 71.384 .001 0 - 5 126 32.6 75 18.7 44 13.7 6 - 1 0 99 25.6 106 26.4 55 17.1 11-15 42 10.9 59 14.7 49 15.2 1 6 - 2 0 46 11.9 71 17.7 76 23.6 20 < 65 16.8 82 20.4 96 29.8 Missing 8 2.1 9 2.2 2 0.6 Season 10 2.6 13 3.2 90 28.9 156.730 .001 Half season 2 0.5 4 1.0 17 5.2 23.217 .001 11 pack 5 1.3 14 3.5 39 12.1 45.340 .001 15 pack 1 0.3 2 0.5 13 4.0 21.592 .001 22 pack 3 0.8 3 0.7 7 2.2 3.492 .139 Group 26 6.7 35 8.7 42 13.0 7.742 .021 Single 331 85.8 345 85.8 163 40.6 155.514 .001 Best medium T.V. 185 47.9 202 50.2 119 37.0 14.043 .001 Radio 80 20.7 91 22.6 78 24.2 1.250 .535 Newspaper 102 26.4 110 27.4 113 35.1 7.488 .024 Website 45 11.7 55 13.7 30 9.3 3.296 .192 Magazine 5 1.3 3 0.7 1 0.3 2.150 .314 Game colleague Alone 9 2.3 11 2.7 14 4.3 2.630 .268 Spouse 124 32.1 134 33.3 128 39.8 5.080 .079 Child 43 11.1 49 12.2 53 16.5 4.799 .091 Family 95 24.6 110 27.4 86 26.7 .828 .661 Relatives 17 4.4 22 5.5 21 6.5 1.545 .462 Co-worker 32 8.3 47 11.7 72 22.4 31.531 .000 Neighbour 5 1.3 5 1.2 7 2.2 1.245 .537 Friends 208 53.9 252 62.7 181 56.2 6.690 .035 Sport club 5.506 .064 Yes 138 35.8 161 40 143 44.4 No 248 64.2 241 60 179 55.6 Sport club Soccer 17 4.4 21 5.2 12 3.7 .946 .623 Hockey 26 6.7 51 12.7 65 20.2 28.476 .001 Golf 6 1.6 11 2.7 14 4.3 5.053 .080 Fitness 15 3.9 14 3.5 11 3.4 .138 .933 Hobby Golf 37 9.6 33 8.2 47 14.6 8.380 .016 Hockey 56 14.5 77 19.2 93 28.9 22.935 .001 Ski & Board 24 6.2 37 9.2 16 5.0 5.442 .066 137 Fitness 25 6.5 25 6.2 20 6.2 .029 .986 Art 17 4.4 18 4.5 19 5.9 1.054 .590 Amateur spots 24 6.2 31 7.7 26 8.1 1.055 .590 Reading 65 16.8 60 14.9 43 13.4 1.682 .431 Favourite activity Golf 53 13.7 59 14.7 66 20.5 6.834 .033 Baseball 41 10.6 42 10.4 34 10.6 .006 .997 Hockey 84 21.8 129 32.1 130 40.4 28.895 .001 Ski & Board 43 11.1 44 10.9 30 9.3 .728 .695 Basketball 42 10.9 35 8.7 27 8.4 1.614 .446 Soccer 53 13.7 61 15.2 40 12.4 1.143 .565 Volleyball 26 6.7 40 10.0 28 8.7 2.655 .265 Outing Movie 247 64.0 263 65.4 205 63.7 .288 .866 Football 71 18.4 73 18.2 56 17.4 .128 .938 Concert 97 25.1 100 24.9 109 33.9 8.975 .011 Amateur sports 39 10.1 31 7.7 27 8.4 1.484 .476 Dinner 31 8.0 42 10.4 35 10.9 1.980 .372 Drinking 22 5.7 25 6.2 25 7.8 1.308 .520 Favourite website Major media 40 10.4 55 13.7 21 6.5 9.799 .007 Sport media 19 4.9 35 8.7 34 10.6 8.165 .017 Computer related 97 25.1 101 25.1 80 24.8 .010 .995 Hockey related 27 7.0 47 11.7 78 24.2 46.229 .001 Play as a child Baseball 83 21.5 103 25.6 91 28.3 4.431 .109 Hockey 109 28.2 141 35.1 127 39.4 10.170 .006 Basketball 76 19.7 64 15.9 61 18.9 2.100 .350 Football 40 10.4 49 12.2 33 10.2 .927 .629 Soccer 128 33.2 132 32.8 105 32.6 .025 .988 Volleyball 37 9.6 34 8.5 31 9.6 .404 .817 Note. Chi-square orF": Reported test statistic is %2, unless otherwise noted. The complete descriptive statistics table for activity pattern is presented in Appendix C. 138 Appendix D Profiles of the Three Clusters on Sports Interests and Activities In Study 2 139 Cluster Cluster 1 Cluster 2 Cluster 3 Variables (n = 291) (n = 233) (n = 161) Mean Mean Mean Chi-square or F" P Game attendance 5.3 3.5 3.8 F=5.04 .007 Years of being a fan 14.3 13.7 10.6 F=8.25 .001 n % n % n % Years of being a fan 40.178 .000 0 - 5 48 16.5 59 25.3 49 30.4 6 - 1 0 73 25.1 59 25.3 48 29.8 11-15 53 18.2 25 10.7 21 13.0 1 6 - 2 0 61 21.0 27 11.6 22 13.7 20 < 55 18.9 60 25.8 18 11.2 Missing 1 0.3 3 1.3 3 1.9 Season 31 10.7 16 6.9 12 7.5 2.716 .257 Half season 5 1.7 4 1.7 2 1.2 .176 .916 11 pack 21 7.2 14 6.0 7 4.3 1.491 .474 15 pack 2 0.7 3 1.3 4 2.5 2.584 .275 22 pack 2 0.7 3 1.3 1 0.6 .694 .707 Group 24 8.2 20 8.6 22 13.7 3.942 .139 Single 221 75.9 176 75.5 128 79.5 .974 .614 Ticket Purchaser I do 162 55.7 110 47.2 85 52.8 3.750 .153 Spouse 23 7.9 14 6.0 7 4.3 2.282 .320 Friend 31 10.7 34 14.6 30 18.6 5.681 .058 Family 53 18.2 32 13.7 20 12.4 3.370 .185 Company 23 7.9 39 16.7 12 7.5 12.931 .002 Best medium T.V. 141 48.5 110 47.2 77 47.8 .080 .961 Radio 51 17.5 54 23.2 30 18.6 2.764 .251 Newspaper 102 35.1 61 26.2 45 28.0 5.396 .067 Website 35 12.0 32 13.7 18 11.2 .639 .727 Magazine 4 1.4 0 0.00 3 1.9 3.890 .143 Others 17 5.8 13 5.6 7 4.3 .475 .789 Game colleague Alone 11 3.8 9 3.9 6 3.7 .005 .997 Spouse 88 30.2 80 34.3 38 23.6 5.222 .073 Child 31 10.7 35 15.0 10 6.2 7.592 .022 Family 101 34.7 53 22.7 39 24.2 10.772 .005 Relatives 18 6.2 11 4.7 8 5.0 .620 .733 Co-worker 28 9.6 33 14.2 20 12.4 2.631 .268 Neighbour 5 1.7 4 1.7 1 0.6 1.029 .598 Friends 189 64.9 129 55.4 106 65.8 6.425 .040 Others 11 3.8 16 6.9 5 3.1 3.928 .140 Member of sport club Yes 153 52.6 94 40.3 43 26.7 28.988 .000 No 138 47.4 139 59.7 118 73.3 Sportclub (which one?) Soccer 16 5.5 14 6.0 3 1.9 4.079 .130 Hockey 49 16.8 29 12.4 13 8.1 7.126 .028 140 Golf 8 2.7 6 2.6 4 2.5 .032 .984 Fitness 12 4.1 9 3.9 5 3.1 .299 .861 Hobby Golf 24 8.2 26 11.2 13 8.1 1.631 .442 Hockey 72 24.7 24 10.3 28 17.4 18.275 .000 Ski & Board 25 8.6 18 7.7 11 6.8 .454 .797 Fitness 28 9.6 15 6.4 6 3.7 5.696 .058 Art 7 2.4 13 5.6 5 3.1 3.884 .143 Amateur spots 34 11.7 14 6.0 14 8.7 5.095 .078 Reading 39 13.4 39 16.7 22 13.7 1.302 .521 Favourite activity Golf 44 15.1 32 13.7 23 14.3 .206 .092 Baseball 34 11.7 27 11.6 9 5.6 4.917 .086 Hockey 101 34.7 63 27.0 42 26.1 5.209 .074 Ski & Board 35 12.0 26 11.2 27 16.8 2.981 .225 Basketball 30 10.3 15 6.4 24 14.9 7.570 .023 Soccer 48 16.5 34 14.6 21 13.0 1.021 .600 Volleyball 27 9.3 13 5.6 13 8.1 2.514 .285 Outing Movie 194 66.7 135 57.9 104 64.6 4.411 .110 Football 45 15.5 36 15.5 36 22.4 4.143 .126 Concert 84 28.9 61 26.2 38 23.6 1.518 .468 Amateur sports 31 10.7 22 9.4 14 8.7 .496 .780 Dinner 33 11.3 23 9.9 10 6.2 3.154 .207 Drinking 23 7.9 17 7.3 7 4.3 2.155 .340 Use of internet 9.436 .009 Yes 204 70.1 134 57.5 110 68.3 No 87 29.9 99 42.5 51 31.7 Favourite website Major media 23 7.9 20 8.6 22 13.7 4.342 .114 Sport media 33 11.3 10 4.3 13 8.1 8.566 .014 Computer related 88 30.2 54 23.2 37 23.0 4.427 .109 Hockey related 55 18.9 18 7.7 17 10.6 15.386 .000 Play as a child Baseball 72 24.7 65 27.9 28 17.4 5.865 .053 Hockey 112 38.5 66 28.3 50 31.1 6.488 .039 Basketball 52 17.9 40 17.2 33 20.5 .756 .685 Football 37 12.7 23 9.9 12 7.5 3.204 .201 Soccer 100 34.4 65 27.9 64 39.8 6.210 .045 Volleyball 30 10.3 17 7.3 11 6.8 2.242 .326 Note. Chi-square or F" : Reported test statistic is X > unless otherwise noted. The complete table for the three clusters on sports interests and activities is presented in Appendix D. 141

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
United States 18 2
China 13 1
Germany 3 0
Tunisia 3 0
Spain 2 0
India 2 0
Canada 2 0
Sweden 2 0
Poland 1 0
Republic of Korea 1 0
United Kingdom 1 0
City Views Downloads
Beijing 8 0
Unknown 6 0
Mountain View 6 1
Nanjing 5 0
Stuttgart 3 0
Redmond 2 0
Toronto 2 0
Waltham 2 0
Stockholm 2 0
Valencia 2 0
Miami 2 0
Pittsburgh 2 0
Slippery Rock 2 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}

Share

Share to:

Comment

Related Items