Chapter 5. Data Analysis, Results and Discussion

Size: px
Start display at page:

Download "Chapter 5. Data Analysis, Results and Discussion"

Transcription

1 Chapter 5 Data Analysis, Results and Discussion 5.1 Large-scale Instrument Assessment Methodology Data analysis was carried out in two stages. In the first stage the reliabilities and validities of the instruments were assessed while in the second stage the proposed hypotheses were tested. Validity evaluates the accuracy level of an instrument while the reliability of an instrument evaluates the quality and consistency of the instrument. Bagozzi (1980) and Bagozzi and Philips (1982) advocate a process to evaluate measurement instruments that checks for reliability and validity, factor structure, the dimensionality of the scale and the validation of higher order constructs (if any). The methods used for these analyses include: corrected-item total correlation (for convergent validity), Cronbach s alpha (for reliability), average variance extracted (for discriminant validity) and CFA (to check the factor structure and the dimensionality of the scales) (Gerbing and Anderson, 1988). The scale items to measure the constructs used in the model were cleansed in the beginning with the help of Corrected Item-to-Total Correlation (CITC) scores with respect to each of the dimensions of a particular construct (refer to Table 5.1). As Churchill (1979) suggested, the main reason for undertaking this step was to remove the so called garbage items before conducting the actual EFA. The CITC score measures the extent of each item s contribution to the internal consistency of the instrument while the overall consistency of the instrument is captured by Cronbach s Alpha (Cronbach, 1951). For a construct to be considered as consistent, Cronbach s alpha value should be 0.70 or above (Robinson et al., 1991; Robinson and Shaver,

2 1973). The reliability analysis option of SPSS 20.0 was employed to calculate the CITC values for each of the constructs. All the items of the one-dimensional constructs were used simultaneously to calculate the CITC values. For constructs having more than one dimension, CITC scores for each of the sub-dimensions were calculated. A widely accepted rule is that any item having CITC value of less than 0.5 should be removed from the measurement instrument. This study also followed the same rule. However, past literature also suggests that certain items with CITC value above 0.50 can also be eliminated from the instrument if by doing so the Cronbach s alpha value improves considerably. Such elimination decisions were taken on the basis of the Alpha if deleted score (Nunnally, 1978). After making changes in the measurement instruments on the basis of the CITC and Cronbach Alpha values Exploratory Factor Analysis (EFA) of the items for each of the constructs were undertaken to examine the dimensionality of each of the constructs. SPSS 20.0 software for Windows was used for this EFA. EFA is generally used to identify the latent dimensions of a construct with help of the correlations between the observed items. For the purpose of this study Principal Component Analysis was used to identify the factor structure while orthogonal VARIMAX rotation was used to classify the items in their respective factors. This procedure finds support in the study conducted by Li et al. (2007). Additionally replace missing values with mean command was applied so that none of the responses get eliminated from the analysis (Li, 2007). Each of the constructs was verified using EFA even if a prior knowledge of dimensionalities of these constructs were known. It was expected that any uni-dimensional instrument would come up with only a single factor. In case of emergence of multiple factor structures for these uni-dimensional constructs, the logical explanation of such factor structure was carefully sought and theoretical rationalizations were explored.

3 In EFA an accepted value of factor loadings of greater than 0.30 is considered to be valid for a sample size of more than 50 respondents. But, loadings of 0.40 are generally used as the cut-off value while any factor loading above 0.50 is considered as very good (Hair et al. 1998). To adhere to the most accepted norm of EFA a factor loading of 0.50 was used as the cut-off score. Additionally, presence of high cross loading (i.e. factor loading of more than 0.50 for an item on more than one factors) led to elimination of those items from the measurement instruments. The initial set of instruments consisted a total of 70 items. Exploratory factor analysis (EFA) with VARIMAX rotation was carried out on these 70 scale items. Following the methodology suggested by Hair et al. (1998), four decision rules: factor loadings greater than 0.5, eigenvalue greater than 1 and simple factor structure were employed to identify the underlying dimensionality of the different scales used in this study. CITC was used to purify the scale items. Two items were dropped from the Team Brand Image scale suggested by Bauer et al. (2008) following the above analyses. Factor analysis on the remaining 68 items revealed that the scales were indeed measuring the constructs they were supposed to measure. However, the reliability analysis for the scales indicated that some of the items could be eliminated leading to an increased reliability of the scales. The CITC values were checked to identify the convergent validity of the items. CITC values measure the correlations between each of the items in a construct and the total score for that construct. A positive CITC value for each of the items indicated internal consistency of a construct. The logic behind such a conclusion is that the items included in an instrument to measure a particular construct should have a high correlation with that construct. However, if the correlations are small and/or non-significant, those items are not as much helpful to explain the construct as much as the other items. Such items were dropped from the measurement tools.

4 Table 5.1 presents the factor loadings and CITC for all the constructs used in the study which was derived from the EFA and the reliability analysis of the measurement instruments to test the convergent validity. This analysis was conducted using the data collected during the pre-testing. Table 5.1: Convergent Validity of the Scales from Pre-Tests S. No Items Factor loading Corrected item-tototal correlation Team Brand Identity 1 It is important to me that my friends see me as a fan of my favorite team My friends and my family recognize me as a fan of my favorite team When someone praises my favorite team, it feels like a compliment When I talk about my favorite team, I usually say we rather than they I began following my favorite team because of my friends and family members It is important to follow the same team as my friends do I follow my favorite team because my friends support the same team I possess a great deal of information about my favorite team If I were to list everything I know about my favorite team, the list will be quite long Compared to other teams, I consider myself an expert about my favorite team Team Brand Image 1 Team history and Tradition Logo and Team Colors Stadium Name

5 4 Team Anthem Main Sponsor or The Owner Composition of the current team Presence of star players in the team The performance of the Coach Performance of your favorite team in the just concluded IPL Performance of your favorite team during the last year s IPL Watching, reading, and talking about my favorite team provides temporary escape from life's problems My favorite team s games are exciting and entertaining Thinking of my favorite team brings back good memories Brand Association 1 The quality of players in my favorite team is low compared to the other teams. (Rev. Coded) My favorite team brand is a valuable brand My favorite team brand is unpleasant to others. (Rev. Coded) My favorite team s performance is quite good Overall, I think that my favorite team brand is inferior to others. (Rev. Coded) Self Image-Team Brand Image Congruency 1 It is clear to me what the team stands for I can easily connect with the image of my favorite team There is a natural fit between what the team stands for and my self-image I have learnt something new about my favorite team since I started following it

6 5 The image of my favorite team is in accordance with my self-image Brand Love 1 My favorite team is a wonderful team My favorite team makes me feel proud My favorite team is totally awesome I have neutral feelings about my favorite team. (Rev. Coded) My favorite team makes me very happy I love this team! I have no particular feelings about my favorite team. (Rev. Coded) I am passionate about my favorite team My favorite team is a pure delight I am very attached to my favorite team Team Brand Loyalty 1 I support my favorite team at all situations I am extremely committed to my favorite team There is nothing that could change my commitment to my favorite team I will not change my affiliation from my favorite team to another team because of a prolonged failure of my favorite team I would defend my favorite team in public even if this causes problems I will not shift my support towards my favorite team to another team just because my friends try to convince me to. 7 It is really important to me that my favorite team continues playing in the major tournaments The long-term success of my favorite team is important to me

7 9 I prefer to attend games of my favorite team live in the stadium if I have the ability to buy tickets I prefer not to watch games of my favorite team on TV I often follow reports about my favorite team s players, coaches, managers, etc. in the media I purchase a lot of my favorite team-related merchandise. 13 I often wear the colors and/or the logo of my favorite team. 14 I have often participated in discussions about my favorite team Word of Mouth (WOM) 1 I have recommended my favorite team to lots of people. 2 I talk up about my favorite team to my friends. 3 I try to spread the good-word about my favorite team. 4 I always talk positive about my favorite team Purchase Intention 1 I will buy a match ticket the next time my favorite team plays a match in my city I will ask my friends to buy tickets for my favorite team s match in my city I will buy the official merchandise of my favorite team whenever they are available. 4 I will definitely try out any new merchandise with a logo of my favorite team The Kaiser-Meyer-Olkin (KMO) Measure of Sampling Adequacy was calculated for each of the dimensions and levels in EFA. The second step was to examine the uni-dimensionality of the latent constructs or factors. Hair et al. (1998) defined uni-dimensionality as the feature of a set of items which are part of a single

8 latent construct or of a common concept. On the basis of a review of literature this study proposed some relationships between the items or observed variables and the common underlying trait(s) and tested them empirically. Confirmatory factory analysis (CFA) is a vital tool to measure the goodness of fit indices of a proposed model. We used the powerful tool called Structural Equation Modeling (SEM) to test the CFA models proposed in the study 1. This study used AMOS 20.0 for conducting SEM analysis. There are a set of goodness-of-fit indices that test the fit between the model and the data. Goodness-of-fit indices test the hypothesis that the observed covariance or correlation matrix is the same as that of the covariance/correlation matrix predicted by the model. Goodness-of-fit measures can be categorized into three groups: (1) absolute fit measures (used for the purpose of assessing the overall fitness of the model); (2) Incremental fit measures (used for the purpose of comparing between the proposed model and a different model proposed by the researcher called null model ); and (3) Parsimonious fit measures (used for the purpose of verifying whether the goodness-of-fit indices of the model achieve the values required to conclude that the model is a good fit. Parsimonious fit measures also help to identify the extent to which each of the goodness-of-fit indices has achieved the fit). The most commonly used test to check the fitness of a model is Chi-square Fit Index. Chi-Square Fit index calculates the extent to which the observed covariance and the calculated covariance differs from each other. For a model to have a good fit, the Chi-Square Fit index should not be significant. But, this test has an inherent problem. The problem is that as the size of the sample increases the chances of Type II error also increases, i.e. the probability of the Chi-Square Index 1 The CFA models can be first order or higher order in nature. First order models include a smaller number of factors at a single level with which the observed variables have high correlations. These factors latent variables which are in the same level are termed as primary or first-order factors. The basic factor structure for a higher order construct is a second order factor. Second-order models explore the possibility of representing a set of first order factors into a still smaller number of underlying dimensions.

9 being significant increases. Additionally this Chi-Square index is also very sensitive to multivariate normality assumption. To overcome these issues, chi-square/degree of freedom (χ 2 /df) was used where values less than 3 for the chi-square/degree of freedom statistic indicates a good fit (Carmines and McIver, 1981). AMOS 20.0 provides various other statistics to test the overall model fit of which the most important ones are: goodness of fit index (GFI), adjusted goodness of fit index (AGFI), comparative fit index (CFI), normed fit index (NFI), root mean square residual (RMR) and root mean square error of approximation (RMSEA). GFI value gives the extent to which the variance and covariance present in the model can be explained by the model. The range of GFI is 0 to 1. AGFI is similar to GFI but considers the number of degrees of freedom present in the model. NFI is used to get a comparative assessment between the proposed and the null model. CFI gives a comparison between the fit of the proposed model and the fit of the null model. Higher is the difference in fit between the proposed and the null model higher is the value of CFI suggesting a better fit between proposed model and data. CFI gives a better result compared to NFI as NFI has a tendency to give erroneous results if the sample size is small. The widely accepted values of these goodness-of-fit indices vary in the range of While the values of 0.80 and above are considered as reasonable fit; scores above 0.90 indicate a good fit for the model (Hair et al., 1998; Bentler and Bonett, 1980). RMR measures the extent to which, on an average, the sample covariance matrix differs from the predicted covariance matrix generated by the model. The range of RMR is between 0 and 1. A RMR value closer to zero is an indicator of a better model (cut-off value being 0.05) (Browne and Cudeck, 1993). RMSEA has been identified as one of the most informative model fit index. By taking the error of approximation present in the population and representing it relative to the degree of freedom present in the model RMSEA becomes susceptible to the number of

10 parameters to be estimated. RMSEA value below 0.05 is considered a good fit (Browne and Cudeck, 1993). Experts suggest that the alteration in the model should be done on a step by step basis and not more than one item should be modified in each step. Following their suggestion, this study also ensured that repeated modifications for both first-order and second-order models were made on a step by step basis to improve the model fit indices. Careful decisions were made to remove items from the constructs wherever necessary on the basis of sufficient theoretical and empirical evidence. This modification process of the instruments was repeated till the desirable values of the model fit indices were achieved. The reliability of items was measured using Cronbach s alpha. As suggested by Nunnally (1978), a Cronbach s alpha score of 0.70 was taken as the cut off score for the reliability analysis in this study. 5.2 Large Scale Measurement Results This section deals with the validation of the instruments used in this study as well as to test the validity and reliability of the scales used to measure the constructs. The following acronyms have been used to mark the items in the sub-constructs: Team Identification Team Brand Image Team Brand Association Image Congruency Brand Love Team Brand Loyalty Word of Mouth Purchase Intentions ti tbi tba ic bl tbl wom pi

11 5.2.1 Team Identification Team Identification (ti) is a uni-dimensional construct represented by 10 items. CITC scores indicated that the 5 th item i.e. I began following my favorite team because of my friends and family members was far less than 0.5 (0.311 to be precise) and the resulted Cronbach s Alpha was 0.692, this item was removed from the analysis. Although CITC values for the 6 th and the 7 th item were also less than 0.5 (0.483 and respectively), these items were not removed as it is the norm that one item should be removed at each step. The reliability analysis of the scale was carried out again after removing the 5 th item from the scale. The results indicated an improved Cronbach s Alpha value of However, the CITC values for 6 th item i.e. It is important to follow the same team as my friends do and the 7 th item i.e. I follow my favorite team because my friends support the same team were found to be less than the pre assigned value of 0.5 (0.469 and respectively). Hence, the 6 th item was been dropped from the study. After dropping the 6 th item the same procedure was repeated once more and this time the results indicated a further improved Cronbach s Alpha value of After this round of analysis it was found that only the 7 th item was having a very low value of CITC (0.248) and hence was a suitable candidate for deletion from the scale. After dropping the 7 th item from the analysis, the results indicated a good CITC values (all above 0.5). Additionally the reliability status for the 7 item measure was higher than that of the 10 item measure (Cronbach s Alpha values of compared to 0.692). The CITC of all the remaining items and their corresponding code names are shown in Table 5.2.

12 Table 5.2a: Purification of Team Identification Scale Team Identification Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct ti1 It is important to me that my friends see me as a fan of my favorite team ti2 My friends and my family recognize me as a fan of my favorite team ti3 When someone praises my favorite team, it feels like a compliment ti4 When I talk about my favorite team, I usually say we rather than they ti6 It is important to follow the same team as my friends do ti7 I follow my favorite team because my friends support the same team ti8 I possess a great deal of information about my favorite team ti9 If I were to list everything I know about my favorite team, the list will be quite long ti10 Compared to other teams, I consider myself an expert about my favorite team ti5 I began following my favorite team because of my friends and family members.411 Item removed after purification Table 5.2b: Purification of Team Identification Scale Team Identification Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct ti1 ti2 It is important to me that my friends see me as a fan of my favorite team My friends and my family recognize me as a fan of my favorite team ti3 When someone praises my favorite team, it feels

13 like a compliment ti4 When I talk about my favorite team, I usually say we rather than they ti7 I follow my favorite team because my friends support the same team ti8 I possess a great deal of information about my favorite team ti9 If I were to list everything I know about my favorite team, the list will be quite long ti10 Compared to other teams, I consider myself an expert about my favorite team ti6 It is important to follow the same team as my friends do.469 Item removed after purification Table 5.2c: Purification of Team Identification Scale Team Identification Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct ti1 It is important to me that my friends see me as a fan of my favorite team ti2 My friends and my family recognize me as a fan of my favorite team ti3 When someone praises my favorite team, it feels like a compliment ti4 When I talk about my favorite team, I usually say we rather than they ti8 I possess a great deal of information about my favorite team ti9 If I were to list everything I know about my favorite team, the list will be quite long ti10 Compared to other teams, I consider myself an expert about my favorite team ti7 I follow my favorite team because my friends support the same team.248 Item removed after purification

14 An exploratory factor analysis (EFA) was conducted thereafter using principal components as the means of extraction provide the results are shown in Table 5.3. The KMO score of indicated an acceptable sampling adequacy. The total variance explained by the single factor was %. All the items were loaded on a single factor with factor loadings above 0.50 with no cross loadings. The composite reliability (CR = 0.892) and the Convergent Validity (AVE= 0.542) were found to be acceptable, thus acknowledging the factor structure. Table 5.3: Factor Structure of Team Identification Scale Kaiser-Mayer-Olkin (KMO) Measure of Sampling Adequacy = Item Team Identification Cronbach s Alpha for the construct ti ti ti ti ti ti ti Eigenvalue % of Variance % The next step was to test the 7 item Team Identification (ti) scale in CFA to measure the model fit. The CFA model for Team Identification was tested using AMOS The results indicated acceptable model fit indices: χ2/df =1.576, RMR =.025, GFI =.997, AGFI =.984, NFI =.996, RMSEA =.026 and CFI =.998; eliminating the need of any further modifications. The model for Team Identification (TI) is shown in Fig The factor loadings (λ) were all found to be above.50 and significant.

15 ti1 ti2 ti3 ti4 ti Team Identification ti9 ti Fig. 5.1: Confirmatory Factor Analysis Model of Team Identification (ti) Team Brand Image Team Brand Image was initially represented by 17 items. However during the pre-testing stage it was found that the respondents as well as the experts in the field of sports in India 2 suggested the items Team Culture and Values and Fans Behavior On and Off the Field to be removed. This suggestion was logical because of the context of this study. As this study deals with Indian Premier League (IPL) which is an event only 6 years old, it could be expected that the team culture and values might not be having any significant impact on the fans psychological as well as physical behavior. Hence these two items were not considered as effective items to measure Team Brand Image. The pre-test results also supported the deletion of these two items as the results show very low CITC values of less than 0.5 for both of these two items. As suggested by Bauer et al. (2008) Team Brand Image was divided into three sub-categories e.g. product related attributes (5 items), non-product related attributes (5 items) and brand benefits (5 items). The 2 One sports journalist of a leading English news paper, ex-legal advisor of BCCI, Joy Bhattacharya (managing director of Kolkata Knight Riders, an IPL franchise). Mr. Arumugam (member of Hockey India League).

16 reliability analysis results indicated that though the items measuring non-product related attributes and product related attributes have Cronbach s Alpha values of and respectively, which were within the acceptable limit, the items measuring brand benefits had a very low Cronbach s Alpha value of while the item My favorite team brings prestige to the community was found to have a very low CITC value of (much less than the acceptable value of 0.5). After removing this item from the instrument, the reliability of the instrument increased to The results are presented in Table 5.4. An exploratory factor analysis (EFA) was then conducted using principal components as the means of extraction. The results are shown in Table 5.5. The KMO score of indicated an acceptable sampling adequacy. The total variance explained by the three factors was 54.27%. All the items were loaded on the factors they are supposed to load with factor loadings above 0.50 with no cross loadings, thus indicating a good factor structure. Team Brand Image Table 5.4: Purification of Team Brand Image Scale Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct Non-Product Related Attributes tbi1 Team history and Tradition.585 bi2 Logo and Team Colors.610 tbi3 Stadium Name.626 tbi4 Team Anthem tbi5 Main Sponsor or The Owner.645 Product Related Attributes tbi6 Composition of the current team.617 tbi7 Presence of star players in the team.535

17 tbi8 The performance of the Coach tbi9 tbi10 Tbi11 Brand Benefits Performance of your favorite team in the just concluded IPL Performance of your favorite team during the last year s IPL Watching, reading, and talking about my favorite team provides a temporary escape from life's problems Tbi12 My favorite team s games are exciting and entertaining Tbi13 Thinking of my favorite team brings back good memories Tbi14 I have fond memories of following my favorite team with friends and/or family members tbi15 My favorite team brings prestige to the community.332 Item removed after purification Table 5.5: Factor Structure of Team Brand Image Kaiser-Mayer-Olkin (KMO) Measure of Sampling Adequacy = Item Non-Product Related Attributes Product Related attributes Brand Benefits Cronbach s Alpha for the construct tbi1.646 tbi2.681 tbi3.713 tbi tbi5.731 tbi6.731 tbi7.644 tbi tbi9.744 tbi tbi11.731

18 tbi12 tbi13 tbi Eigenvalue % of Variance % % % % (Total) The composite reliability (CR = 0.838) and the Convergent Validity (AVE= 0.510) of Non- Product Related Attributes were found to be acceptable. Similarly the composite reliabilities and the Convergent Validities of Product Related Attributes and Brand Benefits were found to be acceptable (CR = and respectively; AVE= and respectively), thus accepting the factor structures. The results of the discriminant validity showed that all variances extracted were found to be higher than the squared correlation of each pair indicating that each dimension was empirically distinct. Table Discriminant validity and squared correlation between the constructs Non-Product Related Attributes Product Related Attributes Brand Benefits Non-Product Related Attributes (0.50)** Product Related Attributes 0.32* (0.55)** Brand Benefits 0.13* 0.26* (0.52)** * - p<.01; ** - diagonal values represent the average variance extracted of the constructs. The next step was to test the sub-categories of the Team Brand Image (tbi) scale in CFA to measure the model fit. The CFA model for the first sub category i.e. non-product related attributes was tested using AMOS The results indicated acceptable model fit indices: χ2/df

19 =8.049, RMR =.052, GFI =.992, AGFI =.987, NFI =.982, RMSEA =.043 and CFI =.984; eliminating the need of any modifications. The model for Non-Product Related Attributes (NPA), a sub-category of Team Brand Image (tbi) is shown in Fig The factor loadings (λ) were all found to be above.50 and significant. tbi tbi3 tbi4 tbi NPA tbi Fig. 5.2: Confirmatory Factor Analysis Model of Non-Product Related Attributes (NPA): A Sub-Category of Team Brand Image Scale The CFA model for the second sub-category i.e. product related attributes was tested using AMOS The results indicated acceptable model fit indices: χ2/df =2.209, RMR =.016, GFI =.998, AGFI =.984, NFI =.996, RMSEA =.038 and CFI =.988; eliminating the need of any modifications. The model for the sub-category of Team Brand Image (tbi), Product Related Attributes (PA) is shown in Fig The factor loadings (λ) were all found to be above.50 and significant.

20 tbi6 tbi7 tbi8 tbi9 tbi PA Fig. 5.3: Confirmatory Factor Analysis Model of Product Related Attributes (PA): A Sub- Category of Team Brand Image Scale The CFA model for the third and final sub-category i.e. brand benefits was tested using AMOS The results indicated acceptable model fit indices: χ2/df =2.281, RMR =.023, GFI =.997, AGFI =.986, NFI =.981, RMSEA =.023 and CFI =.984; eliminating the need of any modifications. The model for the sub-category of Team Brand Image (tbi), Non-Product Related Attributes (NPA) is shown in Fig The factor loadings (λ) were all found to be above.50 and significant. tbi11 tbi12 tbi13 tbi BB Fig. 5.4: Confirmatory Factor Analysis Model of Brand Benefits (BB): A Sub-Category of Team Brand Image Scale The next step was to test the first order CFA model for Brand Image scale. Although all λ coefficients for the initial model were found to be greater than.60, the model fit was very poor: χ 2 /df =4.695; RMR = 0.1, GFI = 0.946, AGFI = and RMSEA = indicating a possibility of error correlation (Table 5.6). Modification indices indicated high error correlation

21 between tbi5 and tbi12, tbi13 and tbi9. Hence, it was decided to delete item tbi5 from the model. The model after removing tbi5 showed an improvement with satisfactory χ 2 /df value (4.331), high RMR (.090) and RMSEA values (0.064). However, there was still some scope to make further modifications. AMOS modification indices showed a high error correlation between tbi12 and tbi6, tbi7, tbi8, tbi9, tbi10 and tbi11. tbi12 was hence removed from the model. The model finally showed a good fit (except RMR): (χ 2 /df = 4.28, RMR = 0.083, GFI = 0.993, AGFI = and RMSEA= 0.049) and hence there was no need of further modifications. Table 5.6: CFA Model Fit Statistics for Team Brand Image Scale Fit Indices χ 2 χ 2 /df RMSEA RMR GFI AGFI NFI CFI Initial Model after removing tbi5 Model after removing tbi5 and tbi

22 The first order CFA model for Brand Image scale is shown in Fig tbi tbi tbi3 tbi npa tbi6 tbi tbi8 tbi pa 0.79 tbi tbi11 tbi13 tbi bb Fig. 5.5: 1 st Order Model of Team Brand Image Scale In the next step, the second order model for Team Brand Image (tbi) scale was tested to identify whether the three sub-categories of tbi scale, i.e. Product Attributes (pa), Non Product Attributes (npa) and Brand Benefits (bb) lead to a single higher order construct of Team Brand Image (tbi). The second order model of tbi showed very good model fit indices apart from the RMR value: χ 2 /df = 2.861; RMR = 0.078, GFI = 0.975, AGFI = 0.955, NFI = 0.952, CFI = and RMSEA = The standardized coefficients were 0.71 for Non Product Attributes (npa), 0.64 for Product Attributes (pa) and 0.85 for Brand Benefits (bb) and all were found to be statistically

23 significant suggesting the second order model to be suitable for the analysis (Fig. 5.6). tbi tbi tbi3 tbi npa 0.71 tbi tbi tbi8 tbi pa 0.64 tbi tbi10 tbi11 tbi tbi14 Fig. 5.6: 2 nd Order Model of Team Brand Image Scale bb Team Brand Association Team Brand Association (tba) is a single dimension construct measured by 6 items. CITC scores indicated that the 2 nd item i.e. My favorite team is absolutely worthless was far less than 0.5 (0.202 to be precise) and the resulted Cronbach s Alpha is Hence, this item was removed from the analysis. The reliability analysis of the scale was again done after removing the 2 nd item from the scale. The results indicated an improved Cronbach s Alpha value of After dropping the 2 nd item from the analysis, the results indicated a good CITC values (all above 0.5) as well as a higher reliability status for the 5 item measure instead of the 6 item measure (Cronbach s Alpha values of compared to 0.785). The CITC of all the remaining items and

24 their corresponding code names are shown in Table 5.7. In this case purposefully three items out of the five items were reverse coded to check the consistency of the responses. Those items were the 1 st item i.e. The quality of players in my favorite team is low compared to the other teams, the 3 rd item i.e. My favorite team brand is unpleasant to others and the 6 th item Overall, I think that my favorite team brand is inferior to others. Table 5.7: Purification of Team Brand Association Scale Brand Association Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct tba1 The quality of players in my favorite team is low compared to the other teams tba3 My favorite team is absolutely worthless tba4 My favorite team brand is a valuable brand tba5 My favorite team brand is unpleasant to others tba6 My favorite team s performance is quite good tba2 Overall, I think that my favorite team brand is inferior to others.202 Item removed after purification An exploratory factor analysis (EFA) was then conducted using principal components as the means of extraction. The results are shown in Table 5.8. The KMO score of indicated an acceptable sampling adequacy. The total variance explained by the single factor was %. All the items were loaded on the factors they were supposed to load with factor loadings above 0.50 and with no cross loadings. The composite reliability (CR = 0.877) and the Convergent Validity (AVE= 0.588) were found to be acceptable, thus accepting the factor structure.

25 Table 5.8: Factor Structure of Team Brand Association Scale Kaiser-Mayer-Olkin (KMO) Measure of Sampling Adequacy = Item Brand Association Cronbach s Alpha for the construct tba tba tba tba tba Eigenvalue % of Variance % The CFA model for Team Brand Association was tested using AMOS The results indicated acceptable model fit indices: χ2/df =1.156, RMR =.018, GFI =.998, AGFI =.992, NFI =.996, RMSEA =.014 and CFI =.989; thus eliminating any need of further modifications. The model for the Team Brand Association (tba) is presented in Fig The factor loadings (λ) were all found to be above.50 and significant. tba1 tba tba4 tba5 tba Team Brand Association Fig. 5.7: Confirmatory Factor Analysis Model of Team Brand Association (tba) Scale Image Congruency Image Congruency (ic) is a single dimension construct measured by 5 items. The reliability analysis of the scale indicated a Cronbach s Alpha value of All the items measuring

26 Image Congruency were found to be either equal or above the acceptable CITC value of 0.5. The CITC of all the items and their corresponding code names are shown in Table 5.9. Table 5.9: Purification of Image Congruency Scale Coding Item Name CITC Cronbach s Alpha for the construct ic1 It is clear to me what the team stands for.534 ic2 I can easily connect with the image of my favorite team.631 ic3 ic4 There is a natural fit between what the team stands for and my self-image I have learnt something new about my favorite team since I started following it ic5 The image of my favorite team is in accordance with my self-image.500 An exploratory factor analysis (EFA) was then conducted using principal components as the means of extraction. The results are shown in Table The KMO score of indicated a slightly low but acceptable sampling adequacy. The total variance explained by the single factor was %. All the items loaded on the factors they were supposed to load with factor loadings above 0.50 and with no cross loadings indicating a good factor structure.

27 Table 5.10: Factor Structure of Image Congruency Scale Kaiser-Mayer-Olkin (KMO) Measure of Sampling Adequacy = Item Image Congruency Cronbach s Alpha ic ic ic ic ic Eigenvalue % of Variance % The CFA model for image congruency was tested using AMOS The results indicated acceptable model fit indices: χ2/df =4.713, RMR =.030, GFI =.995, AGFI =.991, NFI =.992, RMSEA =.047 and CFI =.994; cancelling out any further modifications. The model for the Image Congruency (ic) is presented in Fig The factor loadings (λ) were all found to be above.50 and significant. The composite reliability (CR = 0.856) and the Convergent Validity (AVE= 0.544) were found to be acceptable, thus accepting the factor structure. ic ic2 ic3 ic4 ic Image Congruency Fig. 5.8: Confirmatory Factor Analysis Model of Image Congruency (ic) Scale

28 5.2.5 Brand Love Brand Love (bl) is a single dimension construct measured by 10 items. CITC scores indicated that the 9 th item i.e. I have no particular feelings about my favorite team, which was a reverse coded question, was far less than 0.5 (0.164 to be precise) and the resulted Cronbach s Alpha was 0.808, and hence this item was removed from the analysis. Although CITC values for the 4 th and the 10 th item were also less than 0.5 (0.398 and respectively), these items were not removed as it is the norm that one item should be removed at each step. The reliability analysis of the scale was again done after removing the 9 th item from the scale. The results indicated an improved Cronbach s Alpha value of However, the CITC values for 4 th item i.e. I have neutral feelings about my favorite team and the 10 th item i.e. I am very attached to my favorite team were found to be less than the pre assigned value of 0.5 (0.438 and respectively). Hence, the 4 th item was also dropped from the study. After dropping the 4 th item the same procedure was repeated once more and this time the results indicated a further improved Cronbach s Alpha value of After this round of analysis it was found that only the 10 th item was having a very low value of CITC (0.445) and hence was a suitable candidate for deletion from the scale. After dropping the 10 th item from the analysis, the results indicated a good CITC values (all above 0.5) as well as a higher reliability status for the 7 item measure instead of the 10 item measure (Cronbach s Alpha values of compared to 0.808). The CITC of all the remaining items and their corresponding code names were shown in Table 5.11.

29 Table 5.11a: Purification of Brand Love Scale Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct bl1 My favorite team is a wonderful team bl2 My favorite team makes me feel proud bl3 My favorite team is totally awesome bl4 I have neutral feelings about my favorite team bl5 My favorite team makes me very happy bl6 I love this team! bl7 My favorite team is a pure delight bl8 I am passionate about my favorite team bl10 I am very attached to my favorite team bl9 I have no particular feelings about my favorite team.164 Item removed after purification Table 5.11b: Purification of Brand Love Scale Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct bl1 My favorite team is a wonderful team bl2 My favorite team makes me feel proud bl3 My favorite team is totally awesome bl5 My favorite team makes me very happy bl6 I love this team! bl7 My favorite team is a pure delight bl8 I am passionate about my favorite team bl10 I am very attached to my favorite team bl4 I have neutral feelings about my favorite team.438 Item removed after purification

30 Table 5.11c: Purification of Brand Love Scale Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct bl1 My favorite team is a wonderful team bl2 My favorite team makes me feel proud bl3 My favorite team is totally awesome bl5 My favorite team makes me very happy bl6 I love this team! bl7 My favorite team is a pure delight bl8 I am passionate about my favorite team bl10 I am very attached to my favorite team.445 Item removed after purification An exploratory factor analysis (EFA) was then conducted using principal components as the means of extraction. The results are shown in Table The KMO score of indicated an acceptable sampling adequacy. The total variance explained by the factor was %. All the items were loaded on the factors they are supposed to load with factor loadings above 0.50 and have no cross loadings. The composite reliability (CR = 0.882) and the Convergent Validity (AVE= 0.518) are found to be acceptable, thus accepting the factor structure.

31 Table 5.12: Factor Structure of Brand Love Scale Kaiser-Mayer-Olkin (KMO) Measure of Sampling Adequacy = Item Brand Love Cronbach s Alpha for the construct bl bl bl bl bl bl bl Eigenvalue % of Variance % The CFA model for image congruency was tested using AMOS The results indicated acceptable model fit indices: χ2/df =3.895, RMR =.027, GFI =.995, AGFI =.992, NFI =.994, RMSEA =.049 and CFI =.995; thus eliminating any need of further modifications. The model for the Brand Love (bl) is presented in Fig The factor loadings (λ) were all found to be above.50 and significant. Brand Love bl1 bl2 bl3 bl5 bl6 bl7 bl8 Fig. 5.9: Confirmatory Factor Analysis Model of Brand Love (bl) Scale

32 5.2.6 Team Brand Loyalty Team Brand Loyalty is initially represented by 14 items. It was expected that as this study dealt with Indian Premier League (IPL) which is an event only 6 years old, the loyalty issues might not have a clear impact on the fans attitudes as well as on their physical behavior. As suggested by Day et al. (1969) Brand Loyalty was divided into two sub-categories e.g. attitudinal loyalty (8 items) and behavioral loyalty (6 items). The reliability analysis results indicated that the items measuring attitudinal loyalty and behavioral loyalty have Cronbach s Alpha values of and (within the acceptable limit) respectively. However, it was found that in case of both the constructs, some items were having very low CITC values, thus becoming candidates for deletion from the analysis Purification for Attitudinal Loyalty Scale Attitudinal Loyalty is a single dimension construct measured by 8 items. CITC scores indicated that the 7 th item i.e. It is really important to me that my favorite team continues playing in the major tournaments was less than 0.5 (0.490 to be precise) and the resulted Cronbach s Alpha is Hence, this item was removed from the analysis. Although CITC values for the 6 th and 8 th item were also less than 0.5 (0.428 and to be precise), these items were not removed as it is the norm that one item should be removed at each step. The reliability analysis of the scale was again done after removing the 7 th item from the scale. The results indicated a slightly lower Cronbach s Alpha value of However, the CITC values for the 6 th item i.e. I will not shift my support towards my favorite team to another team just because my friends try to convince me to and the 8 th item i.e. The long-term success of my favorite team is important to me were found to be less than the pre assigned value of 0.5 (0.411 and to be precise). Hence, in the next stage the 8 th item was also dropped from the study. After dropping the 8 th item the same

33 procedure was repeated once more and this time the results indicated a further improved Cronbach s Alpha value of After this round of analysis it was found that only the 6 th item was having a very low value of CITC (0.394) and hence was a suitable candidate for deletion from the scale. After dropping the 6 th item from the analysis, the results indicated a good CITC values (all above 0.5) as well as indicated a slightly higher reliability status for the 5 item measure instead of the 8 item measure (Cronbach s Alpha values of compared to 0.822). The CITC of all the remaining items and their corresponding code names are shown in Table Table 5.13a: Purification of Team Brand Loyalty (Attitudinal Loyalty) Scale Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct tbl1 I support my favorite team at all situations tbl2 I am extremely committed to my favorite team tbl3 There is nothing that could change my commitment to my favorite team tbl4 I will not change my affiliation from my favorite team to another team because of a prolonged failure of my favorite team tbl5 I would defend my favorite team in public even if this causes problems tbl6 I will not shift my support towards my favorite team to another team just because my friends try to convince me to tbl8 The long-term success of my favorite team is important to me tbl7 It is really important to me that my favorite team continues playing in the major tournaments.490 Item removed after purification

34 Table 5.13b: Purification of Team Brand Loyalty (Attitudinal Loyalty) Scale Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct tbl1 I support my favorite team at all situations tbl2 I am extremely committed to my favorite team tbl3 There is nothing that could change my commitment to my favorite team tbl4 I will not change my affiliation from my favorite team to another team because of a prolonged failure of my favorite team tbl5 I would defend my favorite team in public even if this causes problems tbl6 I will not shift my support towards my favorite team to another team just because my friends try to convince me to tbl8 The long-term success of my favorite team is important to me.373 Item removed after purification Table 5.13c: Purification of Team Brand Loyalty (Attitudinal Loyalty) Scale Coding Item Name Initial CITC Final CITC Cronbach s Alpha for the construct tbl1 I support my favorite team at all situations tbl2 I am extremely committed to my favorite team tbl3 tbl4 There is nothing that could change my commitment to my favorite team I will not change my affiliation from my favorite team to another team because of a prolonged failure of my favorite team tbl5 I would defend my favorite team in public even if this causes problems tbl6 I will not shift my support towards my favorite team to another team just because my friends try to convince me to.394 Item removed after purification

35 An exploratory factor analysis (EFA) was then conducted using principal components as the means of extraction. The results are shown in Table The KMO score of indicated an acceptable sampling adequacy. The total variance explained by the two factors was %. All the items were loaded on the factors they are supposed to load with factor loadings above 0.50 having no cross loadings. The composite reliability (CR= 0.878) and the Convergent Validity (AVE= 0.593) of Attitudinal Loyalty were found to be acceptable, thus accepting the factor structure. Table 5.14: Factor Structure of Team Brand Loyalty (Attitudinal Loyalty) Scale Kaiser-Mayer-Olkin (KMO) Measure of Sampling Adequacy = Item Attitudinal Loyalty Cronbach s Alpha for the construct tbl tbl tbl tbl tbl Eigenvalue % of Variance % The CFA model for attitudinal loyalty was tested using AMOS The results indicated an acceptable model fit indices: χ2/df =2.817, RMR =.012, GFI =.997, AGFI =.980, NFI =.998, RMSEA =.047 and CFI =.997; thus eliminating any need of further modifications. The model for the Attitudinal Loyalty is presented in Fig The factor loadings (λ) were all found to be above.50 and significant.

36 tbl1 tbl tbl3 tbl4 tbl Attitudinal Loyalty Fig. 5.10: Confirmatory Factor Analysis Model of Attitudinal Loyalty: a sub-category of Team Brand Loyalty Scale Purification for Behavioral Loyalty Scale Behavioral Loyalty is a single dimension construct measured by 6 items. CITC scores indicated that the 1 st item i.e. I prefer to attend games of my favorite team live in the stadium if I have the ability to buy tickets was less than 0.5 (0.327 to be precise) and the resulted Cronbach s Alpha was Hence, it was decided to remove this item from the analysis. Although CITC values for the 2 nd item was also less than 0.5 (0.205 to be precise), this item was not removed as it is the norm that one item should be removed at each step. The reliability analysis of the scale was again done after removing the 1 st item from the scale. The results indicated a slightly lower Cronbach s Alpha value of However, the CITC value for the 2 nd item i.e. I prefer to watch games of my favorite team on TV was found to be less than the pre assigned value of 0.5 (0.211 to be precise). Hence, in the next stage the 2 nd item was also dropped from the study. After dropping the 2 nd item the same procedure was repeated once more and this time the results indicated a further improved Cronbach s Alpha value of After this round of analysis it was found that the results indicate a good CITC values (all above 0.5) as well as a slightly higher reliability status for the 4 item measure instead of the 6 item measure (Cronbach s Alpha values