Cause-related marketing: the role of team identification in consumer choice of team licensed products.

Lee, Jaedeock ; Ferreira, Mauricio

Introduction

Over the past decades, cause-related marketing (CRM) has been considered as one of the most promising communication tools in the United States (IEG, 2009). Recent estimates projected corporate spending in cause-related initiatives to reach $1.51 billion in 2009 (IEG, 2009). Following the cause-related marketing literature, CRM refers to initiatives where firms contribute a specified amount to a cause contingent upon the consumer buying the company's product (Varandarajan & Menon, 1988). This type of marketing initiative is to be distinguished from sponsorship of causes, where the contribution to a cause does not depend on the consumers' purchases (Cornwell & Coote, 2005).

Many studies have demonstrated that the impact of CRM on consumer choice can be influenced by many factors (Barone, Miyazaki, & Taylor, 2000; Bloom, Hoeffler, Keller, & Meza, 2006; Pracejus & Olsen, 2004; Strahilevitz & Myers, 1998). One factor is the degree of perceived fit between the firm contributing to a cause and its beneficiary. Higher degrees of perceived fit between the firm and the beneficiary can aid consumers' information processing and have been shown to have a positive impact on consumer choice (Pracejus & Olsen, 2004). Another factor of influence is the trade-offs that individuals are willing to make when making a purchase that would benefit a cause (Barone et al., 2000). The degree to which consumers are willing to make trade-offs may depend on how large the tradeoffs are and how the brand engaging in CRM compares to other brands on other features that are also important to the purchase decision (Barone et al., 2000). In addition, the way consumers evaluate firms' motivations to engage in CRM, whether they are socially or profit-motivated, can influence how much value they assign to a "socially responsible" feature of a brand (Barone et al, 2000). Furthermore, the product category, whether more frivolous or more practical, can influence whether individuals would choose to make a contribution to a charity or obtaining price discounts through purchases (Strahilevitz & Myers, 1998).

The way individuals identify with organizations may also play an important role in explaining consumer choice and the success of a cause-related marketing initiative. Bloom et al. (2006) have shown that the degree of affinity individuals have with different types of brand affiliations (e.g., sport teams, social causes, events, and arts) can impact the importance individuals assign to the affiliation itself as an attribute of the brand. Using conjoint analysis, they have observed that when an affiliation is perceived as too "commercial," like professional sport teams, individuals were more likely to consider the affiliation as unimportant or negative to the brand. This was observed even when a condition of high-fit affiliation (e.g., between a beer brand and a sport team) was examined. Both a high-fit commercial affiliation (e.g., beer and sport team) and a low-fit commercial affiliation (e.g., beer and Sunday night movie on network television) were weighted negatively toward the brand. The opposite was true when the affiliations were cause-related (e.g., beer and a designated driver program or children's reading program).

This study seeks to extend the current work on the influence of CRM on consumer choice by providing evidence that team identification can impact the relationship between CRM and consumer choice of team-licensed products. Instead of examining the degree of affinity between commercial (e.g., sport team) versus cause affiliations (e.g., a reading program) as considered by Bloom et al. (2006), this study examines the extent to which affinity can vary between two cause-related affiliations, that and the impact that team identification may have in this relationship. Particularly, we implemented a discrete choice experiment (Louviere & Woodworth, 1983), which indicated that preference for purchasing team-licensed products that supported a cause can depend on the levels of identification individuals have with the team.

Theoretical Background

Brand-Cause Fit

Fit or congruence in a social marketing context is the perceived link between the causes and firms (Becker-Olsen, Cudmore, & Hill, 2006). Congruence is defined as "the extent to which a brand association shares content and meaning with another brand association" (Keller, 1993, p. 7). The notion of congruence is important because it can impact how well consumers can process information related to brand and its affiliations, and improve clarity of firms' positioning (Simmons & Becker-Olsen, 2006). A perceived high fit between firms and causes would be consistent with individuals' prior expectations regarding firms' actions, which in turn facilitates how they process the information and form brand images. An example of high fit would be a beer manufacturer sponsoring a designated driver program or a retailer of home improvement and construction sponsoring a program to help the homeless (Becker-Olsen et al., 2006; Bloom et al., 2006). Professional sport leagues and community-based sports may also represent a high fit relationship. On the other hand, perceived low fit associations can lead to confusion and negative attitudes toward the firm (Becker-Olsen et al., 2006). An example of low fit would be a beer manufacturer sponsoring a children's reading program, a retailer of home improvement and construction sponsoring a domestic violence prevention program, or a sport team sponsoring art events for youths.

Becker-Olsen et al. (2006) showed that a low fit between a firm and a cause is likely to diminish purchase intentions, overall attitude toward the sponsor, and perceptions of credibility. They also showed that most positive outcomes were observed for high-fit and socially-driven initiatives as opposed to profit-driven initiatives. Moreover, Pracejus and Olsen (2004) showed that high fit between brand and beneficiary can directly and positively impact consumer choice.

Although brand-cause fit has been shown to impact consumer choice, its impact may be moderated by other factors. We contend here that identification with the brand and cause can play a role in the impact that fit can have on choice.

Social Identity Theory

Social identity theory (Tajfel & Turner, 1979) is one theory that may be useful in understanding CRM initiatives. Basically, social identity theory holds that people define themselves in terms of membership to social categories (Ashforth & Mael, 1989). Tajfel (1982) explained that "social identity is the individual's knowledge that he belongs to certain social groups together with some emotional and value significance to him of the group membership" (p. 31). Thus, according to this theory, people tend to have in-group favoritism and needs for positive distinctiveness from others. Therefore, this theory would suggest that individuals might be willing to engage in purchasing a product if the purchase is perceived as a way to support an organization they care about.

Social identity theory has received a great amount of attention in high involvement contexts like sports. Fans highly identified with teams are likely to evaluate other fans of the same team (in-group members) more positively than out-group members (Branscomb & Wann, 1994; Wann & Dolan, 1994), as well as more likely to purchase their team licensed products (Fisher & Wakefield, 1998), and attend games of teams with which they are identified (Fisher & Wakefield, 1998). Several studies have found fan identification to have an impact on various constructs, such as bias toward in-group fans (Branscomb & Wann, 1994; Wann & Dolan, 1994), self-esteem (Wann, Royalty & Roberts, 2000), emotional responses to team performance (Wann & Branscomb, 1992; Wann, Royalty, & Rochelle, 2002), aggression (Dimmock & Grove, 2003; Wann, Peterson, Cothran, & Dykes, 1999), biased predictions of player performance (Wann et al., 2006), basking-in-reflected-glory (BIRGing) behavior (Madrigal, 1995; Sloan, 1989), and impulse buying of sport team licensed merchandise (Kwon & Armstrong, 2002). Thus, a sport fan identified with a particular sport team may be willing to purchase a product if the purchase is perceived as a way to support their team. This form of behavior can be a way for individuals to reinforce their in-group favoritism toward those who are also fans of the same team.

Identification has also started to receive some attention in the CRM literature. For example, Gupta and Pirsch (2006) conducted two experiments examining the role of customer identification with the company in the brand-cause fit relationship. The results of two experiments indicated that a high congruence between a brand and a cause improved attitudes toward the CRM and increased purchase intention. Moreover, the effects of congruence were enhanced when the respondents were highly identified with the company as well as the cause. Barone, Norman, & Miyazaki (2007) also found that customer affinity toward a social cause was a significant moderating variable between retailer-cause fit and consumer evaluations on CRM programs.

Based on the previous findings, it is expected that a cause with a high degree of fit with a brand to be evaluated more positively than a cause with a low degree of fit with the brand. However, it is also expected that this evaluation will be either exacerbated or mitigated depending on the degree of identification individuals have with the brand or cause. Thus, individuals may be more willing to purchase products if the purchase is perceived as a way to support a cause they care more about. That is, preference for a beneficiary congruent with the brand can be more evident among those who are highly identified with the brand.

Method

Discrete Choice Method

Previous CRM research, with the exception of a few (Barone et al., 2000; Pracejus & Olsen, 2004), emphasizes the importance of examining the effectiveness of CRM initiative by using trade-off methods. The choice framework employed here is based on well-established random utility theory, which indicates that individuals are utility maximizers. That is, individuals form overall preferences for products based on their preferences associated with each relevant and important feature of the product and choose the one product from which they can derive the most benefits. This choice framework also indicates that choices are stochastic in nature, which means that there is a degree of randomness in the choice process such that individuals will not always choose the product that will maximize utility.

Following this framework, cause-related marketing programs were considered as one of the attributes of a product from which individuals can derive benefits. By doing so, this study could directly evaluate the contribution a cause-related marketing program can make to consumers' overall choices, and how team identification can influence that relationship.

We developed a discrete choice experiment where students were asked to assume they were shopping for Major League Baseball (MLB) sport team-licensed baseball caps. The use of a product like team-licensed baseball caps allowed us to assess the influence of team identification on the trade-offs consumers are willing to make--such as product quality, features and price--to support a team that also supports a cause. A donation to charity was presented as a transaction-based CRM program (Varadarajan & Menon, 1988), where donations were contingent on consumer purchases of a team-licensed baseball cap.

Subjects and Data Collection

To test the influence of team identification on the relationship between CRM and choice, 119 students currently enrolled in three online sport management classes at a large Southwestern university were asked to participate in an online survey. Because the goal of this study was to test relationships based on theory, the use of a relevant and homogeneous sample was preferred and a representative sample of consumers was not required (Calder, Phillips, & Tybout, 1981). College students constitute a homogeneous sample that belongs to one of the main market segments of team licensed products of baseball (SBRnet, 2009).

The questionnaire was comprised of a set of discrete choice experiment scenarios, team identification measures, and demographics. Identification with the professional teams was measured using the Sport Spectator Identification Scale (SSIS) developed by Wann and Branscomb (1993). The SSIS, successfully used in several countries (Wann, Melnick, Russell, & Pease, 2000), consists of seven items shown in Table 1. The SSIS was used to measure team identification for each team individually (Astros and Rangers).

In the discrete choice part of the questionnaire, respondents were presented with a series of choice scenarios, with attribute levels for baseball caps varying systematically according to an experimental design. The approach essentially assumes that any product (baseball cap in our case) can be described by its features and that individuals' preferences for them can be inferred by the choices they make in the experiment. In each scenario shown, students were presented with three options: two options described by varying attribute levels and a "none" option. Respondents were asked to indicate independently for each scenario which of the three options they would choose if they were the only options available to them. An example of a scenario and choice question is shown in Figure 1.

[FIGURE 1 OMITTED]

Development of the Attributes

Two exploratory studies were conducted to elicit the attributes and levels for the discrete choice experiment and to establish a pair of equally important and relevant social causes representing conditions of high and low fit with MLB teams. Attributes and levels associated with the choice of team-licensed baseball caps were identified primarily by the first exploratory study. In this first exploratory study, a direct questioning technique was used to elicit all the attributes students consider when purchasing a team-licensed cap. A convenient sample of 33 students was presented with four open-ended questions adapted from Louviere (1988) designed to elicit features that make baseball caps attractive and unattractive to them. The responses were coded and the frequencies of occurrences counted. Four main attributes emerged as the most frequently elicited by students: a) Team Logo, b) Front Design, c) Back Design, and d) Price.

In this first exploratory study, using an unaided recall method, we also elicited a list of social causes that students were able to recall and that they considered important and relevant. The following organizations emerged as the most frequently recalled: Susan G. Komen Breast Cancer Foundation, American Red Cross, American Cancer Society, UNICEF, and amfAR.

The second exploratory study (N=65) was designed to evaluate the strength of the relationship between two MLB teams (Houston Astros and Texas Rangers) and the social causes previously identified in the first study to establish pairs of equally important and relevant social causes: one social cause that represented a high fit and another that represented a low fit with the MLB teams. The reason why only two franchises were chosen is that these two professional teams were at least familiar to the subjects in Texas, where the study took place. In addition, for this second study, two cause-related programs (Baseball Tomorrow Fund and Boys and Girls Club) in which the MLB is currently involved were added to the list of causes. The addition of these two programs allowed us to include programs that were more relevant to the MLB brand, yet students were less familiar (not identified in the first study) with these two programs. Fit between MLB and eight social causes were measured using four Likert-scale items (1 = Does not fit at all/7 = Fit very strongly, 1 = Not similar at all/7 = Very similar, 1= Not consistent at all/7 = Very consistent, 1 = Not complementary at all/7 = Very complementary), adopted from Becker-Olsen et al. (2006). In this pretest, Baseball Tomorrow Fund (BTF) was identified as the cause with the highest fit with MLB, and Susan G. Komen Breast Cancer Foundation (SKB) as the cause with lowest fit with MLB in all four constructs. Fit varied as expected (high fit M=5.42 and low fit M=2.02). As a result of this study, two programs were selected to be included in the discrete choice experiment: (a) Baseball Tomorrow Fund (BTF), which was a high-fit program, and (b) Susan G. Komen Breast Cancer Foundation (SKB), which was a low-fit program. They were presented in the discrete choice experiment as levels of an attribute related to a transaction-based CRM program: "$1 donation to a [high-fit or low fit] charity," or "Not related to social causes." The attributes and levels used in the experiment are listed in Table 2.

Nevertheless, it is important to point out that SKB was a cause previously identified by the students as relevant and important to them, whereas BTF was not recalled by students. Consequently, these two causes inevitably created an interaction condition between fit and appeal: (a) a condition of high-fit, narrow appeal (low-familiarity/relevance)--BTF, and (b) a condition of low-fit, broad appeal (high-familiarity/relevance)--SKB. Given these conditions, it would be expected that SKB may be preferred over BTF because it is a cause with a broader appeal (more familiar, and potentially more relevant and important to the students). However, it would also be expected that those identified with the teams and the sport of baseball may favor BTF as a way to show their in-group favoritism.

Experimental Design

The choice experiment was designed using the balanced overlap method in Sawtooth Software's (2008) Choice-Based Conjoint software. The computer program generated 119 different versions of 12 choice scenarios with three alternatives each. Each participant in the study was randomly assigned to one of the different versions of the design (sometimes referred to as random designs). To discourage consumers from engaging in some form of task simplification (Orme, 2009), the design permitted some degree of level overlap, where an attribute like team logo could have identical levels across all concepts. With this design, a loyal fan did not always face a trivial choice task between two baseball caps of opposing teams. However, the design did not permit two identical concepts on all attributes to appear within the same task. Across all versions generated, each level of each attribute occured an approximately equal number of times within and across all versions, ascertaining near orthogonality between the attributes. One of the main advantages of this design is that it conveniently creates different combinations of features across many versions, which facilitates the estimation interaction effects between features. A discussion and comparison of different design strategies, including random designs, strategies using design plans such as those listed in Addelman (1962) and discussed in Louviere, Hensher, and Swait (2000), and efficient designs using computer search algorithms (Kuhfeld, Tobias, and Garratt, 1955) can be found in Chrzan and Orme (2000).

Determining the Sample Size

Determining the sample size for a discrete choice experiment study depends on many factors such as cost, heterogeneity of the sample, and statistical requirements from sampling theory (Orme, 2010). We followed the recommendations by Louviere, Hensher, and Swait (2000), Johnson and Orme (1996), and Orme (2010) to determine the minimum number of respondents and scenarios to be used in the discrete choice experiment. In a discrete choice, sample size can increase by either increasing the number of respondents in the study or by increasing the number of choice scenarios each respondent will evaluate. Higher number of scenarios becomes more economical from a sampling point of view, but also can increase respondent fatigue. Johnson and Orme (1996) recommend the number of scenarios to be between 10 and 20 and the number of total observations per number of cells to be 500 for an aggregate model. This study with 119 respondents and 12 scenarios with 3 alternatives each provided 4,284 total observations. Johnson and Orme (1996) propose to take the total number of observations excluding the "none" option, which is 2,856 (119 x 12 x 2), and divide it by the number of analysis cells (i.e., the largest number of levels for any one attribute in a main-effects model). In our main-effects model, this resulted in 952 observations per main effect, which is about two times higher than the rule of thumb (more than 500) recommended in Orme (2010).

According to the method described in Louviere et al. (2000), a sample size of 119 respondents with each responding to 12 scenarios also allowed us to predict a choice probability (or share) for any baseball hat to be approximately in the range of 20 percent or higher with a relative accuracy of 10 percent with a 95 percent confidence interval. However, it is important to recognize that these sample size calculations are just approximations since respondents were not a random sample of the population and were given multiple scenarios that were not completely independent. However, as Louviere et al. (2000) ascertain, practice has shown that well designed studies based on replicated scenarios per individual can yield similar parameter estimates that are similar to models based on single choices.

Data Analysis

An aggregate conditional logit model (McFadden, 1974) was first conducted to understand the relative impact of product attributes to students' decisions to purchase baseball caps as well as the moderating effects of identification. There were a total of 4,284 observations available for the model estimation (119 individuals x 12 scenarios x 3 alternatives per scenario). In a conditional logit model, choice probabilities depend only on the attributes of choice alternatives as the explanatory variables, and not on the characteristics of the individual making the choice.

In our study, team identification is an explanatory variable that is specific to individuals and constant throughout the choice experiment. It is exogenous to the choice model, and at a different level of variation. Because a respondent's level of team identification is the same across the choice experiment, team identification cannot help predict why a respondent chose one cap over another. However, at the aggregate level, it is possible to have interactions between team identification and attributes by creating product terms and entering them in the model (Allison, 1999). These interactions allow us to examine if the importance of an attribute varies according to individuals' characteristics. For example, when the individual variable is discrete, like gender, we can see if an attribute is more important for males versus females. For team identification, the interaction operates in a way that the importance of attributes increase or decrease according to the levels of the moderating variable. In this study, we expect that preference for team logos to increase with higher levels of individuals' team identification and test whether preference for a cause is also influenced by team identification.

One advantage of using an aggregate model is that it allows us to perform classical hypothesis testing. To test for the moderating effects of identification, we performed a likelihood ratio chi-square test (Hedges & Olkin, 1985) to compare model fit between the main effects model and the interaction model. The test measures the difference in deviance fit scores (-2 times log-likelihood) between the main and interaction effects model. A significant likelihood ratio statistic rejects the null hypothesis that the two models have an equal fit. The significance of the interaction terms provides evidence for the role that identification can play in determining consumer choice, especially in moderating the impact of a social cause program on choice.

Although using interaction terms in an aggregate model is a plausible solution to incorporate individuals' characteristics into the model and allow classical testing, we recognize that it is a pooled model that assumes individuals have similar preferences when comes to choosing a baseball hat. This can be a severe limitation as generally consumers are different and have different preferences for products. Therefore, we also reported the results of a multilevel model that overcomes this limitation and more naturally addresses individual interactions. For the past decade, the hierarchical Bayesian (HB) logit model (Allenby & Ginter, 1995; Rossi & Allenby, 2003) has become the most popular multilevel model to analyze discrete choice data. The HB model is a multilevel model because it simultaneously explains choice as a function of both an upper-level model, which is pooled across respondents, and a lower-level model, which is an individual, within-respondents model. The lower-level model explains choice based on the attributes from the experimental data following a logit model. The upper-level-model explains the variation on the lower-level coefficients (heterogeneity) across the population of respondents. Intuitively, the HB model optimally uses the information for the entire sample to adjust the utility estimates for each individual. When individual variables like team identification are entered in an HB model, they are entered in the upper-level model to explain the variation of lower-level coefficients. Therefore, they allow the individual coefficient estimates to depend on individual characteristics. A non-technical introduction to HB and its estimation procedures can be found in Howell (2009) and Orme and Howell (2009).

The predictive validity of the models was compared by reserving the last choice scenario of each respondent for predictive testing. The models were re-estimated without the individual responses from the last choice scenario, and the parameter estimates were used to predict the individual choices for the holdout scenario. The predicted choices were then compared to the actual choices to assess model internal consistency. The results from the most predictive model were used to determine the relative impact of attributes and the interactions on students' choices. The relative importance of the attributes was estimated by using a probability analysis method discussed in Lancsara, Louviere, and Flynn (2007), which is relatively simple to implement. The analysis consists of calculating the probability of choosing an alternative given a particular level of an attribute. To implement the analysis, we first calculated the probability of choosing a baseball cap alternative in a base case scenario. We created a base case scenario composed of two baseball cap offerings (Astros vs. Rangers cap, where all other attributes were set to their mean values), and a "none" option. We then systematically changed the levels of each attribute, one at a time, and recorded the absolute percent change in probabilities relative to the base case. These percent changes were compared across all attributes to provide the relative importance of the attributes.

We also used probability analysis using the same aforementioned base case to graphically portray the impact of the interaction between team identification and the social cause attribute on respondents' choices. In the analysis, we varied social cause with different levels of team identification and recorded the resulting probabilities.

Results

Respondent Characteristics

There were a total of 119 participants in this study. The sample was comprised of 60.5% males and 39.5% females. The average age of respondents was 21.33, and the ethnic background included primarily Caucasians (79%) and Hispanics (9.2%). Regarding team identification, respondents reported a higher mean team identification score for the Astros (M=3.12, SD=2.06) compared to the scores for the Rangers (M=2.33, SD=1.80). The difference in mean scores was statistically significant (t=2.857, p<0.01).

Estimation Results

Table 3 shows the results of the aggregate conditional logit model including both main and interaction effects and the HB model (1) with team identification included as a covariate. The models generally agree in terms of what attributes had significant impact on students' choice. However, the HB model had a better predictive validity than the aggregated model. The HB model correctly predicted 88% of the individual choices in the holdout tasks, compared to only 60-65% of choices correctly predicted using either one of the aggregated models.

The result of the likelihood ratio chi-square test comparing the aggregate main-effects model to the aggregate interaction-effects model indicates that the interaction effects had a statistically significant contribution to model fit ([chi square] = 472.79, df=16, p< .001). The same significant interaction estimates are also evidenced in the HB model. (2) As expected, the interactions between team identification and logo were statistically significant for both Astros and Rangers. Therefore, the higher the identification for a team was, the more students preferred a baseball cap with the respective team's logo. The interaction between identification with Astros and social cause was also significant, indicating an increasing preference for BTF, a high-fit and narrow appeal (low-familiarity) beneficiary, as identification with Astros increased. The interaction between identification with Rangers and social cause was not significant, although the coefficients had the same sign as the interaction with Astros. This can be potentially explained by the fact the sample had a significantly higher affinity toward the Astros compared to the Rangers as explained above.

Relative Importance of Attributes

Table 3 shows the results of the choice probability analysis using the HB model. The base case was comprised of three alternatives: a) Astros cap with all other attributes at their mean values, b) Rangers cap with all other attributes at their mean values, and c) a "none" option. Predicted probabilities for the base case across the three alternatives were 30.2%, 23.2%, and 46.6% for Astros, Rangers, and "none" options respectively. The table shows the percent change in probability from the base case to the case including each attribute one at a time. The implied order of 'impact' is presented in the last column. As shown in Table 4, front design had the highest impact on students' choices with a 65.6% absolute change in shares as a result of moving from the base case to a front design with a curved peak. Changing the base case to a different logo had the second highest impact (37.4%), followed by changing the base case to SKB, a low-fit/broader appeal cause (33%), and to a low price (32%).

The choice probabilities for each feature shown in Table 4 also provide a preference ordering for each level within an attribute. The levels that garnered the highest choice probabilities within each attribute, holding everything else equal, were the Astros logo, curved peak front design, the elastic back design, the SKB (low-fit/broad appeal) social cause, and low price.

Table 5 illustrates the results of the simulated preference shares for three baseball cap alternatives (Astros, Rangers, and none) at different social causes and identification values, holding everything else constant at the base case. Because in the HB model we have probabilities for each individual, we can estimate probabilities for groups that have different identification levels. We examined probabilities for individuals classified into three groups based on the identification with Astros: high ID (individuals that averaged 6 or 7 in the identification scale with Astros), medium ID (individuals that averaged 3 to 5 in the identification scale with Astros), and low ID (individuals that averaged 1 or 2

in the identification scale with Astros). As expected given the interaction of logo and team identification, preference shares for the Astros baseball increase and shares for the Rangers decrease with higher identification levels with the Astros. However, in Figure 2, we can visualize the nature of the interaction between team identification and social cause. On average, we see a higher preference for SKB over BTF among respondents. However, when we take identification into account, that preference for SKB is mostly evident within those that are lowly identified with the Astros, and dissipates among those that are highly identified with the Astros. Although the probability of purchasing the Astros cap among those lowly identified with the Astros is small, the likelihood of this group purchasing it to benefit SKB (26.4%) is two times greater than purchasing it to benefit BTF (12%).

Discussion

This study sought to examine the moderating role of identification on the effectiveness of cause-related programs. Using a discrete choice experiment, this study investigated how students evaluate CRM initiatives against other attributes when they buy team licensed products. Three main findings were noteworthy. First, the offering of a cause program was an important attribute that influenced students' decisions to purchase baseball caps. Second, confirming our expectations, students were more likely to choose a CRM program with a broad appeal, despite its lower degree of fit with the sponsoring organization. Third, team identification moderated the evaluations of different cause-related programs and their impact on choice.

The first finding replicates previous studies indicating that a cause-related marketing strategy can have an effective influence on consumer choice (Barone et al., 2000; Pracejus & Olsen, 2004). However, the cause-related program was identified as one of the top most important attributes; showing that the students were willing to trade off other attributes to support a cause of their interest. This may be explained by the product category used in this study. According to Strahilevitz and Myer (1998), charity incentives work better for frivolous products where easier trade-offs are made (e.g., less expensive products, low risk, etc.) than for practical ones when consumers respond to CRM programs. This would also be consistent with the results of Barone et al. (2000), which suggest that harder trade-offs between social causes and quality or price for categories like personal computers could be more difficult to make (larger trade-offs) than the trade-offs respondents experienced for a baseball cap (smaller trade-offs).

The second finding, that students selected more baseball caps that supported Susan G. Komen Breast Cancer Foundation, warrants further discussion. It is important to note that individuals were observed to have distinct preferences between two cause-related programs regardless of their degree of fit with the brand. Supporting expectations, students preferred the cause with which they were more familiar and which was relevant as opposed to the one with a higher degree of fit. This result showed that in this study, being familiar with and considering the cause relevant and important superseded judgments of brand fit. Although the implication of the finding seems straightforward, essentially implying that sport organizations are better off seeking organizations with broad appeal to their CRM campaigns, our results can't generalize to the many different types of cause organizations that exist. In particular, it is possible that fit could have been a more influential driving force, as the literature suggests, if a high-fit organization that is also as familiar and relevant as SKB were also included in the study.

More significantly, team identification has been found to play a moderating role in the relationship between CRM and consumer choice. Although preference for an Astros cap supporting SKB was higher than an Astros cap supporting BTF, that difference is almost nullified for respondents highly identified with Astros. This finding contrasts the original argument that highly identified individuals would be more willing to purchase products if the purchase is perceived as a way to support a cause they care more about. Although we saw that preference for a beneficiary congruent with baseball, like BTF, increased with team identification, we did not observe a difference in preference between the two programs at the highest levels of team identification. We did observe that preference improved for BTF and decreased for SKB as identification increased. The results seem to support the notion that highly identified individuals were mostly driven by their sole desire to purchase an Astros cap as a way to demonstrate their affiliation regardless of the support of a cause included in our test. Among those highly motivated individuals, a social cause had little impact. The impact of a social cause was more evident among those who were lowly identified with the Astros, and in this case a more familiar and relevant cause, SKB, was more preferred. This seems to indicate that social causes exert more influence and provide an extra reason to buy team-licensed products when motivation to buy is low, like among those lowly identified with a team.

From the practitioner standpoint, selecting proper social causes to support is a critical decision that can impact the effectiveness of a CRM program. Previous CRM literature supported the notion that high fit associations (e.g., baseball team supports youth baseball leagues) have positive impacts on consumer behavior (Becker-Olsen et al., 2006; Pracejus & Olsen, 2004). In addition, as discussed in Varandarajan and Menon (1988), both the salient characteristics of the cause-related effort such as image fit and the appeal potential of the cause are important factors that influence firms' partnering choices. However, this study shows that when familiarity and relevance are not equal between two social programs, they can supersede the impact of fit on consumer choice. Therefore, it would be more advantageous for sport marketers to not consider brand fit in a vacuum, but also examine and measure the degree of familiarity and relevance of different causes (their appeal potential). This can be particularly beneficial when targeting consumers that have low team identification.

Despite the contributions of this study, there were some limitations that should be mentioned. First, this study was limited to testing conditions of either high familiarity but low fit, or less familiarity and high fit. Therefore, other conditions of high-fit/high familiarity and low-fit/low familiarity should be assessed in future studies to arrive at more conclusive answers regarding the impact of fit, familiarity, and team identification on choices. In addition, the diversity of cause organizations is also an important aspect to be examined in future studies. Some causes may be considered as creating more "good" than others, while other causes may be perceived to be controversial, potentially having a negative impact on consumer choice. Furthermore, future studies should consider including other influential factors not included in this study such as motivations to engage in CRM program (Barone et al. 2000), timing (Becker-Olsen et al., 2006), and donation magnitude (Strahilevitz, 1999). Finally, this study was also limited to a small and more homogeneous sample, consisting of college students. People belonging to different genders, age groups, education levels, and socioeconomic status may respond differently toward different CRM initiatives.

In conclusion, this study shows that identification plays an important role in the effectiveness of CRM programs. The results were noteworthy and warrant further investigations to enhance our understanding of the role it can play in a broader theoretical context.

References

Addelman, S. (1962). Orthogonal main effects plans for asymmetrical factorial experiments. Technometrics, 4, 21-46.

Allenby, G. M., & Ginter, J. L. (1995). Using extremes to design products and segment markets. Journal of Marketing Research, 32, 392-403.

Allison, P. D. (1999). Logistic regression using the SAS system: Theory and application. Cary, NC: SAS Institute and Wiley.

Ashforth, B. E., & Mael, F. (1989). Social identity theory and the organization. Academy of Management Review, 14, 20-39.

Barone, M. J., Miyazaki, A. D., & Taylor, K. A. (2000). The influence of cause-related marketing on consumer choice: Does one good turn deserve another? Journal of the Academy of Marketing Science, 28, 248-262.

Barone, M. J., Norman, A. T., & Miyazaki, A. D. (2007). Consumer response to retailer use of cause-related marketing: Is more fit better? Journal of Retailing, 83(4), 437-445.

Becker-Olsen, K. L., Cudmore, B. A., & Hill, R. P. (2006). The impact of perceived corporate social responsibility on consumer behavior. Journal of Business Research, 59(1), 46-53.

Bloom, P. N., Hoeffler, S., Keller, K. L., & Meza, C. E. B. (2006). How social-cause marketing affects consumer perceptions. MIT Sloan Management Review, 47(2), 49-55.

Branscomb, N. R., & Wann, D. L. (1994). Collective self-esteem consequences of outgroup derogation when a valued social identity is on trial. European Journal of Social Psychology, 24(6), 641-657.

Calder, B. J., Phillips, L. W., & Tybout, A. M. (1981). Designing research for application. Journal of Consumer Research, 8(2), 197-207.

Chrzan, K., & Orme, B. (2000). An overview and comparison of design strategies for choice-based conjoint analysis. Retrieved from http://www.sawtoothsoftware.com/download/techpap/desgncbc.pdf

Cornwell, T. B., & Coote, L. V. (2005). Corporate sponsorship of a cause: The role of identification in purchase intent. Journal of Business Research, 58(3), 268-276.

Dimmock, J. A., & Grove, J. R. (2003). Relationship of fan identification to determinants of aggression. Journal of Applied Sport Psychology, 55, 37-41.

Fisher, R. J., & Wakefield, K. (1998). Factors leading to group identification: A field study of winners and losers. Psychology & Marketing, 15, 23-40.

Gupta, S., & Pirsch, J. (2006). The company-cause-customer fit decision in cause-related marketing. Journal of Consumer Marketing, 23(6), 314-326.

Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. New York: Academic Press, INC.

Howell, J. (2009). CBC/HB for beginners. Retrieved from http://www.saw toothsoftware.com/download/techpap/CBCHBbeginners.pdf

International Event Group (IEG). (2009). Sponsorship spending. Retrieved from http://www.sponsorship.com/resources/sponsorshipspending.aspx

Johnson, R. M., & Orme, B. K. (1996). How many questions should you ask in choice-based conjoint studies? Retrieved from http://www.sawtoothsoftware.com/download/techpap/howmanyq.pdf

Keller, K. L. (1993). Conceptualizing, measuring, managing customer-based brand equity. Journal of Marketing, 57(1), 1-22.

Kuhfeld, W., Tobias, R. D., & Garratt, M. (1995, November). Efficient experimental designs with marketing research applications. Journal of Marketing Research, 31 , 545-557.

Kwon, H. H., & Armstrong, K. L. (2002). Factors influencing impulse buying of sport team licensed merchandise. Sport Marketing Quarterly, 11(3), 151-163.

Lancsara, E., Louviere, J. J., & Flynn, T. (2007). Several methods to investigate relative attribute impact in stated preference experiments. Social Science & Medicine, 64, 1738-1753.

Louviere, J. J. (1988). Analyzing decision making: Metric conjoint analysis. Newbury Park, CA: Sage Publications, Inc.

Louviere J. J., Hensher, D. A., & Swait, J. D. (2000). Stated choice methods: Analysis and applications. New York: Cambridge University Press.

Louviere, J. J., & Woodworth, G. (1983). Design and analysis of simulated consumer choice or allocation experiments: An approach based on aggregate data. Journal of Marketing Research, 20, 350-367.

Madrigal, R. (1995). Cognitive and affective determinants of fan satisfaction with sporting event attendance. Journal of Leisure Research, 27, 205-227.

McFadden, D. (1974). Conditional logit analysis of qualitative choice behavior. In P. Zarembka, Frontiers in Econometrics (pp. 105-142). New York: Academic Press.

Orme, B. (2009). Fine-tuning CBC and adaptive CBC questionnaires. Retrieved from http://www.sawtoothsoftware.com/download/techpap/finetune.pdf

Orme, B. (2010). Getting started with conjoint analysis: Strategies for product design and pricing research (2nd ed.). Madison, WI: Research Publishers LLC.

Orme, B., & Howell, J. (2009). Application of covariates within Sawtooth Software's CBC/HB program: Theory and practical example. Retrieved from http://www.sawtoothsoftware.com/download/techpap/HBC ovariates.pdf

Pracejus, J. W., & Olsen, G. D. (2004). The role of brand/cause fit in the effectiveness of cause-related marketing campaigns. Journal of Business Research, 57(6), 635-640.

Rossi, P. E., & Allenby, G. M. (2003). Bayesian statistics and marketing. Marketing Science, 22, 304-328.

Sawtooth Software, Inc. (2008). CBC v6.0 technical paper. Retrieved from http://www.sawtoothsoftware.com/download/techpap/cbctech.pdf

SBRnet. (2009). Baseball (MLB) sports logo apparel: Summary report. Retrieved from http://www.sbrnet.com/research.asp?subrid=700

Simmons, C. J., & Becker-Olsen, K. L. (2006). Achieving marketing objectives through social sponsorships. Journal of Marketing, 70(4), 154-169.

Sloan , L. R. (1989). The motives of sports fans. In J. H. Goldstein (Ed.), Sports, games, and play: Social & psychological viewpoints (2nd ed., pp. 175-240). Hillsdale, NJ: Lawrence Erlbaum Associates.

Strahilevitz, M. (1999). The effects of product type and donation magnitude on willingness to pay more for a charity-linked brand. Journal of Consumer Psychology, 8(3), 215-241.

Strahilevitz, M., & Myers, J. G. (1998). Donations to charity as purchase incentives: How well they work may depend on what you are trying to sell. Journal of Consumer Research, 24, 434-446.

Tajfel, H. (1982). Social identity and intergroup relations. New York: Cambridge University Press.

Tajfel, H., & Turner, J. C. (1979). An integrative theory of intergroup conflict. In W. G. Austin & S. Worchel (Eds.), The social psychology of intergorup relations (pp. 33-47). Monterey, CA: Brooks Cole.

Varadarajan, P. R., & Menon, A. (1988). Cause-related marketing: A coalignment of marketing strategy and corporate philanthropy. Journal of Marketing, 52(3), 58-74.

Wann, D. L., & Branscomb, N. R. (1992). Emotional responses to the sports page. Journal of Sport and Social Issues, 16(1), 49-64.

Wann, D. L., & Branscomb, N. R. (1993). Sports fans: Measuring degree of identification with the team. International Journal of Sport Psychology, 24, 1-17.

Wann, D. L., & Dolan, T. J. (1994). Spectators' evaluations of rival and fellow fans. Psychological Record, 44(3), 351-359.

Wann, D. L., Melnick, M. J., Russell, G. W., & Pease, D. G. (2000). Sport fans: The psychology and social impact of spectators. New York: Routledge.

Wann, D. L., Peterson, R. R., Cothran, C., & Dykes, M. (1999). Sport fan aggression and anonymity: The importance of team identification. Social Behavior and Personality: An International Journal, 27, 597-602.

Wann, D. L., Royalty, J., & Roberts, A. (2000). The self-presentation of sport fans: Investigating the importance of team identification and self-esteem. Journal of Sport Behavior, 23(2), 198-201.

Wann, D. L., Royalty, J., & Rochelle, A. (2002). Using motivation and team identification to predict sport fans' emotional responses to team performance. Journal of Sport Behavior, 25(2), 207-215.

Endnotes

(1) Table 3 shows the HB estimates from the upper-level model with two sets of weights: the mean population utility estimates associated with each attribute when the team identification variables are zero (constants of the upper-level model equivalent to main-effects), as well as the regression weights associated with the two identification variables (interaction-effects) that represent the adjustment in the population mean for each attribute that is expected for each unit increase in the identification variables.

(2) Comparing an HB model with and without the team identification variables also indicated an improvement in model prediction represented by lower deviance score (-2 log likelihood) when team identification variables are included in the model.

Jaedeock Lee, PhD, is an assistant professor in the Department of Sport Management at East Stroudsburg University of Pennsylvania. His research interests include corporate social responsibility in sport and its effects on consumer behavior. Mauricio Ferreira, PhD, is the head of marketing science at Hypothesis Group and a faculty member at the University of San Francisco. His research interests are centered on understanding consumer behavior with a particular emphasis on consumer choice.

Table 1.
The modified scale measuring fan identification with Houston Astros

In this section, I would like to start by asking you to answer
the following questions regarding the Houston Astros. On a scale
from 1 to 7, where 1 is "not agree at all" and 7 is "completely
agree," please indicate the degree to which you agree to the
following statements regarding the Houston Astros.

 Disagree Agree

To me, it is important that the Houston 1 2 3 4 5 6 7
Astros win.

I see myself as a fan of the Houston 1 2 3 4 5 6 7
Astros.

My friends see me as a fan of the Houston 1 2 3 4 5 6 7
Astros.

During the season, I follow the Houston 1 2 3 4 5 6 7
Astros via ANY of the following: in person
or on television, on the radio, or
televised news or a newspaper.

Being a fan of the Houston Astros is 1 2 3 4 5 6 7
important to me.

I dislike the greatest rivals of the 1 2 3 4 5 6 7
Houston Astros.

I display the Houston Astros' name or 1 2 3 4 5 6 7
insignia at my place of work, where I
live, or on my clothing.

Table 2.
Attributes and levels

Attributes Levels

Team logo Houston Astros
 Texas Rangers

Front Design Curved Peak
 Flat Peak

Back Design Buckle
 Velcro
 Elastic

Social Cause $1 donated to Baseball Tomorrow Fund
 (High Fit/Narrow Appeal)
 $1 donated to Susan G. Komen Breast
 Cancer Foundation (Low Fit/Broad Appeal)
 No CRM

Price Low ($17.99)
 Medium ($24.99)
 High ($31.99)

Table 3.
Estimation Results

Variables Main-Effects
 Conditional Logit
 Model

 [beta] SE

Team logo (Astros) 0.186 *** 0.038
Front design (curved peak) 0.506 *** 0.040
Back design (buckle) 0.081 0.057
Back design (velcro) -0.210 *** 0.059
Social Cause--(BTF) (b) 0.021 0.057
Social Cause--(SKB) 0.198 *** 0.056
Price (low) 0.201 *** 0.056
Price (medium) -0.001 0.057
Astros ID * team logo (Astros)
Astros ID * front design (curved peak)
Astros ID * back design (buckle)
Astros ID * back design (velcro)
Astros ID * social cause (high fit)
Astros ID * social cause (low fit)
Astros ID * price (low)
Astros ID * price (medium)
Rangers ID * team logo (Astros)
Rangers ID * front design (curved peak)
Rangers ID * back design (buckle)
Rangers ID * back design (velcro)
Rangers ID * social cause (high fit)
Rangers ID * social cause (low fit)
Rangers ID * price (low)
Rangers ID * price (medium)
None 0.247 *** 0.058
Astros ID * None
Rangers ID * None
-2 log likelihood 2888

Variables Interaction-Effects
 Conditional Logit
 Model

 [beta] SE

Team logo (Astros) 0.183 *** 0.047
Front design (curved peak) 0.681 *** 0.048
Back design (buckle) 0.061 0.065
Back design (velcro) -0.276 *** 0.067
Social Cause--(BTF) (b) 0.006 0.065
Social Cause--(SKB) 0.243 0.065
Price (low) 0.282 0.064
Price (medium) 0.031 0.066
Astros ID * team logo (Astros) 0.270 *** 0.024
Astros ID * front design (curved peak) -0.008 0.023
Astros ID * back design (buckle) -0.055 0.033
Astros ID * back design (velcro) -0.079 * 0.034
Astros ID * social cause (high fit) 0.087 ** 0.033
Astros ID * social cause (low fit) -0.089 ** 0.033
Astros ID * price (low) -0.018 0.032
Astros ID * price (medium) -0.055 0.033
Rangers ID * team logo (Astros) -0.388 *** 0.030
Rangers ID * front design (curved peak) 0.054 0.028
Rangers ID * back design (buckle) -0.063 0.040
Rangers ID * back design (velcro) -0.149 *** 0.041
Rangers ID * social cause (high fit) 0.056 0.039
Rangers ID * social cause (low fit) -0.02 0.040
Rangers ID * price (low) -0.05 0.038
Rangers ID * price (medium) -0.052 0.040
None 0.514 *** 0.065
Astros ID * None
Rangers ID * None
-2 log likelihood 2415

Variables Hierarchical Bayes
 Model, with
 Covariates a

 Mean [beta] SD

Team logo (Astros) 0.863 +++ 0.222
Front design (curved peak) 2.629 +++ 0.359
Back design (buckle) 0.292 0.239
Back design (velcro) -0.927 +++ 0.250
Social Cause--(BTF) (b) -0.044 0.212
Social Cause--(SKB) 0.715 ++ 0.245
Price (low) 0.889 +++ 0.229
Price (medium) 0.160 0.209
Astros ID * team logo (Astros) 0.987 +++ 0.128
Astros ID * front design (curved peak) 0.038 0.145
Astros ID * back design (buckle) -0.063 0.121
Astros ID * back design (velcro) -0.381 ++ 0.131
Astros ID * social cause (high fit) 0.191 ++ 0.105
Astros ID * social cause (low fit) -0.197 ++ 0.121
Astros ID * price (low) -0.095 0.104
Astros ID * price (medium) -0.180 ++ 0.107
Rangers ID * team logo (Astros) -1.248 +++ 0.154
Rangers ID * front design (curved peak) 0.088 0.163
Rangers ID * back design (buckle) -0.248 + 0.141
Rangers ID * back design (velcro) -0.452 ++ 0.146
Rangers ID * social cause (high fit) 0.192 0.122
Rangers ID * social cause (low fit) -0.114 0.136
Rangers ID * price (low) -0.180 0.127
Rangers ID * price (medium) -0.066 0.124
None 1.972 ++ 0.546
Astros ID * None 0.105 0.255
Rangers ID * None -0.276 0.291
-2 log likelihood 919

* p<.05. ** p<.01. *** p<.001.

(a.) These are the upper-level estimates for the Hierarchical
Bayes model. They are based on the average of the last 10,000
draws generated in the CBC/HB program. The estimates show both
the mean population utility estimates when the team
identification variables are zero (constants), as well as the
regression weights associated with the two identification
variables (interaction-effects) that represent the adjustment in
the population mean that is expected for each unit increase in
the identification variables. We have highlighted with an
asterisk (*) any param- eters where the draws are >95% positive
or >95% negative, with two asterisks (**) where the draws are
>99% pos- itive or >99% negative, and three asterisks (***) where
the draws are >99.9% positive or >99.9% negative.

(b.) BTF--Baseball Tomorrow Fund. SKB--Susan G. Komen Breast
Cancer Foundation

Table 4.
Choice Probability Analysis: Relative Importance of Attributes

 Probability of Choice

Attribute Astros Rangers None

Base Case (a) 30.2% 23.2% 46.6%
Team Logo
 Team logo (Astros) 30.2% 21.8%
 Team logo (Rangers) 18.9% 23.2%
Front Design
 Front design (curved peak) 49.9% 38.2%
 Front design (flat peak) 20.0% 15.2%
Back Design
 Back design (buckle) 33.6% 25.8%
 Back design (velcro) 27.2% 22.2%
 Back design (elastic) 34.2% 26.4%
Social Cause
 BTFb 28.7% 22.1%
 SKB 38.2% 31.1%
 No CRM 25.8% 18.0%
Price
 Price (low) 38.0% 28.6%
 Price (medium) 31.9% 26.2%
 Price (high) 24.6% 17.1%

 Absolute Percent Change

Attribute Astros Rangers Maximum Order
 Impact
Base Case (a)
Team Logo
 Team logo (Astros) 0.0% 6.3% 6.3% 12
 Team logo (Rangers) 37.4% 0.0% 37.4% 2
Front Design
 Front design (curved peak) 65.6% 64.2% 65.6% 1
 Front design (flat peak) 33.6% 34.8% 34.8% 3
Back Design
 Back design (buckle) 11.5% 11.1% 11.5% 10
 Back design (velcro) 9.8% 4.6% 9.8% 11
 Back design (elastic) 13.4% 13.5% 13.5% 8
Social Cause
 BTFb 4.7% 4.8% 4.8% 13
 SKB 26.8% 34.0% 34.0% 4
 No CRM 14.6% 22.5% 22.5% 7
Price
 Price (low) 25.9% 22.9% 25.9% 6
 Price (medium) 5.7% 12.8% 12.8% 9
 Price (high) 18.4% 26.3% 26.3% 5

(a.) Base case for each alternative includes the respective logo
and all other attributes set to their mean values.

(b.) BTF--Baseball Tomorrow Fund. SKB--Susan G. Komen Breast
Cancer Foundation

Table 5.
Choice Probability Analysis: Interaction between
social cause and team identification

Social Cause Attribute/ Choice Probabilities
Team ID for Baseball Cap Options

Baseball Tomorrow Fund (BTF) Astros Rangers None

 Low Astros ID 12.00% 37.49% 50.51%
 Medium Astros ID 41.67% 17.01% 41.32%
 High Astros ID 53.99% 0.88% 45.13%

Susan Susan G. Komen (SKB) Astros Rangers None

 Low Astros ID 26.45% 29.39% 44.16%
 Medium Astros ID 47.91% 12.36% 39.73%
 High Astros ID 55.10% 0.91% 43.99%

Figure 2.
Probabilities of choosing an Astros cap at different levels of Astros
Team Identification and social causes (Baseball Tomorrow Fund and
Susan G. Komen)

Probability of Choice

Astros Team ID

 Low Medium High

BTF (High Fit/Narrow Appeal) 12.0% 41.7% 54.0%
SKB (Low Fit/Broad Appeal) 26.4% 47.9% 55.1%

Note: Table made from line graph.