The worldwide sponsorship market was valued at $46.3 billion in 2010, when total spending by U.S. and Canadian companies hit $17.2 billion, up from the $14.91 billion in 2007. The range of sponsored activities has increased steadily, though sports remains the most important area, garnering 68% of total spending (International Events Group 2010). The increase in sponsorship activities reflects the growing awareness that developing a brand through associations with an event may build brand equity more effectively than can traditional marketing communications, such as advertising (Keller 2003). Corporations thus seek to increase brand equity and revenues by building brand awareness, brand image, and brand loyalty. However, whereas the influence of sponsorship activities on brand awareness and brand image is relatively well established, their effects on brand loyalty remain uncertain. With this study, we therefore outline and explain the impact of sponsorship on an ultimate communication objective, namely, brand loyalty. Our focus is on sponsorship effects specifically, not on a comparison of sponsorship and advertising effects.

As noted, mechanisms and measures to determine the influence of sponsorship on brand awareness and brand image are extensive (Dean 1999; Javalgi et al. 1994; Lardinoit and Derbaix 2001; for a review, see Cornwell 2008), yet we know little about the mechanisms by which sponsorship influences brand loyalty or the magnitude of this influence. Both academics and professionals have called for more research on the effects of sponsorships on brand loyalty (Cliff and Motion 2005; Cornwell et al. 2001). To the best of our knowledge though, only one study has explored the impact of sponsorship on brand loyalty (Sirgy et al. 2008), showing that self-congruity with a sponsored event has a small and positive influence on customer loyalty toward the sponsor brand. Even these results are somewhat inconsistent though, because no effect emerges for two of the five samples in that study. Although Sirgy and colleagues offer an important initial consideration of sponsorship effects on brand loyalty, their cross-sectional study contains only one antecedent of loyalty (i.e., self-congruity) and cannot reveal the mechanisms through which sponsorship influences brand loyalty. We aim to address this gap.

Furthermore, we note the many constructs identified in prior sponsorship literature that influence consumer reactions to sponsorships, such as brand-event fit (Rifon et al. 2004; Speed and Thompson 2000), affect toward the event (Crimmins and Horn 1996; Speed and Thompson 2000), and attitude toward the sponsorship (Meenaghan 2001; Quester and Thompson 2001). Research has established sponsorship effects on outcomes such as brand affect and purchase intent (Cornwell 2008), though not brand loyalty (cf. Sirgy et al. 2008). Olson (2010) even highlights that most sponsorship research remains focused on a limited number of causal relationships; thus there is a need for comprehensive models with several causal relationships to understand and explain sponsorship effects more fully. Beyond sponsorship literature, research has revealed that brand affect and brand trust influence brand commitment and loyalty (Chaudhuri and Holbrook 2001; Dick and Basu 1994; Kim et al. 2008). Therefore, we build on established relationships from both sponsorship literature and consumer-brand relationship research to extend previous models of sponsorship effects.

In so doing, we fill two critical gaps in the sponsorship literature. First, we seek to demonstrate the causality between sponsorship and brand loyalty in a realistic setting and to model the process, such that we enable a better understanding of the effects of a brand’s event sponsorship on consumer loyalty to that brand. Second, we show that consumers’ self-congruity with the event and perceptions of the fit between the sponsor and the event help explain the impact of sponsorship on brand loyalty. The first process suggests an affective transfer from the event to the sponsored brand, which influences brand loyalty. The second process influences both brand affect, through attitude toward sponsorship, and brand trust, which in turn exert impacts on brand loyalty. Brand affect appears therefore to be an important mediating variable of sponsorship effects on brand loyalty.

In the remainder of this article, we review key concepts from the sponsorship and consumer-brand relationship literature to build our conceptual model of the impact of sponsorship on brand loyalty. With this theoretical foundation, we formulate a set of hypotheses that describe the relationships among the model constructs. We then detail our experimental study, based on a major sporting event (Olympic Games), which we conducted to test our model of sponsorship effects on brand loyalty. Finally, we provide the research results, limitations, and directions for further research.

Main concepts and research hypotheses

We define commercial sponsorship as an investment, in cash or kind, in an event, person, or idea for the purpose of exploiting the commercial potential of this association (Meenaghan 1983). The main communication objectives of sponsorships are to increase brand awareness and improve brand image or consumer attitudes toward the brand. Other objectives may include motivating personnel or building closer relationships with partners (e.g., suppliers, customers, prospective customers) (Erikson and Kushner 1999). Most sponsorship research therefore concentrates on modeling and understanding the process by which sponsorships help firms achieve their communication objectives (Cornwell et al. 2005), according to three main mechanisms: congruence between the sponsor and the event, signaling theory, and affect transfer. These theories underlie our proposed conceptual model as well, in which we predict that sponsor-event congruence and self-congruency with the event influence brand affect, brand trust, and thus ultimately brand loyalty.

Sponsor-event congruence

Sponsor-event congruence (or fit)Footnote 1 increases sponsorship efficiency. Speed and Thompson (2000, p. 230) define this construct as “the degree to which the pairing [of an event and sponsor] is perceived as well matched or a good fit, without any restriction on the basis used to establish fit.” Sponsorship research also has adopted a consumer-based conceptualization of congruence, as in the brand extension literature (Aaker and Keller 1990), such that the perception by spectators or participants determines the degree to which sponsors and the event match, belong to the same world, or seem likely to engage in joint business or communication efforts. In the endorsement literature, the match-up hypothesis also suggests that a strong fit between an endorser and a product type creates favorable attitudes toward the endorsed brand (Kamins 1990; Till et al. 2008). Sponsor-event fit thus represents a key influence on consumer responses to sponsorship.

The composition of a sponsor-event pair provides a natural form of congruity or incongruity that influences attitudes; it is important primarily because of its implications for partner selection and its influence on consumer relationships with the brand. Sponsorships with high fit are consistent with consumers’ expectations of the firm. Therefore, they increase the likelihood that spectators can identify the correct sponsor (Johar and Pham 1999; Johar et al. 2006), encourage positive brand attitudes (Becker-Olsen and Simmons 2002; Gwinner and Eaton 1999; Roy and Cornwell 2003), augment purchase intentions (Olson and Thjomoe 2009), and increase brand equity (Olson 2010; Simmons and Becker-Olsen 2006). Sponsorships with low fit instead make negative associations more accessible; through attitude priming (Fazio and Williams 1986; Houston and Fazio 1989), this negative affect is likely to engender other negative thoughts and result in an unfavorable attitude toward the sponsorship by that brand.

However, Olson and Thjomoe (2009), using realistic sponsorship stimuli, and Trendel and Warlop (2005), using implicit measures, demonstrate that low fit sponsors actually may benefit from stronger identification than do high fit sponsors. Because people find some incongruence interesting, such low fit could generate positive effects (Meyers-Levy and Tybout 1989), particularly if spectators view the sponsorship as philanthropic (D’Astous and Bitz 1995), consider the sponsored event important and significant (Speed and Thompson 2000), or regard the association as funny and creative and if they exhibit a high need for cognition (Masterson 2005).

Faced with these conflicting results, we follow the dominant findings: Simmons and Becker-Olsen (2006) and Olson (2010) show that high fit sponsorships lead to more favorable attitudes. Meenaghan (2001) also proposes that sponsorship intervenes in the emotional relationship between consumers and the event. When consumers perceive the association between a sponsor and an event as congruent, they tend to assume a goodwill motivation and build positive feelings toward the brand. We predict:

  1. H1:

    Perceived fit between the brand and the event relates positively to attitude toward the sponsorship by the brand.

Following Chaudhuri and Holbrook (2001), we define brand affect as the positive emotional feelings of consumers toward the brand. Accordingly, positive attitudes toward the brand, which are the consequences of the perceived fit between the event and the brand, result in higher brand affect. Olson (2010) demonstrates the positive relationship between attitude toward the sponsorship and sponsor equity, which represents the strongest relationship in his model. Two of the three measures of sponsor equity are positive feelings toward the brand and brand overall liking, both of which link to brand affect. We therefore include brand affect in our model as a generally strong predictor of brand loyalty. It thus follows that:

  1. H2:

    Attitude toward the sponsorship relates positively to brand affect.

Brand trust is another well-established determinant of brand loyalty (Ganesan and Hess 1997; Morgan and Hunt 1994). This form of trust refers to consumers’ expectations about the brand’s reliability in a risky situation or willingness to rely on the brand to perform its stated functions (Chaudhuri and Holbrook 2001; Delgado-Ballester et al. 2003). Trust comprises three dimensions that reflect consumers’ beliefs about the brand’s integrity, benevolence, and competence (Ganesan and Hess 1997; Schlosser et al. 2006).

Furthermore, signaling theory suggests that sponsorship influences brand trust (Inman 1995; Kirmani and Rao 2000; Kirmani and Wright 1989; Spence 1973). That is, extrinsic marketing cues influence consumers’ brand expectations. Consumers encode the information provided by marketers to form their beliefs about the company or brand, which then evoke a tentative inference that enables consumers to move beyond the information provided. In a sponsorship context, event-specific factors such as the size of the event may offer cues that influence consumers’ beliefs about sponsors. For example, Clark et al. (2002) suggest that a sponsorship effectively signals manufacturing ability and financial stability; companies seemingly should not be able to sustain a high-level sponsorship deal unless they have appropriate financial resources, earned through market success. Market success in turn might signal quality, brand trustworthiness, and the presence of satisfied consumers. Therefore, sponsorship may be a good signal of the ability of the brand to meet its obligations, especially when the sponsorship is consistent with what consumers expect of the firm (i.e., high fit between the event and the brand). The brand’s perceived ability to meet its obligations then is linked to perceptions of the brand’s competence and benevolence.

Sponsor-event fit also implies that spectators or participants believe a brand would be likely to sponsor the event, which should send a signal of the brand’s integrity, defined as the brand’s honesty and adherence to a set of moral principles (Mayer et al. 1995). Sponsoring an incongruent event instead might signal that the brand is pursuing a hidden agenda or is not honest in its support of the event or its expression of core values. Low sponsor-event fit thus could represent ambiguous information that creates confusion among consumers and limits their commitment to or trust in the brand (Dhar 1997). Moreover, this unexpected information could decrease perceptions of brand predictability, an important predictor of trust (Hurley 2006; Vanhonacker 2007).

  1. H3:

    High perceived fit between the brand and the sponsored event relates positively to brand trust.

Consumers’ self-congruity with the sponsored event

Consumers often purchase goods to express their identity (e.g., Aaker 1996; Aaker 1997; Malhotra 1981, 1988) and thus might evaluate brands or events according to the match between their symbolic attributes and the consumers’ own self-concept. This matching process is referred to as self-congruity (Sirgy 1985). Self-congruity plays an important role in both pre- and post-purchase behaviors (Johar and Sirgy 1991), and according to self-congruity theory, a match between the brand and the consumer’s self-image arises when a value-expressive brand triggers the consumer’s self-schema, which contains self-knowledge related to the product’s perceived image (Sirgy 1985). However some studies offer inconsistent findings regarding the self-congruity hypothesis (Barone et al. 1999; Dolich 1969; Hugues and Guerrero 1971).

Sirgy et al. (2008) have extended self-image congruence research to a corporate sponsorship context by demonstrating that the relationship between corporate sponsorship and brand loyalty is mediated by self-congruity with a sponsorship event. Also, self-congruity with a brand positively influences brand affect (Fournier 1998; Johar and Sirgy 1991; Sirgy et al. 1991). By analogy, we posit that self-congruity with an event encourages a favorable affective attitude toward the event. It follows that:

  1. H4:

    Customers’ self-congruity with a sponsored event has a positive influence on their affect toward that event.

Sponsorship also offers a prominent tool for leveraging the brand (Keller 2003; Roy and Cornwell 2003; Ruth and Simonin 2003). The process of association enables brands to borrow the equity of other entities through knowledge transfers (e.g., attitudes, thoughts, images, feelings, awareness, experience) (Keller 2003). The transfer process might reflect Heider’s (1958) balance theory, which claims that people prefer a balanced state in their lives. Dalakas and Levin (2005, p. 91) also assert that “individuals will tend to like whatever is associated with what they already like and will tend to dislike whatever is associated with what they already dislike; otherwise, there will not be balance.” This supposition also relates to the confirmation bias; that is, once positive attitudes toward a brand have been generated, consumers act to sustain those positive attitudes (McKenzie 2006; Raghunatham et al. 2006). Alternatively, Speed and Thompson (2000) use classical conditioning to explain how affect toward an event transfers to the sponsor. Pairing a brand with a positive affective stimulus (e.g., the event) invokes an evaluative conditioning procedure (De Houwer 2009), such that the brand benefits from the positive affective stimulus (Sweldens et al. 2010). Therefore, the more a consumer likes a sponsored event, the more he or she generates positive affect toward the sponsoring brand. The sponsorship that links the brand to the event in turn should facilitate the transfer of positive affect from the event to the sponsor. Therefore:

  1. H5:

    Affect toward the event relates positively to brand affect.

Determinants of brand loyalty

We have proposed thus far that sponsorship influences brand affect and brand trust through different processes. Because the goal of our model is to explain the impact of sponsorship on brand loyalty, we next relate brand trust and affect to brand loyalty, according to Chaudhuri and Holbrook’s (2001) model, in which both brand trust and brand affect positively relate to brand loyalty (purchase and attitudinal loyalty). Brand trust and affect also are well-established determinants of brand commitment, purchase intentions, and brand loyalty (e.g., Ganesan and Hess 1997; Morgan and Hunt 1994; Sirgy et al. 1991; Thomson et al. 2005). To extend knowledge about sponsorship effects beyond attitudes and encompass brand loyalty, we therefore assume parallel findings about purchase loyalty and propose:

  1. H6:

    Brand affect relates positively to brand loyalty.

  2. H7:

    Brand trust relates positively to brand loyalty.

We outline our proposed model and hypotheses in Fig. 1.

Fig. 1
figure 1

Conceptual model

Full size image

Method

We study the effects of exposure to sponsorship on brand loyalty using a repeated measures experiment, conducted with real sponsorships. The event we study is the 2008 Summer Olympics, a well-known, widely visible event with a very positive image for heterogeneous audiences (Stipp 1998). Because we are interested in the impact of the fit between the event and the brand on brand loyalty, we investigated different levels of fit and included two disparate brands in our study: one congruent with the image of the event and one perceived as less congruent. We conducted the study in January 2009, five months after the Summer Olympics. We measured all constructs (brand-event fit, self-congruity, event affect, attitude towards the sponsorship, brand affect, brand trust, and brand loyalty) using scales available in the literature. All scale items appear in Table 1.

Table 1 Measurement scales
Full size table

Pretests

We first needed to select two brand sponsors for the experiment that provoked similar levels of brand attitude, so we could control for prior brand attitude effects. Information integration theory (Anderson 1981) states that attitudes shift as people receive, evaluate, and integrate stimulus information with their existing attitudes. In turn, we looked for a pair of equally liked brands with different levels of fit, such that one brand would be perceived, on average, as more congruent with the Olympics than the other. We ran a pretest with 50 students to identify these brands, in which we first asked respondents which brands sponsored the 2008 Summer Olympics. McDonald’s, Coca-Cola, Adidas, Samsung, and Visa were recalled as sponsors by more than 30% of the study participants. In a second pretest, again with 50 students, we measured brand affect, brand loyalty, self-congruity, and fit between each of these five brands and the Olympics. Three brands (McDonald’s, Samsung, and Visa) were perceived to offer low fit with the Olympics (M < 3.0 on a five-point scale), whereas two brands (Coca-Cola and Adidas) provoked perceptions of high fit (M > 3.0). We chose two brands with similar levels of brand affect, namely, Adidas and Samsung. Consumers equally liked Adidas and Samsung (M = 3.32 and 3.17, respectively; p > .05), and the fit with the Olympics was higher for Adidas than for Samsung (M = 3.16 and 2.88, respectively).

We next conducted a test of the model using Adidas and Samsung. Their actual sponsorship of the 2008 Summer Olympics renders the experiment very realistic and enables us to use real advertising messages, run by the brands during the Olympics, to stimulate participants’ responses. The experiment therefore approximates reality in terms of the event (Summer Olympics), sponsoring brands (Samsung and Adidas), and brand communications (print ads from the sponsors) (Fig. 2).

Fig. 2
figure 2

Adidas and Samsung advertisements for the 2008 Summer Olympics

Full size image

Procedure

We used convenience sampling to recruit Adidas and Samsung consumers through e-mail invitations to approximately 850 potential participants (alumni and students from two French business schools), which asked them to participate in a short, anonymous, academic study about marketing and consumer behavior in general. We also encouraged them to forward the e-mail invitation to relatives and friends, to achieve a snowball sampling procedure. Snowball sampling is a common used technique for both qualitative and quantitative research (Frankwick et al. 1994; He and Li 2011). Although snowball sampling has the limitation of being prone to bias of sample representation, it enabled recruiting a large sample of Adidas and Samsung customers with study participants diverse in age, occupations and gender. We stopped contacting people when 600 consumers of Adidas or Samsung had indicated they were willing to participate. We defined four groups: an experimental group designed to reveal effects linked to the Adidas sponsorship, an experimental group designed to reveal Samsung sponsorship effects, a control group to compare Adidas sponsorship effects with participants not exposed to the Adidas sponsorship, and a similar control group for Samsung. We randomly assigned study participants to the two experimental groups (200 for Adidas and 200 for Samsung), leaving 200 participants for the two control groups (100 for Adidas and 100 for Samsung).

All data were collected online in three sequential steps (Table 2). First, we collected measures of the concepts that we included in the theoretical model. Second, we showed respondents advertisements for several brands, including the two experimental sponsor brands. Third, we measured once more the dependent variables (brand affect, brand trust, and brand loyalty) after participants’ exposure to the sponsors’ ads.

Table 2 Study procedure
Full size table

The measures of brand-related concepts for the four different brand conditions featured either Adidas or Samsung; we also considered event-related concepts (self-congruity, event affect, attitude toward the sponsorship, fit). In subsequent days, the experimental groups received three Internet links, one at a time, every 3 or 4 days. Each link connected to a different slideshow with four photographs from the 2008 Beijing Olympic Games (e.g., athletes performing, pictures from the opening ceremony in the national stadium) to prime their consideration of the event. None of these pictures featured sponsor logos. The photographs were followed by four print ads by official sponsors of the 2008 Beijing Olympics (Adidas, Samsung, Coca-Cola, and McDonald’s) and four unrelated print ads (Apple, Nike, Ford, and Mennen). The ads from the sponsors naturally included the 2008 Beijing Olympics logo. We used real ads but standardized the sizes of the Olympic Game logo across all ads. The order of the Olympic Games photographs and ads was random across study participants. The control group (n = 200) viewed the same slideshows, except without the ads by Adidas and Samsung.

Three days after they viewed the last slideshow, participants who had seen all three slideshows indicated again their brand affect, brand trust, and brand loyalty toward Adidas or Samsung. Of the 844 people we initially contacted, 449 completed all the steps of the procedure and provided useable responses. The final sample sizes were 300 for the experimental conditions (147 for Adidas, 153 for Samsung) and 149 for the control group. Since the study implied five successive contacts with the respondents, the online approach was considered suitable. Several studies show the reliable response quality of online surveys (Cobanoglu et al. 2001; Deutskens et al. 2006; Knapp and Kirk 2003).

The sample included 50.6% women. In terms of age, 27.8% of the participants were between 15 and 34 years, 49.7% were between 35 and 44 years, and 22.5% were older than 44 years. Furthermore, 56.1% were white-collar workers, and 19.2% were students. To confirm the comparability of the experimental groups with the control group, we conducted chi-square tests on the demographic variables (gender, occupation, and age). No significant differences emerged for any of the variables. We also tested the psychometric properties of the scales and examined the distinctiveness of the variables using confirmatory analysis with AMOS 17.0 and maximum likelihood estimation. For all scales, the factor loadings were significant (p < .01), in support of convergent validity. The Cronbach’s alphas were at least .90, which demonstrated good reliability. We also confirmed discriminant validity, because the average variance extracted exceeded the square of the correlations between constructs (Fornell and Larcker 1981), as we detail in Table 3.

Table 3 Psychometric properties of the scales
Full size table

Results

We began by confirming the difference in fit between the two brands in the main study. The respondents perceived Adidas (M = 4.21) as more congruent with the Olympics than Samsung (M = 3.84; F = 21.85, p = .000).

Direct effects of exposure to sponsorship

We checked the effect of the manipulation on sponsorship awareness and found that 70% of the participants in the experimental groups recalled spontaneously both Adidas and Samsung as Olympic sponsors after their exposure to the slideshows; only 21.5% of the respondents in the control groups did so (χ² = 95.217; p = .000). We then tested the effects of exposure to the Adidas and Samsung sponsorships on the dependent variables (brand affect, brand trust, and brand loyalty) using a repeated measures analysis of variance (ANOVA) with one within-subject factor (before and after sponsorship) and one between-subjects factor (experimental versus control group). Noting the convergent validity of the measurements, we summed the scores of the relevant items to form overall indices of brand affect, brand trust, brand loyalty, fit, self-congruity, event affect, and attitude toward sponsorship. The between-subjects factor (experimental versus control group) interacts with the impact of sponsorship on brand affect (F = 5.21, p < .05), brand trust (F = 39.95; p = .000), and brand loyalty (F = 4.07, p < .05). A repeated measures ANOVA with one within-subject factor (before and after sponsorship), conducted only with the control group, shows no significant results for brand affect, trust, or loyalty (p > .05), whereas the repeated measures ANOVA for the experimental group reveals significant increases in brand affect (F = 6.63, p < .01), brand trust (F = 128.79; p = .000), and brand loyalty (F = 24.94; p = .000). When we compare the pre-treatment measures of brand affect, trust, and loyalty, as expected, we find no significant differences between the experimental group and the control group (p > .05). Specifically, for the control and experimental groups, the mean scores are, respectively, 8.89 and 8.91 for brand affect, 7.09 and 7.37 for brand trust, and 6.69 and 6.90 for brand loyalty. After exposure to the sponsorship though, the experimental groups’ mean brand affect rises to 9.41, brand trust to 9.13, and brand loyalty to 8.01 (all differences significant at p = .01), and there are no significant differences in the control groups, as we show in Table 4.

Table 4 Direct effects of Olympic sponsorship on brand affect, brand trust, and brand loyalty
Full size table

In the Adidas and Samsung ads, only the 2008 Summer Olympics logo (and one tagline for Samsung) differentiates these sponsors’ ads from other ads used during the Olympics. Therefore, there is a risk that the sponsorship effects might be confounded with advertising effects. To test the differential effect of sponsorship versus non-sponsorship ads on brand loyalty, we conducted a between-subjects experiment, in which we randomly exposed 120 undergraduate students to the three Adidas ads used in the prior experiment (n = 30), three parallel Adidas ads without any Summer Olympics elements (n = 31), the three Samsung ads from the prior experiment (n = 27), and three parallel Samsung ads without any Summer Olympics elements (n = 32). The respondents exposed to ads with the 2008 Summer Olympics logo represent the sponsorship group, whereas we refer to respondents exposed to the ads without any elements from the Olympics as the advertising group. Using the same scale as in the main study, we measured brand loyalty before and after the respondents’ exposure to the ads; we also measured attitudes toward the ads after their exposure with four items adopted from Holbrook and Batra (1987). These scales and stimuli appeared in another questionnaire, designed for a separate research project unrelated to sponsorship and sport events. Various tasks (questions and manipulations) appeared between the two brand loyalty measures, which were identical for all groups. We also measured involvement in the event as a control variable at the end of the questionnaire. All the scales exhibit excellent reliability (Cronbach’s alpha > .94). To verify comparability between the groups, we conducted ANOVAs and compared the means for involvement, pre-treatment measures of loyalty, and attitude toward the ads. We find no significant differences across the four groups for any variables (p > .05).

Next, a repeated measures ANOVA featured one within-subject factor (before and after exposure) and one between-subjects factor (sponsorship versus advertising group). The between-subjects factor interacted with the impact of the ads on brand loyalty (F = 7.021, p < .01). Another repeated measures ANOVA with one within-subject factor (before and after ads), conducted with only the advertising group, was significant in terms of brand loyalty (F = 9.706, p < .01), but the repeated measures ANOVA with the sponsorship group revealed a greater increase (F = 37.95, p = .000). That is, our results indicate that sponsorship increases brand loyalty more than does advertising, which supports that the study tests mainly sponsorship effects as communicated through ads.

In the main experiment, we included brand as a between-subjects factor in repeated measures (Adidas/Samsung); it interacted with the sponsorship’s impact on brand affect (F = 13.36, p = .000) but not with brand trust or loyalty (p > .05). Therefore, we tested the model and hypotheses with the aggregate experimental sample, using the brand affect, brand trust, and brand loyalty measures collected after respondents’ exposure to the sponsorship ads.

Structural model testing

To estimate the relationships in the model, we used structural equation modeling with AMOS 17.0 and maximum likelihood estimation. Because previous research confirms the correlations between brand affect and brand trust (Chaudhuri and Holbrook 2001), we allowed for this correlation between residuals in the structural model.

The hypothesized, fully mediated model indicated a strong fit of the data to the model: χ²(181) = 450.877 (p = .000), confirmatory fit index [CFI] = .967, normed fit index [NFI] = .947, root mean squared error of approximation [RMSEA] = .058. The estimates of the structural path model appear in Fig. 3. The R-square values for latent endogenous variables were .30 for event affect, .16 for attitude toward sponsorship, .19 for brand affect, .05 for brand trust, and .23 for brand loyalty.

Fig. 3
figure 3

Model and path estimates

Full size image

Hypotheses testing

The high fit between the event and the brand indicates that respondents have favorable attitudes toward the sponsorship (β = .38; p < .001), in support of H1. Attitude toward the sponsorship has a positive influence on brand affect (β = .28; p < .001), in support of H2. We also find support for H3, because fit has a positive impact on brand trust (β = .17; p < .001). Because the more positive attitudes toward Adidas ads than toward Samsung ads also could explain the positive effect of fit (i.e., Adidas is perceived as more congruent with the Olympics), we measured attitudes toward the ads for both brands in a study on a sample of 57 undergraduate students. Results indicate no significant difference in their attitudes toward the ads (p > .05) enabling to rule out this possible explanation of results found for H1 and H3.

In H4, we posit that high customer self-congruity with the sponsored event exerts a positive influence on event affect; in support of our prediction, this path is positive and significant (β = .48; p < .001). Also in support of H5, event affect has a positive impact on brand affect (β = .31; p < .001). Finally, we find support for H6 and H7, because brand affect (β = .48; p < .001) and brand trust (β = .22; p < .01) positively relate to brand loyalty.

We also tested the proposed structural model and the hypotheses for each brand separately. For both Adidas and Samsung, the overall fit statistics demonstrate acceptable model fit (χ²(181) = 315.811, p = .000, CFI = .966, NFI = .925 RMSEA = .058; and χ²(181) = 411.332, p = .000, CFI = .950, NFI = .914, RMSEA = .075, respectively). All the hypothesized paths remain significant and in the hypothesized directions. However, the hypothesized model provides a better fit with the Adidas data, because Adidas is perceived as more congruent with the Olympics.

Studies that rely on structural equation modeling should test alternative models to confirm support for a proposed model (e.g., Bollen and Long 1992; Thompson 2000). Therefore, we tested a partially mediated model (i.e., with paths from each initial variable to each outcome variable) and compared it with the baseline model in terms of chi-square differences (e.g., Bentler and Bonnet 1980). The statistics indicate a good fit of the partial mediation model with the data, with χ²(176) = 414.327 (p = .000), CFI = .971, NFI = .951, and RMSEA = .055. The partially mediated model fits the data better (Δχ²[5] = 36.53, p < .001; see Table 5). Furthermore, the hypothesized paths remain significant and positive. The added direct paths from self-congruity to brand affect and from attitude toward sponsorship to brand loyalty are significant. We discuss these direct effects further in the “Discussion” section.

Table 5 Mediation tests
Full size table

Common method variance

Examining causal relationships with data from the same source may produce biased results, leaving empirical tests of the hypotheses potentially meaningless. Podsakoff et al. (2003) warn that artificial covariances between theoretically related variables may result when the same respondent provides data for both the predictor variables and the criterion variables. Accordingly, we followed their recommendations to address this issue. First, we separated the predictor and criterion variables sections in the survey questionnaires, ensured respondent confidentiality, and used a two-wave data collection procedure that separated the measures of the exogenous variables (fit, attitude toward sponsorship, self-congruity, event affect) from those of the endogenous variables (brand affect, brand trust, brand loyalty). Second, because we measured fit, attitude toward sponsorship, self-congruity, and event affect with self-reports provided at Time 1, then measured brand affect, brand trust, and brand loyalty with self-reports provided at Time 2, we investigated a single-factor model with all the measures as indicators (Harman’s single-factor test). This test reveals a very poor fit to the data (χ²(189) = 6093.932, p = .000, CFI = .283, NFI = .278, RMSEA = .264). Third, we tested the hypothesized model with an additional latent common method factor (LCMF) on which every item in the model was allowed to load (in addition to loading on its respective construct). This model results in a better fit (χ²(160) = 370.663, p = .000), though none of the standardized path coefficients between the latent factors is affected significantly. Ten of the 21 loadings on the LCMF are significant, but the indicator loadings on their theoretical factors all remain significant. In summary, we find some evidence of common method variance bias but not enough to explain the relationships observed (Piercy et al. 2006).

Discussion

Key findings

From a managerial standpoint, our findings indicate that sponsorship has a positive influence on brand trust and brand loyalty; these are key empirical findings, considering the importance of marketing budgets devoted to sponsorship. Our research reveals changes in brand trust and brand loyalty before and after event sponsorship for two major brands in different categories (Adidas and Samsung). Furthermore, the model offers a better understanding of the process by which brand loyalty is influenced by sponsorship activities. We confirm the important roles of self-congruity (between the consumer and the event) and of the fit between the event and the brand, which influence affect toward the event and attitude toward the sponsorship by the brand, respectively. Self-congruity and event–brand fit are both managerially relevant and actionable, and the relationships uncovered highlight the managerial importance of choosing a sponsorship event that is favored by or congruent with target consumers, as well as perceived by the target consumers as congruent with the brand’s own image or personality. Furthermore, affect toward the event and attitude toward the sponsorship are easily measurable concepts from an empirical standpoint, even before the brand chooses from among its different sponsorship alternatives. The impact of these concepts on brand affect, brand trust, and ultimately brand loyalty indicates that simple market research studies to measure these variables among target consumers will be beneficial for sponsorship decision making and budget allocation. The results also suggest that the sponsorship effects may be short-lived since the control group potentially exposed to the sponsorship during the Olympics shows significant lower brand loyalty toward the actual sponsors few months after the 2008 Summer Olympics. In order to increase brand loyalty, it may be more efficient to sponsor one event that happens regularly compared to one event that occurs every 4 years. More longitudinal research is needed to test this assumption.

From an academic standpoint, our model is one of the first to elaborate on brand sponsorships and attempt to explain brand loyalty. It explains 23% of the variance in brand loyalty; Sirgy et al. (2008) were able to explain less than 1% of that variance. We thus highlight the importance of sponsorship activities for brands and brand management. Our model extends previous results that demonstrate the effect of sponsorship on brand awareness, brand image, and attitude toward the brand (e.g., Dean 1999; Lardinoit and Derbaix 2001). In particular, we demonstrate the effect of sponsorship on brand loyalty, which provides a key competitive advantage, and is an ongoing managerial challenge (Dick and Basu 1994; Fournier and Yao 1997). However, we did not measure actual purchase behavior, which is difficult to do when dealing with durable goods and conducting an experiment. The loyalty measure we employed instead is based on declarations of future purchase behavior or repurchase intent. A confirmation of the relationships uncovered with actual behavioral data thus would be helpful.

Building on our preliminary results, we modeled the process by which self-congruity with the event may affect brand loyalty and proposed another process through which the fit between the event and the brand may affect brand loyalty, in line with theories about the importance of such fit in sponsorship literature (e.g., Speed and Thompson 2000). Our model decomposed the impact of sponsorship on brand loyalty into two main routes, each with mediating variables. The first route refers to the impact of self-congruity with the event on affect felt toward the event, which in turn influences the consumer’s affect toward the brand and finally drives brand loyalty through an affective route. This process relies on consumers’ judgments of the event itself (self-congruity, event affect), which then transfer to the brand (brand affect, brand loyalty). A second route encompasses the perceived fit between the event and the brand, which positively affects brand trust and attitude toward the sponsorship of the event by the brand, then influences brand affect and ultimately brand loyalty. This process involves consumers’ judgments of whether the brand appears related to the event. Consideration of the brand as congruent or incongruent with a particular event (event–brand fit, attitude toward the sponsorship by the brand) influences consumer–brand relationship constructs (brand trust and brand affect) (Fournier 1998), which in turn impact brand loyalty.

The empirical evidence of the significance of all our hypothesized relationships clarifies the mechanisms by which sponsorship drives brand loyalty. Some hypothesized mediations are only partial, which should be expected in explanations of complex constructs such as loyalty. Further research might expand on the meanings of the unexpected direct effects and identify omitted mediators with the same sign as the direct effect. For example, attitude toward the sponsorship might influence brand loyalty through brand attributes (e.g., perceived quality, prominence) that move beyond brand affect. Self-congruity with the event also could influence brand affect directly, due to conformity effects or imitation of social norms. The more a person thinks that other members of his or her social group appreciate the purchase of a sponsor’s product, the more that consumer tends to buy the product (Madrigal 2000). When an event fits a social ideal, the consumer thus may exhibit greater liking of the sponsor.

The important role of brand affect, as determined both by affect toward the event and attitude toward the sponsorship, is another key finding of this study, in that brand affect appears to be an important mediator between event-related concepts and brand loyalty. The process of pairing a conditioned stimulus (i.e., brand) with an unconditioned stimulus that evokes favorable affective responses (i.e., event) entails evaluative conditioning (Sweldens et al. 2010). Attitude toward the sponsorship positively influences brand affect through an indirect affective response, which demands conscious awareness of the brand–event relation. Yet a direct affective transfer between the event and the brand also is possible, without establishing a brand–event relation. Sponsorships, by presenting the brand and event simultaneously, could transfer affect directly to the brand. The positive affective disposition toward the event then could spill over to the brand, as in the case of co-branding (Simonin and Ruth 1998).

Limitations and future research

Our experimental methodology, which features measurements performed before exposure to the sponsorship, multiple exposures to the sponsorship over a week-long period, and then additional measurements of the dependent variables (brand affect, brand trust, and brand loyalty) 4 days later, reflects our effort to minimize problems linked to common method variance or priming with possible immediate memory effects. However, the methodology is more quasi-experimental rather than experimental, in that we use real stimuli (Adidas and Samsung ads) in the context of a real event (Olympics). Therefore, some differences across groups may emerge on variables that we do not control (e.g., attitude toward the Olympics, previous exposure to ads). These are the costs of realism, but realism is key for sponsorship research.

Though our overall model fit is satisfactory, as are the intensities of the relationships between the constructs, the role of brand trust requires further research. Sponsorship has a strong effect on brand trust (F = 81.06, p = .000), and the path coefficients to and from brand trust are significant, yet the model still offers little explanation of brand trust in itself. This limitation could reflect the nature of the two brands we used (Adidas and Samsung), measurement issues, or the way we modeled the role of brand trust. Additional research should deal with this issue, as well as adding other predictors that might increase the explained variance of brand loyalty. Adding predictors also could have an impact on the structural relationship we estimate. Furthermore, this research requires replication with other events that offer more variability in terms of prominence and attitudes and in different settings than the French market. The intensity of consumer exposure and the attitude toward sponsorship activities in general and toward Olympics’ sponsorship in particular may vary across countries and cultures (Marshall and Cook 1992). Similarly, the quasi-experimental design implemented here and the use of a convenience sample through a snowball sampling procedure are limitations that call for replications in other conditions.

Although our research purpose is not comparative, our results indicate that sponsorship increases brand loyalty to a higher degree than does advertising. Very little research has attempted such a comparative assessment, though Olson and Thjomoe (2009) empirically offer some rules for converting sponsorship exposure time into television advertising equivalent values. This notion suggests an interesting direction for further research. Finally, dealing with moderating variables of the impact of sponsorship on brand loyalty is of interest. For example, marketing mix variables (as a premium price) and individual consumers’ experience with the product may moderate the effects found in this study.