INTRODUCTION

Country-of-origin (CO) effects refer to the influence that product foreignness has on consumer choice behavior (Samiee, 2011), and is widely investigated in international business (IB) literature (e.g., Samiee, Leonidou, Aykol, Stottinger, & Christodoulides, 2016). However, there is relatively little a firm can do about its CO. In any given emerging market (EM), it will be viewed as either foreign or local (even if some firms use brand names or other mechanisms to mislead consumers regarding CO; Balabanis & Diamantapoulos, 2011). In contrast, firms can precisely define both the customer segments they target (e.g., firms can target consumers who are relatively more ethnocentric) and the marketing mix they apply. In interviews with various senior marketing managers working in EMs (see “Appendix 1”),1 we learned that celebrity endorsers represent an especially relevant aspect of the marketing mix. Accordingly, in this paper, we investigate whether firms should use local or foreign celebrities as endorsers, contingent on the consumer segment they target.

There are two reasons why research on celebrities is very relevant to IB. First, although celebrity endorsements originated in the developed world, use of celebrity endorsements is relatively greater in EMs. As Bergkvist and Zhou (2016: 644) note, a “comparison of 6359 TV ads from 25 countries found celebrities were (used) in less than 15% of the ads in a number of European countries and the US, while the share in Asia ranged from 25% (China) to 61% (South Korea).” Consistent with this, our senior manager interviewees affirmed that many of their advertising campaigns in emerging markets involve celebrities. Second, in many EMs, the choice of celebrity endorser is a CEO-level decision, not only because of the upside in sales due to celebrity-led ad campaigns (e.g., “the importance of the Shah Rukh Khan’s endorsement of Hyundai – in its early days, when Hyundai was relatively less known in India – cannot be overstated” – senior auto executive in India), but also because of the potential downside should a celebrity become embroiled in a scandal (e.g., “Lionel Messi’s tax evasion scandal created … complications for (India’s) Tata Motors” – (another) senior auto executive in India). As such, in EMs, the choice of a celebrity endorser is a CEO-level decision and so is (1) different from other promotion decisions, and (2) worthy of research attention.

Prior research on CO has largely focused on how EM consumers perceive brands and products; however, there is relatively little research on how EM consumers perceive celebrities. Fit between a celebrity’s and (target) consumers’ demographics likely increases consumers’ evaluations (Roy, Guha & Biswas, 2015), such that – for example – consumers exhibit “more positive responses toward ads featuring an endorser with a … similar ethnicity” (Ryu, Park & Feick, 2006: 506). Building from this, we propose that “consumer CO fit” (i.e., match between the celebrity CO and consumers’ CO) increases evaluations; thus, for example, using a local celebrity spokesperson increases (local) consumers’ evaluations.

Next, we consider two factors that may moderate the effects proposed above. First, we examine consumer ethnocentrism, which is the “bias against out-groups to explain customers’ preferences for their home-country’s products” (Dogan & Yaprak, 2017: 1502; also see Shimp & Sharma, 1987). Noting that “ethnocentric customers take pride in their country’s … symbols and culture” (Steenkamp, Batra & Alden, 2003: 56), we propose that consumer ethnocentrism moderates the effects of CO-driven fit, with more (vs. less) ethnocentric customers having higher evaluations when firms use a local celebrity endorser.

Second, we examine consumer cosmopolitanism. Zeugner-Roth, Zabkar, and Diamantapoulos (2015: 30–32) indicate that “Riefler, Diamantapoulos, and Siguaw (2012) … empirically demonstrate that consumer cosmopolitanism … predicts … preferences for foreign products”; furthermore, Dogan and Yaprak (2017: 1506) state that “consumer cosmopolitanism had a … positive effect on willingness to buy foreign products”. Extending this logic to consumers’ preferences for foreign celebrity endorsers, we propose that levels of consumer cosmopolitanism moderates the effects of CO-driven fit, with those more cosmopolitan having higher evaluations when firms choose to use a foreign celebrity endorser. In effect, the moderating effects of consumer ethnocentrism and consumer cosmopolitanism should work in opposite ways, consistent with consumer ethnocentrism levels and consumer cosmopolitanism levels generally being negatively correlated (Dogan & Yaprak, 2017).

The totality of the above discussion suggests that CO effects for brand evaluations and celebrity evaluations may well be similar, as both are impacted by factors like ethnocentrism and cosmopolitanism. Nevertheless, there is one key difference between how consumers evaluate brands versus how consumers evaluate celebrities, in that while consumers may evaluate brands on a standalone basis, consumers evaluate celebrities based on both (1) the “fit” between the celebrity and the consumer and (2) the “fit” between the celebrity and the brand being endorsed. Using the latter conceptualization of fit, some of the data in Roy and Bagdare (2015) indicate that a match between celebrity CO and brand CO (hereafter termed as “brand CO fit”) increases consumers’ evaluations. This suggests that those selling foreign (local) brands should choose to use a foreign (local) celebrity. In the case of foreign brands, however, this point directly conflicts with the propositions in the prior paragraph. Consider, for example, those selling foreign brands – should these firms use a foreign celebrity spokesperson, as this paragraph suggests, or should they use a local celebrity spokesperson, as the prior paragraphs prescribe. Furthermore, when making the trade-off above, what is the role (if any) of consumer ethnocentrism levels and consumer cosmopolitanism levels? Drawing from information integration theory (IIT, Anderson, 1971), this paper examines the questions (above). Table 1 overviews prior research and shows the specific research gaps addressed by this paper.

Table 1 Gaps addressed by current research
Full size table

We propose three sets of contributions. First, we contribute to the work on CO, specifically as applicable to the important domain of celebrity endorsements. This paper presents a framework to evaluate the impact of celebrity endorsements, and what makes this framework very relevant to IB is that it is based on CO congruency. Our empirics show how consumers’ evaluations are contingent on (1) consumer CO fit, based on CO-relevant factors like consumer ethnocentrism and consumer cosmopolitanism, and (2) brand CO fit; furthermore, we identify contingencies whereby one or the other fit-type dominates.

Second, we contribute to work on consumer ethnocentrism and cross-cultural branding. This paper develops a framework for evaluating effectiveness of celebrity endorsements, contingent on segmentation involving differences in consumer ethnocentrism levels. Also, prior work conceptualizes consumer ethnocentrism as a negative bias against foreign-made brands. In this paper, we show that an endorsement from a local celebrity may (partially) counter this bias. Finally, (some) prior research on consumer cosmopolitanism has suggested that those more cosmopolitan prefer foreign goods and symbols (e.g., celebrities) and seek to be a part of a global consumer culture. We show that this desire is not universal, but contingent on the CO of the brand being considered, and we elaborate on this point later in this paper.

THEORETICAL FRAMEWORK

CO effects have been widely investigated by IB researchers. CO-related research focuses on topics like consumer ethnocentrism, consumer cosmopolitanism, and consumer animosity (Diamantopoulos, Florack, Halkias, & Palcu, 2017a; Grinstein & Riefler, 2015; Klein, 2002; Riefler, et al., 2012; Sharma, 2011, 2015; Steenkamp et al., 2003; Verlegh, 2007) and even extends CO-like effects to beyond a single country, to multi-country entities such as “EU” (Diamantapoulos, Herz, & Koschate-Fischer, 2017b). Beyond consumer choice settings, CO effects also influence how firms structure their marketing mixes and supply chains. In this realm, CO factors inform research into localization versus globalization, (global) branding, celebrity spokespeople, and offshoring (Buckley, Doh, & Benischke, 2017; Ghemawat, 2003; Mishra, Roy, & Bailey, 2015; Roy et al., 2015; Steenkamp et al., 2003). Furthermore, studies show that consumer ethnocentrism and consumer cosmopolitanism drive segmentation in international markets (Balabanis & Diamantapoulos, 2016; Riefler et al., 2012). To develop a suitable framework for this paper, we outline relevant prior work related to consumers’ responses to the brand’s CO, the celebrity’s CO, and the fit between brand CO and celebrity CO.

Consumer Responses to Brand CO

A brand reflects a collection of images and ideas (Riley, 2016), including CO information that consumers might either use to infer product quality (Han & Terpstra, 1988; Sharma, 2011), or cue emotional meanings linked to national identity (Verlegh & Steenkamp, 1999). Consumers thus may react negatively or positively to CO information. Consumer ethnocentrism triggers negative reactions, in the form of a bias against foreign brands (Sharma, 2011, 2015; Shimp & Sharma, 1987), foreign symbols and culture (Steenkamp et al. 2003). Ethnocentric consumers are willing to sacrifice “actual, objective gains to avoid outgroup contact” (Steenkamp et al. 2003: 57). Sharma (2015) identifies three types of responses triggered by consumer ethnocentrism: (1) affective, such as suspicion or disdain for outgroups; (2) cognitive, which implies a positive bias toward domestic products; and (3) behavioral, which reflects a preference for domestic goods in terms of willingness to try, desire to spread positive word of mouth, and purchases. Taking a broader view, Siamagka and Balabanis (2015) incorporate both deliberative and automatic responses to foreign brands. Thus, consumer ethnocentrism may moderate the effects of CO fit, such that those more ethnocentric have lower evaluations when firms use a foreign celebrity endorser.

Other factors may also trigger negative responses. Home-country bias correlates with consumer ethnocentrism (Verlegh, 2007), but whereas ethnocentrism is rooted in economic considerations, home-country bias stems from a desire for a positive self-identity. Also, consumer animosity stems from residual resentment related to prior political and economic conflicts (Klein, 2002). While ethnocentric consumers generally are biased against products from any foreign country, consumers exhibiting animosity direct their bias against a specific foreign country (Riefler & Diamantopoulos, 2007).

In contrast, some consumers may express positive reactions to foreign brands. For example, a sense of nostalgia can trigger positive perceptions of foreign brands in historically connected markets (Gineikiene & Diamantapoulos, 2017). In instances wherein consumers make deliberative decisions, some countries may be associated with perceptions of ‘competence’, in turn increasing evaluations (Diamantapoulos et al., 2017). Most important though, cosmopolitan consumers generally are more willing to engage with cultures and brands with foreign COs (Cleveland, Laroche, Takahashi, & Erdogan, 2014; Dogan & Yaprak, 2017; Riefler et al., 2012; Zeugner-Roth et al., 2015). Thus, consumer cosmopolitanism may (also) moderate the effects of CO fit, such that more cosmopolitan customers offer better evaluations when firms use a foreign celebrity endorser. In essence, we predict that both consumer ethnocentrism and consumer cosmopolitanism moderate the effects of CO fit, albeit in opposite ways, and this is consistent with the negative correlations found between consumer ethnocentrism and consumer cosmopolitanism (Dogan & Yaprak, 2017; Siamagka & Balabanis, 2015).

As an aside, we acknowledge other conceptualizations of consumer cosmopolitanism. Researchers have proposed that cosmopolitan consumers have strong local roots and connections and do not necessarily seek foreign products, but rather prefer authentic products and experiences (Cannon & Yaprak, 2002, 2012); we address this point later in this paper.

Consumer Responses to Celebrity CO

Celebrity endorsements are widespread and international (Bergkvist & Zhou, 2016), with powerful economic impact, such that they represent an important consideration for both practitioners and IB researchers (Agrawal & Kamakura, 1995; Ding, Molchanov, & Stork, 2011; La-Ferle & Choi, 2005; Winterich, Gangwar, & Grewal, 2018). Endorsements have greater impact if consumers perceive the endorsed product as “for me” (Deshpande & Stayman, 1994). For example, when a celebrity endorser’s age matches consumers’ chronological age, consumers express stronger perceptions that the advertisement is “for me” (Chang, 2008) and also have higher evaluations (Roy et al., 2015). Among Singaporean respondents, advertisements with Japanese (vs. U.S.) celebrity endorsers improve consumers’ brand attitudes; Singaporeans likely perceive Japanese endorsers as being of similar ethnicity, so may have inferred “that endorsers with similar ethnicity are likely to have similar … preferences” (Ryu et al., 2006: 491). The CO concept is similar to ethnicity, albeit more focused (e.g., Singaporeans and Japanese match in terms of ethnicity, but not on CO). Thus, consumer CO fit, or a match between the celebrity’s CO and consumers’ CO, may heighten consumers’ evaluations (evidenced by improved brand attitudes and higher purchase intentions).

Hypothesis 1:

Higher levels of fit between the celebrity’s CO and the consumer’s CO heighten consumers’ evaluations.

Some may point out that consumer CO fit is implied by prior work on ethnocentrism. For example, Steenkamp et al. (2003) indicated that ethnocentric consumers express pride in symbols that represent their country, so to the extent that a celebrity represents a symbol, research into ethnocentrism may foreshadow the effects that we predict. However, we consider this account of how consumers evaluate celebrity endorsements incomplete, as we discuss next.

Consumer Responses to Consumer CO Fit and Brand CO Fit

When evaluating brands, consumers may well evaluate brands on a standalone basis, and so – for example – more ethnocentric consumers may dislike foreign brands (or, prefer local brands). However, when evaluating ads with celebrity endorsers, consumers evaluate celebrities not only on a standalone basis (based on the “fit” between the celebrity endorser and the consumer), but also based on the “fit” between the celebrity endorser and the brand. We discuss this latter version of fit (i.e., brand CO fit) below.

Perceived congruency between a celebrity endorser and a brand (i.e., high brand CO fit) leads to more favorable evaluations (Kamins & Gupta, 1994). For example, when considering an attractiveness-related product, consumers express higher purchase intentions when the product is endorsed by a more (vs. less) attractive celebrity (Kamins, 1990). Hence, on dimensions other than CO, fit increases evaluations. Next, we consider CO-specific research. Most CO-specific research has focused on how consumers react to products, conditional on CO. In contrast, what has remained relatively unexplored is (i) how consumers react to celebrities, conditional on CO (we covered this point earlier, in the set-up to Hypothesis 1), and (ii) how consumers react to celebrity-brand pairings, conditional on CO; this brand CO fit issue is examined next.

Endorsements by a global (vs. local) celebrity is effective for all global brands, irrespective of CO (Roy & Bagdare, 2015), but a closer review of the data reveals that evaluations were higher when the celebrity CO exactly matched the brand CO; this suggests that high brand CO fit leads to higher evaluations. This point comports with other work in IB, whereby advertising brands in ways congruent with their home country (on dimensions like competence vs. warmth) increases evaluations (Magnusson, Westjohn & Sirianni, 2018). We propose:

Hypothesis 2:

Higher levels of fit between the celebrity’s CO and the brand’s CO heighten consumers’ evaluations.

For local brands, Hypotheses 1 and 2 converge, suggesting higher evaluations when local celebrity endorsers are used. For foreign brands, however, Hypotheses 1 and 2 diverge. Specifically, Hypothesis 1 (focusing on consumer CO fit) indicates higher evaluations when local celebrity endorsers are used, but Hypothesis 2 (focusing on brand CO fit) indicates higher evaluations when foreign celebrities are used. We examine this divergence in the next section.

Information Integration Theory (IIT)

To determine how consumers integrate the divergent influences of consumer CO fit and brand CO fit, we turn to information integration theory (IIT; Anderson, 1971), which proposes a “cognitive algebra” that consumers apply to integrate various information cues, whether they are mutually reinforcing or contradictory. IIT goes beyond balance theory and congruity theory (Anderson, 1971) to assert that each information cue has distinct valence and weight, so consumers combine cues using various mechanisms, e.g., adding, averaging (Anderson, 1971; Ebbesen & Konecni, 1975; Leon, Oden, & Anderson, 1973). A key point is that when integrating contradictory information cues, the negatively valenced information cue gets relatively more weight during evaluations (Anderson, 1965; Miyazaki, Grewal & Goodstein, 2005). IIT has been used in the IB literature to understand how consumers integrate CO information into overall product evaluations (Hastak & Tong, 1991); it has also been used to understand how pioneering products retain advantage over later-entry products (Kardes & Kalyanaram, 1992), and how decision-makers (e.g., jurors) integrate positive and negative information cues (Kaplan & Kemmerick, 1974).

We note two insights. First, while factors such as ethnocentrism may affect assessments of consumer CO fit, they may not influence brand CO fit assessments (there is no linkage to factors like ethnocentrism, in any of the literatures that set up brand CO fit concerns). Second, considering only consumer CO fit, we note that negative reactions towards foreign celebrities should be more extreme in case of those more ethnocentric. Building from IIT, this latter point suggests that consumer CO fit should have relatively more impact on evaluations in the case of those more (vs. less) ethnocentric. We now build from these two insights, examining first the case relating to foreign brands.

Foreign brands. If a foreign celebrity endorser is used, then this implies (1) high brand CO fit (Hypothesis 2), but (2) low consumer CO fit (Hypothesis 1). On the other hand, if a local celebrity endorser is used, then this implies the exact reverse, i.e., low brand CO fit but high consumer CO fit. To better understand how consumers make (relative) tradeoffs between consumer CO fit and brand CO fit, we first consider the case of those more ethnocentric, and then the case of those less ethnocentric.

Consider first the case of those more ethnocentric. When a foreign celebrity endorser is used, then this implies (1) high brand CO fit, a positively valenced information cue, but (2) low consumer CO fit, a negatively valenced cue related to ethnocentrism, and so for those more ethnocentric this cue should be relatively highly weighted during evaluations. Thus, in the case of those more ethnocentric, use of a foreign celebrity endorser is more likely to lead to lower evaluations, with these consumers being more concerned about consumer CO fit. Consider next those less ethnocentric. When a foreign celebrity endorser is used, then this implies (1) high brand CO fit, but (2) low consumer CO fit. However, amongst those less ethnocentric, this negatively valenced consumer CO fit cue – which is related to ethnocentrism – is relatively less weighted during evaluations. Thus, use of a foreign celebrity endorser is less likely to lead to lower evaluations, with less-ethnocentric consumers being relatively less motivated by consumer CO fit concerns.

Local brands. If a foreign celebrity endorser is used, then this implies low consumer CO fit (per Hypothesis 1) and low brand CO fit (per Hypothesis 2). Thus, in all cases, independent of ethnocentrism levels, evaluations are likely to be lower if a foreign celebrity is used.

Hypothesis 3:

(a) For foreign brands, consumer ethnocentrism levels moderate the relative impact of brand CO fit versus consumer CO fit. Specifically, more ethnocentric (vs. less ethnocentric) consumers are more motivated by consumer CO fit concerns, and so for those more ethnocentric using a foreign celebrity endorser should lead to relatively lower evaluations. (b) For local brands, brand CO fit and consumer CO fit concerns align, and thus – irrespective of consumers’ ethnocentrism levels – using a foreign celebrity endorser leads to relatively lower evaluations.

OVERVIEW OF STUDIES

To test our hypotheses, we rely on experimental methods, and we run multiple studies to ensure robustness (Grinstein & Riefler, 2015; Zellmer-Bruhn, Caligiuri, & Thomas, 2016). Each study builds on the prior study, simultaneously (1) replicating prior study results, and (2) examining pending/new issues. In Studies 1A and 1B, we test Hypotheses 1 and 2. Across Studies 2 and 3, we test Hypothesis 3, as well as examine some pending questions from Studies 1A and 1B. Considering that prior literature has found a negative correlation between consumer ethnocentrism and cosmopolitanism, in Study 3 we also examine the role of cosmopolitanism.

While recruiting respondents, we followed similar procedures in all studies. That is, following Bello, Leung, Radebaugh, Tung, and Van Witteloostuijn (2009), we obtained responses from adults recruited by a market research agency, who came to a central facility to complete a paper-and-pencil survey. “Appendix 2” contains detailed sample information. Noting that respondents may not always identify brand CO correctly (Samiee, Shimp, & Sharma, 2005), at the end of every study we asked manipulation check questions where respondents indicated the brand CO and the celebrity CO (note – all respondents, in all studies, correctly identified CO).

The (Key) Independent Variables

All studies involve some version of 2 (brand CO: local vs. foreign) × 2 (celebrity CO: local vs. foreign) between subjects study design. The local brand and the local celebrity have – by definition – the same CO as the (local) consumer. A nuanced examination of the data in Roy and Bagdare (2015) suggests that evaluations are highest when the foreign brand and the foreign celebrity have the exact same CO. We ran a pretest to re-examine this latter point (not explicitly examined and discussed in Roy & Bagdare, 2015).

The pretest involved a 2 (brand CO: India vs. USA) × 3 (celebrity CO: India vs. USA vs. UK) between-subjects design, using Indian participants. Full details are available from the authors; here we present a brief overview. Participants evaluated a man’s watch (Titan from India vs. Fossil from USA) endorsed by a celebrity (Akshay Kumar from India vs. Tom Cruise from the USA vs. Daniel Craig from the UK). A pre-test showed that, in the absence of any specific brand information, Craig (the face of the James Bond franchise) was perceived more positively than Cruise, although – as expected – Kumar was perceived more positively than both Craig and Cruise. When evaluating the local watch brand Titan, participants’ evaluations were highest when Kumar was the celebrity endorser, consistent with both consumer CO fit and brand CO fit concerns. In contrast, when evaluating the foreign (US) watch brand Fossil, participants’ evaluations were highest when Cruise was the celebrity endorser, consistent with the theme of brand CO fit (celebrity CO matched brand CO). Specifically, despite participants generally preferring Craig over Cruise, (presumably) brand CO fit concerns overrode the preference participants had for Craig. Thus, the optimal foreign celebrity is one that shares CO with the foreign brand.

Study Contexts: India and China

We conducted studies in India and China, for four reasons. First, by 2025, China and India are projected to be the second and third largest consumer economies globally (Rapoza, 2017). Second, celebrities from India and China (e.g., Aishwarya Rai, a leading Indian actress and Miss World 1994; Li Na, a leading Chinese tennis player, ranked second in the world in 2014) have emerged as major endorsers, particularly in their home countries. These local celebrities not only endorse major local brands, but also endorse foreign brands like Nike and Coca-Cola, competing strongly with U.S. (and other foreign) celebrities. Third, local Indian brands like Titan (watches) and Mahindra (tractors), and local Chinese brands like Haier (white goods) and Tsingtao (beer), have substantial resources and can credibly seek endorsements from either foreign celebrities or local celebrities, as they deem fit. Fourth, prior IB research on advertising and branding has also focused on India and China (e.g., Heinberg, Ozkaya & Taube, 2017).

STUDY 1

Method

Both Studies 1A and 1B use a 2 (brand CO: local vs. foreign) × 2 (celebrity CO: local vs. foreign) × 2 (country: India vs. China) between-subjects design (“Appendix 3”, Panel A). In Study 1A, we examined men’s watches, contrasting Fossil (foreign brand) with Temporis (local brand, China) or Titan (local brand, India). In Study 1B, we considered cellphones and contrasted Apple (foreign brand) with Haier (local brand, China) or Micromax (local brand, India). Varying both the product category and the familiarity of the foreign brand (Apple is better known than Fossil) enhances the robustness of our results.

To choose appropriate celebrity endorsers, we ran pretests in China and in India. In each pretest, 30 persons similar in profile to the main study respondents listed their favorite local and their favorite foreign (U.S.) male celebrity endorsers. The actors Jackie Chan (China) and Akshay Kumar (India) emerged as the local favorites (21/30 and 23/30 votes, respectively); the U.S. actor Tom Cruise was the favorite foreign celebrity endorser (19/30 votes in China and 17/30 votes in India). Thus, we used Cruise as the foreign celebrity spokesperson and Chan and Kumar as the local celebrity spokespersons in the respective countries. All these celebrity choices had face validity; in 2016 and 2017, Forbes magazine included all three of them in its lists of the top 10 actors in the world, in terms of endorsement earnings. Consistent with their status as local celebrities, both Kumar and Chan obtained most of their earnings from their domestic markets (e.g., Jackie Chan “makes most of his money on mainland (China) movies you’ve probably never heard of”; Forbes, 2018).

In Studies 1A and 1B, respondents read that they were buying a men’s watch (or, cellphone) for a close friend. They saw an advertisement for the local (or, foreign) brand, which featured a local (or, foreign) celebrity endorser (for stimuli exemplars, see “Appendix 4”). The respondents then indicated their purchase intentions (PI, primary dependent variable) and attitude toward the brand (AB, widely used in branding research). Finally, respondents provided demographic information and responded to the manipulation check questions. The PI and AB measures, along with reliability statistics, are described in “Appendix 5”.

Study 1A Results

We ran linear regressions for both PI and AB, including the three independent variables (brand, celebrity endorser, country) and all two-way interactions, as well as the three-way interaction. Table 2 outlines the PROCESS Model 3 (Hayes 2013) coding and output, including conditional effects; we report specific p values (per Meyer, van Witteloostuijn, & Beugelsdijk, 2017). All stated effects are significant, unless otherwise indicated (exact significance levels are provided in the tables).

Table 2 Study 1A results
Full size table

As expected, the regression results for PI and AB were similar (Panels A and B (PI R2 = 0.74; AB R2 = 0.77). We noted a negative main effect of celebrity (PI b = − 3.20; AB b = − 3.39). In both India and China, evaluations were lower when the local brand was endorsed by a foreign (vs. local) celebrity endorser. This was reflected in negatively-signed conditional effects in both countries (1) PI (India − 3.20; China − 3.96), and (2) AB (India − 3.39; China − 3.32).

We also noted a positive brand × celebrity interaction effect (PI b = 4.93; AB b = 6.32). The coding indicates that this interaction effect relates to the foreign brand. For Indian respondents, evaluations were higher for the foreign brand–foreign celebrity endorser pairing than for the foreign brand-local celebrity endorser pairing (conditional effects, PI b = 1.73; AB b = 2.93). The main effect and the interaction effect presented in the two paragraphs above are consistent with Hypothesis 2, and are consistent with brand CO fit concerns.

Finally, we noted a negative three-way interaction effect (PI b = − 2.90; AB b = − 4.60), which indicated that the two-way interaction was effectively suppressed in the case of respondents from China. Respondents in China offered poorer evaluations in the foreign celebrity endorser conditions, irrespective of whether the brand was foreign (conditional effect, PI b = − 1.92; AB b = − 1.60) or local (discussed above). These three-way interaction effects were consistent with both Hypothesis 1 and consumer CO fit concerns.

Study 1B Results

Results were similar to Study 1A. Noting that results for PI and AB were very similar, here we only discuss results for PI (but present results for both PI and AB in Table 3). First, we noted R2 = 0.62. Second, we noted a significantly negative main effect of celebrity (b = − 2.64). In both countries, evaluations were lower when the local brand was endorsed by a foreign celebrity endorser (conditional effects – India − 2.64; China − 2.10). Third, the positive interaction effect of brand × celebrity (b = 4.39) indicated that in India, consumers’ evaluations were higher for the pairing of foreign brand–foreign celebrity endorser pairing than for the pairing of foreign brand-local celebrity endorser (conditional effect, b = 1.75). Thus, the results for Indian respondents (both, the main effect and the interaction effect) were more consistent with Hypothesis 2 and brand CO fit concerns. Finally, the negative three-way interaction effect (b = − 4.46) indicated that the positive two-way interaction was suppressed among Chinese respondents. Evaluations were lower when the endorser was a foreign celebrity, for both a foreign brand (conditional effect, b = − 2.17) and a local brand (discussed above). This negative three-way interaction effect was consistent with both Hypothesis 1 and consumer CO fit concerns.

Table 3 Study 1B results
Full size table

Discussion

The consistent results across Studies 1A and 1B attest to the robustness of our findings. However, two questions emerge. First, for foreign brands, why do Indian (Chinese) respondents react more positively to foreign (local) celebrity endorsers? Second, for regards local brands, evaluations are consistently lower in response to a foreign celebrity endorser, so does that imply local brands should never use foreign celebrity endorsers? We examine these questions in Studies 2 and 3.

STUDY 2

Study 1 results indicate that participants from India are more concerned with brand CO fit, such that PI is higher when a local celebrity endorses the local brand and a foreign celebrity endorses the foreign brand. However, participants from China seem more driven by consumer CO fit concerns, because PI is higher when a local celebrity endorser is used, regardless of whether the brand is local or foreign. The key point of difference emerges in relation to foreign brands: Indian (Chinese) participants prefer foreign (local) celebrity endorsers. For example, in Study 1A, respondents from India expressed relatively higher PI for Fossil endorsed by the foreign celebrity Cruise (vs. local celebrity Kumar, PI b = 1.73), whereas respondents from China indicated lower PI for Fossil when endorsed by Cruise (vs. local celebrity Chan, PI b = − 1.92). Prior research suggests that Indian consumers tend to be relatively less ethnocentric than Chinese consumers (Hsu & Nien, 2008; Pereira, Hsu, & Kundu, 2002), which might explain the Indian respondents’ enhanced evaluations of the foreign brand endorsed by a foreign celebrity. In Study 2, to test this conjecture, we specifically elicit consumer ethnocentrism (individual difference measure). Because key differences pertain solely to the foreign brand, in Study 2 we focus on foreign brands exclusively.

Method

The procedures were similar to those in Study 1A, except that we only examined the foreign brand (Fossil). In addition, we used a nine-point (cf. seven-point) PI scale, to increase scale variance.2 Finally, we explicitly measured consumer ethnocentrism (CET) with a 17-item, nine-point consumer ethnocentrism scale (CETSCALE, similar to Shimp & Sharma’s [1987). The CETSCALE offers unidimensionality (Netemeyer, Durvasula, & Lichtenstein, 1991) and has been verified with both Chinese and Indian populations (Pereira et al., 2002). Thus, Study 2 used a 2 (celebrity endorser origin: local vs. foreign) × 2 (country: China vs. India) × 2 (CET: low vs. high) between-subjects design (“Appendix 3”, Panel B).

Results

First, we tested whether we could replicate Study 1A results. We ran a linear regression for PI (R2 = 0.11), with the celebrity endorser and country as independent variables, and including an interaction term (PROCESS Model 1; see Table 4, Panel A). Replicating Study 1A results, when Cruise endorsed Fossil, (1) Indian respondents had directionally higher evaluations (b = 0.29, t = 1.91, p = 0.06), but (2) Chinese respondents had lower evaluations (b = − 0.99).

Table 4 Study 2 results
Full size table

Next, we ran a linear regression for PI, with independent variables celebrity, country, and CET (median-split by country, per procedures outlined by Steenkamp et al., 2003; p. 60). Also, we included all two-way and the three-way interactions. The PROCESS (Model 3) coding and output is shown in Table 4, Panel B.

The regression for PI (R2 = 0.27) offered two key results. First, we noted a positive main effect of celebrity (b = 0.62); this relates to responses from those less ethnocentric. Considering conditional effects, we found a higher PI when the foreign celebrity endorsed the foreign brand (India b = 0.62; China b = 0.49). Therefore, in both India and China, respondents with lower CET scores appeared to be less motivated by consumer CO fit concerns, and more motivated by brand CO fit concerns.

Second, we noted a negative celebrity × ethnocentrism interaction effect (b = − 0.66). This related to responses from more ethnocentric respondents, and revealed that for those with high CET scores, PI was not any different for the foreign brand–foreign celebrity pairing than the foreign brand–local celebrity pairing. The negative three-way interaction effect (b = − 2.21) indicated that, in China, more ethnocentric consumers expressed significantly lower PI for the pairing of foreign brand–foreign celebrity than that of foreign brand–local celebrity. The conditional effects results clarify the above. Among more-ethnocentric respondents, respondents from India did not indicate a significantly different PI for the foreign brand–foreign celebrity (vs. foreign brand–local celebrity) pairing (b = − 0.05, t = − 0.25, p = 0.81). However, respondents from China indicated a lower PI for the foreign brand–foreign celebrity pairing (vs. foreign brand–local celebrity; b = − 2.37). In general, the above suggests that – in both India and China – more ethnocentric respondents were (relatively-speaking) more motivated by consumer CO fit concerns, with this effect being stronger in China.

Discussion

While Study 1 results suggested some differences between respondents from India and China, Study 2 results offer more nuance, revealing both similarities and differences. That is, for both Indian and Chinese respondents considering foreign brands, those with lower (higher) CET scores are relatively more motivated by brand CO (consumer CO) fit. These results affirm a moderating role of consumer ethnocentrism, as predicted in Hypothesis 3a. Given that our hypotheses are not country specific, we point out that Study 2 results indicated that Hypothesis 3a sustains across both India and China (separately). But we also identify some subtle differences, consistent with conjectures that Indian consumers tend to be relatively less ethnocentric (Hsu & Nien, 2008; Pereira et al., 2002). In this sense, Indian respondents with higher CET scores are relatively indifferent to endorsements from foreign or local celebrities, but Chinese respondents with higher CET scores have lower evaluations when the endorsement comes from a foreign celebrity.

STUDY 3

Two points led us to conduct Study 3. First, consistent with Hypothesis 3b, Studies 1A and 1B show that when evaluating local brands, both Chinese and Indian respondents (examined separately) offer lower evaluations when the endorsement was from a foreign celebrity. However, when evaluating local brands, is it possible that those more cosmopolitan/less ethnocentric may have higher evaluations for advertisements involving foreign celebrities? We investigate this question in Study 3. Because Indians (vs. Chinese) tend to exhibit behaviors that are relatively less ethnocentric, we ran Study 3 only in India. If there is even the slightest chance that – when evaluating local brands – respondents have higher evaluations for advertisements involving foreign celebrities, then such chance will be higher in the case of respondents from India.

Second, in the case of foreign brands, consumers’ ethnocentrism levels moderate the tradeoff between brand CO fit and consumer CO fit concerns (Hypothesis 3a). For robustness purposes, in Study 3 we test if consumers’ cosmopolitanism levels (expected to be negatively correlated with consumer ethnocentrism levels) also moderate the tradeoff between brand CO fit concerns and consumer CO fit concerns.

Method

The method is similar to our previous studies, except that we only included Indian respondents, and that we (additionally) gathered cosmopolitanism scores, using a unidimensional scale from Cleveland et al. (2014) which has been validated across countries. As in Study 2, we elicited PI and AB on nine-point scales, and we obtained CET (and cosmopolitanism) ratings. This is a 2 (celebrity CO: local vs. foreign) × 2 (brand CO: local vs. foreign) × 2 (CET/cosmopolitanism: low vs. high) between-subjects design (“Appendix 3”, Panel C).

Results

Respondents answered the manipulation check questions correctly, indicating that they could suitably identify the brand CO and the celebrity CO. Also, as expected, cosmopolitanism scores and CET scores were negatively correlated (Pearson correlation = − 0.14, p < 0.01; consistent with Dogan & Yaprak, 2017, Table 1 and with Siamagka & Balabanis, 2015, Table 3). We median-split cosmopolitanism and CET scores (following Steenkamp et al., 2003). We present two (separate) analyses, (1) using CET scores, and (2) using cosmopolitanism scores.

Analysis #1 – Consumer Ethnocentrism. Similar to Study 2, we ran linear regressions for PI and AB with three independent variables: celebrity, brand, and CET scores. The regression results for PI and AB were similar. As in Study 2, we present results for PI (but report results for both PI and AB in Table 5, Panels A and B, respectively; PI R2 = 0.81; AB R2 = 0.74).

Table 5 Study 3 results
Full size table

We found a negative main effect of celebrity (b = − 3.04). This related to the local brand, such that the evaluations were lower when the foreign (vs. local) celebrity endorser was used. Conditional effects related to the local brand were negatively signed, indicating that using a foreign celebrity to endorse the local brand hurt evaluations, irrespective of whether respondents were less ethnocentric (b = − 3.04) or more ethnocentric (b = −1.90). The results stated in this paragraph are consistent with Hypothesis 3b.

The positive brand × celebrity interaction effect (b = 6.89) and negative three-way interaction effect (b = −7.21) related to evaluations of the foreign brand. The conditional effects clarify the results above. Among those less-ethnocentric, evaluations were higher for foreign brand–foreign celebrity endorser pairings than for foreign brand–local celebrity endorser pairings (b = 3.85). This was not the case for those more-ethnocentric (b = − 2.23). The results stated in this paragraph are consistent with Hypothesis 3a.

Analysis #2 – Consumer Cosmopolitanism. We ran another set of (note – separate set of) linear regressions for PI and AB, with independent variables celebrity, brand, and cosmopolitanism. Regression results for PI and AB were similar, as we report in Table 5, Panels C and D, respectively (PI R2 = 0.50; AB R2 = 0.48). Below, we discuss PI results.

We found a negative main effect of celebrity (b = − 2.66), related to the local brand, such that consumers’ evaluations were poorer in the local brand–foreign celebrity endorser condition than in the local brand–local celebrity endorser version. The conditional effects results clarified these results, indicating that evaluations for the local brand-foreign celebrity pairing were lower for both those less-cosmopolitan (b = − 2.66) and those more-cosmopolitan (b = − 1.80). Specifically, even those more-cosmopolitan had lower evaluations when a foreign (vs. local) celebrity endorsed local brands. The results in this paragraph are consistent with the Hypothesis 3b, in the sense that – in the case of local brands – consumers’ evaluations are always lower when these brands are endorsed by a foreign celebrity, irrespective of consumer-type (more- versus less-ethnocentric, more- versus less-cosmopolitan etc.). Importantly, this represents an exception to prior IB research, which has indicated that those more cosmopolitan generally seek a connection with a global consumer culture.

The interaction effects of brand × celebrity (b = 2.70) and brand × celebrity × cosmopolitanism (b = 1.73) were all positive, reflecting evaluations of the foreign brand. Considering conditional effects, evaluations of less cosmopolitan respondents were not significantly higher (b = 0.04, t = 0.16, p = 0.87) for the foreign brand–foreign celebrity endorser pairing. In contrast, those more cosmopolitan, who (per prior IB research) seek a connection to a global consumer culture, had higher evaluations for the foreign brand–foreign celebrity endorser pairing (b = 2.63). Noting that cosmopolitanism is negatively correlated with ethnocentrism, the results stated in this paragraph are consistent with Hypothesis 3a.

GENERAL DISCUSSION

Theoretical Contributions

We contribute to the work on CO, specifically as applicable to the important domain of celebrity endorsements. This paper presents a framework to evaluate the impact of celebrity endorsements, and what makes this framework very relevant to IB is that this framework is based on CO congruency. Our empirics show how consumers’ evaluations are contingent on their preference for (1) consumer CO fit, or (2) brand CO fit; furthermore, we identify contingencies whereby one or the other fit-type dominates. This leads to a novel finding, that in the case of those less ethnocentric, brand CO fit concerns can potentially override consumer CO fit concerns (thus, consumers’ evaluations were higher when Cruise (vs. Kumar) endorsed Fossil watches).

Next, we contribute to work on consumer ethnocentrism and cross-cultural branding. This paper develops a framework for evaluating celebrity endorsements, contingent on use of segmentation involving differences in consumer ethnocentrism levels. Also, and important, prior work conceptualizes consumer ethnocentrism as a negative bias against foreign-made brands. In this paper, we show that an endorsement from a local celebrity may (partially) counteract this ethnocentrism bias. As we show in Studies 2 and 3, when advertising to those more-ethnocentric, using a local celebrity endorser can improve evaluations of such foreign-made goods. It is not that consumers start perceiving these “foreign” brands as “local” (in our studies, all respondents identified the CO of the foreign brand correctly), but rather that those more ethnocentric relatively prefer endorsements from local celebrities, and so endorsements from such celebrities improves their evaluations.

Finally, (some) prior research on consumer cosmopolitanism has suggested that those more cosmopolitan prefer foreign goods and symbols (e.g., celebrities) and seek to be a part of a global consumer culture. We show that this desire is not universal, but contingent on the CO of the brand being endorsed. In the case of foreign brands, those more cosmopolitan do indeed react more positively to endorsements from foreign celebrities, consistent with their desire to be part of a global consumer culture. However, in the case of local brands, those more cosmopolitan do not prefer a foreign celebrity endorser; rather, they react relatively more positively to endorsements from local celebrities.

Contributions to Practice

Considerable confusion exists in practice about which celebrities are optimal endorsers. For example, advertising in India and China features all the various possible combinations of local/foreign celebrities endorsing local/foreign brands (“Appendix 6”). This paper provides some guidance, suggesting optimal marketing mix strategies, contingent on segmentation decisions. First, when choosing celebrity endorsers for foreign brands, managers should explicitly prioritize either brand CO fit (align the celebrity with the brand) or consumer CO fit (align the celebrity with the target consumer segment). Specifically, managers of foreign brands (but not local brands) should prioritize brand CO fit for segments involving less ethnocentric consumers (e.g., younger consumers; Liu, Murphy, Li, & Liu, 2006) and use foreign celebrity endorsers. However, they should focus on consumer CO fit for segments involving higher levels of ethnocentrism, and use local celebrity endorsers.

Second, the senior managers we interviewed at the start of this study, also provided feedback on our findings. While these managers broadly concurred with our findings, managers based in China rejected the idea of using a foreign celebrity endorser. One senior manager (who sells European CO white goods in China) identified a strong need to establish the brand as “right for China … using a Chinese celebrity helps.” Yet our results suggest that, contrary to this managerial wisdom, there may be benefits to using foreign celebrity endorsers when marketing foreign brands to less ethnocentric Chinese consumers.

Limitations and Future Research

Firms have little control over brand CO; in any specific market, the brand will be viewed as either foreign or local, and consumers’ evaluations will be dependent on levels of consumer ethnocentrism and cosmopolitanism. This point led us to take brand CO as a given, and focus on the choice of a suitable celebrity endorser; we encourage further research to do the same but consider other elements, beyond brand CO fit and consumer CO fit, that help define suitable celebrity endorsers.

Also, we used controlled studies to examine our hypotheses. Field studies would complement our findings, as also tighten the linkage to practice.

Next, this paper highlights a key tradeoff for brand managers of foreign brands expanding in EMs. On the one hand, brand managers can use a single celebrity endorser, typically a foreign celebrity endorser with “global” appeal, and run a consistent advertising campaign across multiple EM countries (see Özsomer & Simonin, 2004). On the other hand, many EMs have sizes large enough to justify specific advertising campaigns. In such EMs, to the extent that there exist large consumer segments that are relatively ethnocentric, this paper indicates that using local celebrity endorsers is optimal. We leave the examining of such tradeoffs, which is related to the globalization versus localization debate (e.g., Albaum & Tse, 2001) to future research.

Relative to the brands we examined in this paper, EM brands such as Emirates Airlines and Etihad Airlines cater less to consumers in their home market (UAE) and more to global consumers (spanning developed markets like the UK, and EMs like India). For such brands, seeking to sell across multiple countries, local celebrity endorsers may (again) be less useful. One of the UAE-based senior managers we spoke to indicated that this was exactly why “Emirates uses Jennifer Aniston, and Etihad uses Nicole Kidman … to connect with consumers across … Europe and the US … and not for connecting with consumers in the UAE”.

Fifth, future work could take this research further. For example, we examined a foreign brand with origins in the USA, which has many well-known celebrity endorsers. However, future research can test which type of celebrity endorser is optimal for a foreign brand that originates in a country with few well-known celebrities. In this case, when selling into EMs like India and China, should those managing the foreign brand seek an endorser who is well-known globally but has a different (but still foreign) CO, or seek a less known celebrity endorser who shares CO with the foreign brand? Examining such tradeoffs might produce useful insights.

Finally, there is a conflict in the literature on consumer cosmopolitanism. On the one hand, Dogan and Yaprak (2017; 1506) state that “consumer cosmopolitanism had a … positive effect on willingness to buy foreign products”. On the other hand, Cannon and Yaprak (2002; Figure 1) argue that cosmopolitans do not necessarily seek out foreign products, but seek out authentic products and experiences. Study 3 results (Analysis 2) may provide guidance. In cases relating to foreign brands, where both the choice of local celebrity endorsers (due to consumer CO fit) and foreign celebrity endorsers (due to brand CO fit) can be justified, those more cosmopolitan do indeed have higher evaluations when foreign celebrity endorsers are used. However, in cases relating to local brands, there is no justifying the choice of foreign celebrity endorsers, as both consumer CO fit concerns and brand CO fit concerns motivate the use of local celebrity endorsers. In such cases, those more cosmopolitan have lower evaluations when foreign celebrity endorsers are used. It could well be that in cases wherein use of foreign celebrity endorsers cannot be justified, then endorsements from such celebrities is seen as “inauthentic”, and so perhaps this is why those more cosmopolitan do not prefer the use of foreign celebrity endorsers. While more research is needed to more fully understand and resolve this conflict, Study 3 results may shed light into the direction future research must go. Beyond this point, globalization theorists also consider how global cultural forces may become indigenized (“glocalization”; Alden, Steenkamp & Batra, 2006); how glocalization may impact the debate (above) is unexamined, but is worthy of investigation.

Notes

  1. 1

    The first and second authors conducted approximately 20 semi-structured interviews, via e-mail and telephone, that started with two sets of questions to motivate discussions: “In your experience, considering firms that have major brands (e.g., of stature like Titan, Pepsi, Vicks etc.), who makes the final call as regards choice of celebrity endorser? Does it get made at level of brand manager, CMO or whoever heads the marketing function? Or does the CEO have to sign off?” and “When it comes to choosing a celebrity endorser, how important are issues as regards country-of-origin (CO) of brands vs. CO of celebrities? Specifically: [a] In general, do Indian [Chinese] brands generally prefer using Indian celebrities or using foreign (non-Indian [non-Chinese]) celebrities? [b] In general, do Indian [Chinese] subsidiaries of foreign brands generally prefer using foreign celebrities or using Indian celebrities?” Subsequently, we presented our key findings to the managers and probed any areas in which their opinions differed from those findings.

  2. 2

    Expanding the scales does not change the objective, relative standing of any alternative in the choice set, but it increases the perceived discriminability among options (Burson, Larrick, and Lynch 2009).