Towards an international understanding of the power of celebrity persuasions: a review and a research agenda

Abstract

Research into advertising using celebrity has been undertaken for nearly 40 years. It has principally used surveys and experiments to explore how consumers respond to celebrity advertisements. A recent meta-study of 32 papers has demonstrated that different populations respond in different ways to celebrity endorsements. Specifically, both US subjects and college students are more likely to respond in a significant way to the presence of celebrity than subjects who are not from the US, or who are not studying at college. Given that the nationality and student status of subjects matter, this article explores the make up of the samples that have been used to examine celebrity advertising. The article finds that these samples are not representative of US populations (because so many are students), nor of populations outside the US (because so few live beyond it). Furthermore, the history of dominance of US-based student samples, and the citation practices which keep them circulating in academia, suggests that theories of celebrity advertising have for a long time been excessively influenced by ideas tested on this unrepresentative group. This fact will limit the applicability of research into celebrity advertising to the wider world. I explore whether this matters, and how deficiencies might be addressed in further research.

Keywords:

• advertising
• marketing
• celebrity endorsements
• college students
• sample analysis

Introduction

The effect of celebrity on different consumer groups has provoked continual interest in advertising and marketing journals. In US markets, advertising using celebrities represents a significant minority of all television advertising (nearly 10% of television advertising in 2002/03; Choi et al. 2005). Moreover, it has provided some phenomenal success stories (Pringle 2004). Regardless of whether celebrity advertising always works so well, the fact that it can succeed remarkably whets the appetite of advertisers.

It is important to distinguish between how celebrity advertising works with consumers and how people think it works. There is often a great deal of faith in the market in the power of celebrity advertising. When rumours of Michael Jordan’s return to the National Basketball Association hit Wall Street in 1995, firms whose products enjoyed his endorsement enjoyed a 2% increase in their value – worth in total almost $1 billion (Mathur et al. 1997). This was a good return on the £32 million Jordan was being paid annually to endorse products in 1993. Similarly, when firms announce celebrity endorsements in advertising campaigns before the campaigns begin, their stock value tends to rise (Agrawal and Kamakura 1995).

Such faith in celebrity advertising tells us that it can reap dividends. But it does not tell us how celebrity advertising works with consumers. Researchers have used four main theories to explain the effectiveness of celebrity advertising:

1. The ‘source credibility’ model suggests that celebrity endorsements work because of the credibility of the celebrity as a spokesperson for the product. Credibility here has three elements – attractiveness, trustworthiness and expertise (Hovland and Weiss 1951, Ohanian 1991).

2. The ‘source attractiveness’ model explores one of the dimensions of the credibility model – attractiveness – in more depth, given the power of beauty also to connote intelligence and believability (Friedman et al. 1976, Kahle and Homer 1985).

3. The ‘product match-up hypothesis’ examines the fit between the image of the celebrity and the product. This extends ideas developed in the first two theories to consider how the qualities of the celebrity map onto the qualities of the product they are endorsing (Kamins 1990, Kamins and Gupta 1994).

4. Finally there is the more complicated ‘meaning transfer model’ developed by Grant McCracken in the late 1980s (McCracken 1989). His theory posits that the socially ascribed meaning (what we would call now aspects of their brand) which public figures build up over the course of their career is transferred to the product through endorsements. If consumers identify with the celebrity, or at least recognise their socially ascribed meaning, then that meaning will be transferred to the products being endorsed, and from the product to the consumer by the consumer when she purchases it.

Research based on these theories into how celebrity advertising works on different consumer groups has now been undertaken for the best part of 40 years. It is useful at this juncture to consider what has been learnt from this research in combination. There have been few published reviews. One was produced by Erdogan (1999), and a meta-analysis was conducted by Amos et al (2008) which I discuss below. Both were concerned with what could be concluded from the findings of previous studies with respect to the predictions of the theories summarised above.

This article provides a different sort of overview. My concern is to examine the methods used to test these theories, and specifically the samples used. In this article I first explore why these samples matter. Then I consider what the empirical basis of research used to test theories of celebrity advertising has been. I discuss the implications for the biases found and consider how they might be addressed with further research into celebrity advertising.

Samples and methods in celebrity advertising research

Research into celebrity has almost exclusively used quantitative methods to explore the impact of celebrity endorsements. Qualitative studies exist (Tantiseneepong et al. 2012), but are rare. This is in keeping with other marketing research. A recent review of nearly 2600 papers published between 2003 and 2009 in the nine leading marketing journals found only 99 that had used qualitative and quantitative methods (Harrison and Reilly 2011). Furthermore, the bibliographies of quantitative studies do not lead back to qualitative research. Journals such as Qualitative Market Research do not feature in their references. The insights of qualitative research have rarely been used by researchers of celebrity advertising.

With respect to the quantitative methods applied, two have been used most frequently. Experimental approaches present advertisements with or without celebrity to different groups of people and explore their responses. Surveys ask samples of consumers their opinions about celebrity advertising. Both methods have tended to involve large sample sizes; the mean sample size of the studies reviewed here is 225. There have also been proxy measures of advertising effectiveness which have looked at how celebrity advertising altered company share values, and surveys of business leaders’ opinions about celebrity advertising. These again have used quantitative methods.

What does this accumulation of quantitative data tell us about celebrity advertising? In an important paper, Clinton Amos and colleagues conducted a meta-review on 32 studies (culled from 87 on the topic) to identify the most significant effects across the published literature (Amos et al. 2008). A significant effect indicates that the subjects examined responded in some significant way to the presence of the celebrity in the advert, either positively or negatively. An insignificant effect indicated that the presence of the celebrity did not appear to make much difference to subjects’ thinking. The meta-study showed, perhaps unsurprisingly, that credible, attractive and expert celebrities are more effective than those who are not. The most important factor was negative information about a celebrity, which was a powerful predictor of endorsements not working.

However, Amos et al.’s meta-study also identified two other, perhaps more interesting, variables which affected the outcome of the studies. They observed that studies of students are much more likely to find significant celebrity effects than studies of non-student populations, and that studies of US consumers are more likely to observe significant effects of celebrity endorsement than consumers living in other parts of the world (Amos et al. 2008, pp. 221 and 226). Specifically they observe that of the 185 significant effects of celebrity advertising on consumers observed across their survey, 73% derived from US samples and 63% could be attributed to student populations. Of the 81 insignificant effects of celebrity advertising on consumers, 66% were observed in non-US populations and 71% among non-students.

I have redrawn the relevant part of Amos et al.’s table using revised data supplied by the authors (Table 1; Amos, pers. comm., 5 October 2012). This shows the distribution of effects by population group. It is clear that student populations are highly likely to respond significantly to celebrity advertisements because 83% of effects are significant among student samples (Table 1). Non-students are more evenly split: 53% of effects are significant. In US populations, 80% of effects are significant; in non-US populations, only 43% are. In both cases these differences in distribution are statistically significant (chi-square test results are reported in the footnote to Table 1). Or in plain English, US populations and student populations are likely to respond in significant ways to celebrity advertisements. The presence of celebrity in an advertisement appears to affect the way they think or behave. Non-US populations are unlikely to respond in significant ways, whereas non-student populations are more equivocal.

Table 1. Distribution of source effects from Amos et al. (2008).

Sample characteristics	All effects	Significant effects	Non-significant effects
Student	157	129	28
Non-student	145	77	68
Total	302	206	96
US	187	150	37
Non-US	127	55	72
Total	314	205	109

Note: Data courtesy of Clinton Amos, 5 October 2012.

Chi-square test results:

student/non-student, χ² = 29.36; df = 1; p < 0.001; US/non-US, χ² = 45.45; df = 1; p < 0.001.

It is unfortunately not possible to explore how populations who are neither from the US or who are not students compare with the effects demonstrated by populations who are both from the US and who are students. The relevant SPSS file required to produce those data cannot now be located (Amos, pers. comm., 5 October 2012). Nevertheless it is clear that the two results do exist separately from each other, and would be likely to accentuate each other. In both US and non-US samples a majority of the sample were students, so we cannot explain the difference between countries according to their student status. Similarly for both students and non-students a majority of the sample were from the US, so we cannot explain differences according to nationality.

These differences are important. They matter because if quantitative methods such as experiments and surveys are used to test theories, then we have to be sure that the samples used are representative of a broader population. This is a prerequisite of the quantitative methods used thus far. If they are not representative of a broader population, then we cannot safely extrapolate to other groups. US students are not representative of the rest of the US with respect to their response to celebrity, because students respond differently from non-students. Nor are US subjects representative of populations from different countries for the same reason. Theories of celebrity advertising which want to be relevant for these other consumer groups will need to be tested on different, more appropriate, populations.

If research into celebrity endorsements has been using inappropriate samples, then it is possible that models of celebrity advertising are being built on false pretenses. Advertising and marketing professionals, and students on advertising and marketing courses will read papers about the power and effectiveness (or otherwise) of celebrity advertising but be unable to determine rigorously which population groups it is appropriate to apply those theories to. There is a risk of over-extrapolating the significance of celebrity advertising’s effects by testing theories of celebrity advertising on populations who respond more to celebrity than do other groups.

It is therefore quite important to quantify more precisely how many of the people asked about celebrity advertising in surveys or experiments are students or based in the US, and how many are that more common type of consumer – people who are not students or who are not living in the US. We need, in other words, to explore in detail the make-up of the samples that lead to these publications. That task provides the main substance of this article. It presents one general research question and three sub-questions:

• Research Question 1: Of which populations are samples collected to test theories about celebrity advertising representative?

  • Sub-Question 1: What proportion of the samples is composed of US subjects?

  • Sub-Question 2: What proportion of the samples is composed of students?

  • Sub-Question 3: What proportion of the samples is composed of US students?

It is also important to examine how this knowledge has evolved over time. Is research based on US students dated? Is the work on non-US subjects very recent? Are there any trends in the development of the collective sample of celebrity advertising research that might affect the body of knowledge produced by it? This generates the second research question:

• Research Question 2: How has the constitution of the samples used in celebrity advertising research changed over time?

Finally, we are assuming that this knowledge matters because it is read and noted by the academe and by marketing and advertising professionals. However, we cannot assume that each paper is equally impactful. It might be quite possible for the samples used to test a theory to be unrepresentative of a larger population, but for readers to gain a more representative picture by using the papers selectively. They might, for example, take much more interest in results drawn from non-student populations, and this may counter a tendency to over-sample students. This provides the last research question:

• Research Question 3: How have data based on different sorts of samples been used by other researchers working on theories of celebrity advertising?

Methods

To survey the celebrity advertising literature I first turned to existing surveys. I included all the studies I could obtain from the Amos et al (2008) review, and from Erfgen’s (2011) and Erdogan’s (1999) reviews of these literatures. Then I searched for other more recent papers (since 2004) in the journals in which these authors had found their studies. I also included theses I have come across in other literature and studies on the impact of celebrity on political campaigns. The journals I surveyed are presented in Table 2.

Table 2. Journals surveyed for this research.

Journal title
Advances in Consumer Research
American Journal of Business
Annual Review of Psychology
International Journal of Advertising
Journal of Advertising
Journal of Advertising Research
Journal of Business Research
Journal of Consumer Marketing
Journal of Consumer Psychology
Journal of Consumer Research
Journal of Managerial Finance
Journal of Marketing Research
Journal of Political Marketing
Journal of Popular Culture
Journal of Product and Brand Management
Journal of Services Management
Journal of the Academy of Marketing Science
Marketing Management
Psychology and Marketing

This search produced 96 relevant papers and theses, of which 83 were published papers. For each document I recorded the number of people who took part in the experiment/survey, whether they were students or not (whether graduates or undergraduates), and which country they lived in. Recording the country of the study was not always straightforward because it was not always mentioned, particularly in some of the earlier papers. However, since the authors of these papers were based in US universities, were generally describing work undertaken with students, and would often refer to the students’ university according to whether it was ‘East Coast’ or ‘Mid-West’ and so forth, it was plain where the study was undertaken. I also recorded the samples’ mean age and their gender, but not enough studies provided this information for it to be used in the analysis.

I have presented in the following information about the samples from the published papers only, because it is these which are most accessible in the public domain. These documents are most able to exert influence on the thinking of advertising and marketing professionals, and students, who may encounter them. Therefore the make-up of their samples is important. The papers used for the following data are marked with an asterisk in the Reference list.

To gauge the nature of a paper’s influence I collected citation data from the Web of Knowledge for each published article. The flaw in this method is that it can result in unpopular papers scoring highly, even though they are only referred to because of their flaws. However, having read these papers it is clear that this happens rarely, if at all.

Results

The vast majority of the papers in this literature review are based on work undertaken by scholars working in the US (Table 3). Of the 83 papers published on celebrity advertising, 76% were based on work there. The samples for the other papers are scattered all over the world, with Europe providing the main secondary concentration of research (nine published papers).

Table 3. Countries in which the work in the published papers analysed in this article was conducted.

Region	Country	Number of papers
North America	USA	63
	Canada	2
Europe	UK	2
	Germany	2
	Ireland	2
	Norway	1
	Austria	1
	Netherlands	1
Africa	South Africa	1
Asia	China	1
	Hong Kong	1
	India	1
	Iran	1
	Japan	1
	South Korea	1
	Pakistan	1
	Singapore	1
Total		83

Unsurprisingly, the vast majority of the people constituting the samples of surveys, or taking part in experiments, were resident in the US (Table 4). Most of the studies have also worked with samples of students, although students are only a minority of the non-US subjects. The vast majority of these samples were from single-country studies. There are only a couple of incidences when authors have tried to make international comparisons and surveyed people outside their own countries.

Table 4. Number of students and non-students from the US and non-US countries making up the samples of the published papers.

Country	Non-student	Student	Total
Non-US	3098	3298	6396
USA	3314	14,598	17,912
Total	6412	17,896	24,308

The tendency to survey US populations, and US students, here is notable (Table 4). Seventy-three per cent of the people sampled in these studies were from the US, and 79% of those subjects were students. Altogether 58% of all the people making up the samples in all of the published papers were US students, and only 13% of all the samples represent people who are neither from the US nor college students.

Please note, however, that samples taken from outside the US have a much more even split of students to non-students (48% to 52%). This is still a higher proportion of the population than students actually constitute, but nonetheless is more representative than the US samples. This means also that US students dominate the student samples. Only 23% of students who were sampled for quantitative research on celebrity came from outside the US.

More striking still than the stark US-student bias of this research is the fact that US populations were exclusively sampled by the authors of this literature until relatively recently (Table 5). Non-US populations only began to be studied from 1998 onwards. The geographical spread of subjects in the samples outside the United States is now quite diverse, with no particular country dominating, although Europe is the best region represented (Table 6). But this is a recent phenomenon. The fact that such samples now constitute as much as 25% of the total sample population is due in part to just four Asian surveys, which account for half of the non-US samples assembled here. Without those four surveys, US populations would still constitute nearly 90% of the sample population.

Table 5. Sampling of different countries and populations over time.

Total	Non-US		US		Total	Cumulative totals			Cumulative proportions (%)
	Non-student	Student	Non-student	Student		Cumulative total	Cumulative US subjects	Cumulative US students	% US subjects	US students of US subjects	US students of all subject
1951	–	–	–	223	223	223	223	223	100	100	100
1976	–	–	–	150	150	373	373	373	100	100	100
1977	–	–	200	–	200	573	573	373	100	65	65
1978	–	–	75	166	241	814	814	539	100	66	66
1979	–	–	360	–	360	1174	1174	539	100	46	46
1981	–	–	–	99	99	1273	1273	638	100	50	50
1982	–	–	–	135	135	1408	1408	773	100	55	55
1983	–	–	196	335	531	1939	1939	1108	100	57	57
1985	–	–	–	277	277	2216	2216	1385	100	63	63
1989	–	–	139	–	139	2355	2355	1385	100	59	59
1990	–	–	265	852	1117	3472	3472	2237	100	64	64
1991	–	–	343	268	611	4083	4083	2505	100	61	61
1992	–	–	100	–	100	4183	4183	2505	100	60	60
1994	–	–	–	1446	1446	5629	5629	3951	100	70	70
1996	–	–	–	147	147	5776	5776	4098	100	71	71
1997	–	–	–	1086	1086	6862	6862	5184	100	76	76
1998	100	–	–	842	942	7804	7704	6026	99	78	77
1999	–	–	–	200	200	8004	7904	6226	99	79	78
2000	–	–	152	425	577	8581	8481	6651	99	78	78
2001	141	–	–	–	141	8722	8481	6651	97	78	76
2002	–	–	–	864	864	9586	9345	7515	97	80	78
2003	–	880	–	106	986	10,572	9451	7621	89	81	72
2004	–	130	218	249	597	11,169	9918	7870	89	79	70
2005	280	706	–	232	1218	12,387	10,150	8102	82	80	65
2006	–	249	–	339	588	12,975	10,489	8441	81	80	65
2007	–	500	–	962	1462	14,437	11,451	9403	79	82	65
2008	631	342	566	1556	3095	17,532	13,573	10,959	77	81	63
2009	100	–	–	1409	1509	19,041	14,982	12,368	79	83	65
2010	316	319	–	–	635	19,676	14,982	12,368	76	83	63
2011	1530	–	700	503	2733	22,409	16,185	12,871	72	80	57
2012	–	172	–	1727	1899	24,308	17,912	14,598	74	81	60
Total	3098	3298	3314	14,598	24,308

Hovland and Weiss 1951; Friedman et al. 1976; Fireworker and Friedman 1977; Friedman et al. 1978; Sternthal et al. 1978; Friedman and Friedman 1979; Mowen and Brown 1981; Freiden 1982; Klebba and Unger 1982; Atkin and Block 1983; DeSarbo and Harshman 1985; Kahle and Homer 1985; Kamins 1989; Kamins et al. 1989; Kamins 1990; Misra and Beatty 1990; Ohanian 1990; Langmeyer and Walker 1991; Ohanian 1991; Gail et al. 1992; Heath et al. 1994; Kamins and Gupta 1994; Lynch and Schuler 1994; Moore et al. 1994; Tripp et al. 1994; Basil 1996; Basil and Brown 1997; Nataraajan and Chawla 1997; O’Mahony and Meenaghan 1997/8; Till and Shimp 1998; TIll and Busler 1998; Cronley et al. 1999; Lafferty and Goldsmith 1999; Goldsmith et al. 2000; Till and Busler 2000; Erdogan et al. 2001; Lafferty et al. 2002; Louie and Obermiller 2002; Stafford et al. 2002; Pornpitakpan 2003; Priester and Petty 2003; Silvera and Austad 2003; Batra and Homer 2004; Bush et al. 2004; Chao et al. 2005; Forehand and Perkins 2005; Jackson and Darrow 2005; La Ferle and Choi 2005; Biswas et al. 2006; Money et al. 2006; Choi and Rifon 2007; Saleem 2007; Van der Waldt et al. 2007; Wood and Herbst 2007; Austin et al. 2008; Chan and Prendergast 2008; Jackson 2008; Klaus and Bailey 2008; Lee and Thorson 2008; Marshall et al. 2008; Martin et al. 2008; Siemens et al. 2008; Till et al. 2008; Zahaf and Anderson 2008; Edwards and La Ferle 2009; Koernig and Boyd 2009; Samman et al. 2009; Wheeler 2009; White et al.2009; Eisend and Langner 2010; Ranjbarian et al. 2010; Veer et al. 2010; Hung et al. 2011; Newman et al. 2011; Nownes 2011; Pughazhendi et al. 2011; Becker 2012; Choi and Rifon 2012; Miller and Allen 2012; Rice et al. 2012; Rossiter and Smidts 2012.

Table 6. Sample size from countries other than the US.

Region	Country	Sample size
North America	Canada	654
Europe	Austria	280
	Germany	224
	Netherlands	172
	Ireland	200
	Norway	130
	UK	503
Africa	South Africa	200
Asia	China	1030
	Hong Kong	631
	India	500
	Iran	193
	Japan	249
	South Korea	250
	Pakistan	300
	Singapore	880
Total		6396

I have presented this information in two ways in Table 5. The left-hand side of the table presents the raw data, and the right-hand side presents the cumulative proportions of the make up of the collective samples by US subjects and US students respectively. The cumulative proportions are important because they give an indication of what anyone reviewing the literature will encounter as they look back through it. Thus, someone reading the literature at the start of 2002 would encounter samples totalling over 9400 subjects, 97% of whom were American, and 78% of whom were US college students.

Furthermore, not only is there a historically dominant body of US-based, and US-student-based, knowledge about celebrity advertising, but papers based on such subjects are much more influential within the literature than papers based on work in other countries, and on other population groups. Papers based on US subjects account for 96% of all citations. Those based on US students account for 72% of citations, whereas US students only constitute 58% of the collective sample (Table 7).

Table 7. Citation count of different sample types.

	Non-student	Student	Total
Non-US	52	78	130
USA	725	2136	2861
Total	777	2214	2991

Such dominance is to be expected given that papers earlier on in this sequence could only cite US-based work; no other work existed. However, since papers based on other countries have begun to be published they have tended to be cited less than papers published based on US subjects which were published in equivalent years. Papers based on non-US populations are cited on average 0.6 times for each year of their existence, and those on US populations 2.2 times per year.

For the purpose of this argument it matters not why these non-US papers are cited less. They may simply deserve little attention. What matters is that samples used to test theories of celebrity advertising are based largely on US subjects, and that this effect is exaggerated by the attention paid to these samples in the literature.

Discussion

Geographical and historical differences

There is one important caveat we must consider with respect to Amos et al.’s findings before we can consider further implications of these results. It is possible that the difference between US and non-US populations which Amos et al. (2008) identify reflects a historical difference not a geographical difference. No research on celebrity advertising was conducted outside the US before 1998. The difference between the US and non-US populations could therefore reflect a difference between results produced before and after that year. It is theoretically possible for both US and non-US samples to be producing few significant results after 1998. This would be possible if the non-significant US effects produced after 1998 were masked by the scale of the significant results found in US samples before 1998. If this has happened then it would suggest that US populations are taking less notice of celebrity now than then, and have become more like non-US consumers.

Interest in celebrity can wax and wane over time. But the scenario I have just sketched is an unlikely possibility. I have not come across any indication that interest in celebrity has been declining in the US in the last 14 years. If anything, the popular perception is that the reverse is true. Certainly within academia there has been an increase in the power and reach of celebrity over this same time period (Beer and Penfold-Mounce 2010). Assuming therefore that the differences between the countries are real, we can proceed to discuss their implications

The implications of the US/student bias

There are obvious reasons for the biases demonstrated here. The US houses a large number of advertising and marketing professors, who command the ready attention of a large number of students in their classes. They can offer the students course credits for taking part in their experiments and surveys (a common practice) and thus can produce a large amount of fact about responses to celebrity advertising for free.

However, this method presents a problem for the accumulated celebrity advertising literature because its findings are derived from the most celebrity-orientated population. Theories about celebrity advertising have been based on the collective views of a population which is not representative of the rest of the US. Furthermore, the US populations are not representative of the rest of the world. The extent to which these theories will be useful for talking about the responses to celebrity advertising from different population groups is more limited than has thus far been realised.

Note that I am not criticising any individual author of any particular paper. Most papers I have read were quite careful to consider the limitations of their sample for the conclusions they could draw. Most recognised that students were not representative of the population as a whole – but were still an important market and deserved attention for that reason. These are sound arguments. Their cumulative effect, however, is to produce an unrepresentative picture of responses to celebrity advertising.

These effects will have been particularly marked up until relatively recently. Since 2000, a significant minority, but a growing proportion, of the subjects sampled has not been US subjects (Figure 1). Since most people mostly read the more recent literature, this may diminish the some of the domination of the US focus of the exclusively American era of advertising research pre 2000. However, anyone reading back though the literature will quickly encounter work that was either only composed of US samples or was written informed by such work. In that respect the influence of a biased sample early on in the formation of a body of literature has an enduring effect. There is inertia towards such knowledge. In any case, while recent literature is more balanced than before, it is still unbalanced. US subjects still account for most of the sampling conducted for most of the papers published about celebrity advertising. Moreover, their dominance in the literature is accentuated by citation practices which pay more attention to research conducted on US populations.

Figure 1. Trends in sampling in celebrity marketing literature from 2000 onwards.

It is not possible to determine from these data whether the limited social basis of tests about theories of celebrity advertising matters very much. We know that should business, advertising and marketing students, practitioners and researchers explore this literature, then they will be exposed to unrepresentative results. Their understanding of the power of celebrity advertising may be skewed by the domination of a celebrity-orientated population’s views about it in the literature which they read. On the basis of the evidence presented here, students and researchers reading about the effects of celebrity advertising will not have been able to glean a reasonable picture of what celebrity advertising does to most people. We cannot tell, however, how influential those theories actually are either in the minds of these students or in the practices of advertising and marketing professionals.

Even if these results have been influential, then would it matter? Do the opinions of non-US subjects matter with respect to celebrity advertising? Not if we are concerned only with US markets. Indeed, given that such consumers behave so differently, people interested in US markets and US consumers should actively ignore research based elsewhere. The presence of non-US research in recent years would simply be a distraction for such people. All that is required is to undertake more research into non-student populations, because the current constitution of the collective sample is not representative of US populations.

However, regardless of its actual influence, it is unsatisfactory to have a body of theory based on such limited samples and for those limitations to persist unrecognised. It could distort readers’ views of the world. Moreover, if non-US markets are interesting, then we will need to take more interest in research based in them. With globalisation these markets are becoming more important. Internationally famous figures are being used to endorse brands across different countries, and at international gatherings (the Olympics, the FIFA World Cup) which concentrates attention of audiences from numerous countries. As brands (of products and celebrities) expand their international reach, advertisers will need to build a stronger empirical base upon which to base the value of their associations.

Globalisation, however, does not mean that we can assume the world is becoming more uniform. It is more connected, but the different economies and societies being connected can vary considerably. There is good reason to think that they vary in response to celebrity. The few studies which have explicitly compared US with non-US samples have noted the value of doing so. For example, Silvera and Austad, comparing Norwegian and US students, suggested that the lack of influence of celebrity could reflect the lower prominence of celebrity culture, and celebrity credibility generally, in Norwegian (and Nordic) countries more generally (Silvera and Austad 2003). I have demonstrated elsewhere that, with respect to British audiences, public reception of celebrity advocacy for good causes is surprisingly muted, at the same time as it is still widely believed to be powerful (Brockington 2014, Brockington and Henson 2014).

Conversely there are also comparisons with Asian countries, where celebrity advertising is apparently much more extensive and, by implication, more powerful than in the US. Choi and colleagues compared the use of celebrity in advertisements in South Korea and the USA (Choi et al. 2005). Celebrity was used much more in the former country, with celebrity appearing in 57% of 841 South Korean adverts (the samples were collected in 2002 and 2003). They also note that the tone of the adverts differs, with celebrities more likely to play themselves in the US and to act roles in South Korea. Money et al. (2006) report Kilburn’s findings that in the late 1990s in Japan 70% of all advertisements, and 90% of all that were ‘likable, popular or memorable’, featured a celebrity. In contrast, in Choi et al.’s survey, just 9.3% of 2354 US commercials on the three major TV channels featured celebrities.

Clearly therefore the categorisation of ‘US’ and ‘non-US’, encompassing as it does celebrity-sceptical countries and celebrity-receptive countries, is an unsatisfactory one. Moreover, even those continental sub-divisions are unsatisfactory. I make them because the knowledge base beyond the US is thin, not because they are inherently reasonable. Rather than using crude regional generalisations, theories of the power of celebrity advertising will need to take into account different national celebrity cultures and different advertising regimes, as the other papers in this volume demonstrate. Work such as that by Choi et al. (2005), which is deliberately and systematically building up an understanding of advertising in South Korea, will be necessary in order to explore how responses to advertising vary within and between countries.

A clear research agenda therefore emerges from this analysis. We need to have nationally and regionally differentiated studies of the variable power of celebrity persuasions. We have an authoritative body of knowledge on the responses of US college students. Celebrity advertising research now needs to expand its horizons and examine how different audiences respond (or fail to respond) to celebrity endorsements in other countries, and particular demographics within those countries.

But my suggestion, in closing, is that in expanding their sample base, celebrity advertisement researchers also expand their methodological repertoire. Thus far researchers’ preference for using quantitative methods does not appear to be justified by the wealth of insights these methods afford. It is clear that the knowledge built up has not properly addressed the challenges of sampling upon which effective quantitative methods hinge. It would be possible for new research to address these problems and proceed as before, but simply in a broader variety of countries, and with fewer students in the US. With the methods currently employed, this work will ultimately produce a list of significant and insignificant results among different specified portions of different countries’ populations. This list could then be subject to further meta-analyses of the sort Amos and colleagues conducted, producing a refined and more significant set of results.

However, it is plain from the research conducted thus far that these methods will produce little more than such a list. While this might be a reasonably interesting resource for advertisers, it will not really explain how particular effects become significant (or not) through consumers’ consumption practices. Qualitative methods are particularly useful when addressing the meaning of consumption practices, which is likely to be central to any understanding of how celebrity works (Marshall 1997, Turner et al. 2000, Rojek 2001, Turner 2004, Ferris 2007). These will be obviously useful when exploring how celebrity works in contexts foreign to researchers, but also provide a good means of understanding societies with which we think we are familiar. In current marketing practice, the utility of more qualitative, and specifically anthropological and ethnographic, methods is increasingly well recognised (Anderson 2009, Jackson 2009, Radjou 2009, Reinan 2009).

These methods could explore the ways in which celebrity endorsements work among different groups – what sort of identity work they involve and how meaning is constructed between celebrity and consumer brands in the minds of different publics. They could explore the different consequences of direct and indirect endorsements. This would entail in-depth interviews, focus groups, diaries, q methodology and participant observation. Through such methods it would be possible to learn more about the differentiated consumption practices that different populations demonstrate in their following, and not following, celebrity endorsements.

If qualitative methods can be countenanced by the community of researchers exploring the impacts of celebrity advertising, then quite an exciting research programme could unfold. A mixed-methods approach that, based on representative samples, combines the predictive power of quantitative methods with the quality of insight that qualitative methods afford would be a powerful thing. Driven by such approaches, it might be possible for a representative and rigorous theory of celebrity advertising to emerge.

Conclusion

Research into the effectiveness of celebrity advertising has largely been based on US students. The samples assembled are, collectively, not representative of the US population as a whole. Nor are they representative of other parts of the world. Theories and models of celebrity advertising which are built upon the collective wisdom of the research undertaken thus far can only safely be applied to a relatively small proportion of consumers.

A weighty body of knowledge has built up based on these samples, and apparently in ignorance of the problems in the sample base. There is evidence to suggest that these flaws have not been noticed before. Amos and colleagues discussed the different responses of these groups, but not their consequences for the samples collected. It is particularly notable that the total dominance of US research until the late 1990s excited no comment. Indeed, current citation practices still accentuate the bias towards US-based samples. It is quite difficult, even for diligent students, to learn about how celebrity advertising may work in other parts of the world or other populations. The bias of earlier work is likely to have an enduring effect unless it is explicitly recognised.

This means that theories of celebrity advertising have a much more limited scope than has previously been realised. It does not mean that they are wrong, or that their scope cannot get broader. The far greater prevalence of celebrity advertising in some Asian countries could mean that US students, who are more attentive to celebrity than their compatriots, may make a good empirical base from which to learn about other population groups. My point is simply that current theories have been tested in a limited sector of the market, and have to be extrapolated with care.

More insight into the power of celebrity advertising internationally therefore will require more research in other countries. Current research cannot be safely applied there. My plea, however, is that such work should undertake more qualitative research into the meaning of consumption practices which celebrity enables or stymies. Even when applied to the US student populations of which they are most representative, some of the current conclusions are not particularly surprising. I am not convinced that advertisers needed to be told that credible, attractive and expert celebrities are effective endorsers. Understanding which groups of the population are most swayed by such endorsers in other countries would have only limited value. Exploring the richer ways in which consumers interact with celebrity through qualitative research and mixed methods may provide more useful insights.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Funding

Funding for this research was generously provided by the Economic and Social Research Council [RES 070-27-0035].

Notes on contributors

Dan Brockington

Dan Brockington is a Professor of Conservation and Development at the University of Manchester. Trained as an anthropologist in London, he has worked previously in the geography departments of the Universities of Oxford and Cambridge. His research has covered the social impacts of conservation policy, the relationships between capitalism and conservation, the work of media and celebrity in development, and long-term livelihood change in East Africa. He has conducted most of his research in Tanzania but has also worked in South Africa, Australia, New Zealand and India, as well as conducting global overviews of the social impacts of protected areas, media and conservation and continental wide examinations of the work of conservation NGOs in sub-Saharan Africa. Dan is happiest conducting long-term field research in remote areas but also learns much from studying plush fundraising events. He has recently published Celebrity Advocacy and International Development (2014) and has also written Celebrity and the Environment (2009), Nature Unbound (2008) (with Rosaleen Duffy and Jim Igoe) and Fortress Conservation (2002), and co-edited (with Rosaleen Duffy) Capitalism and Conservation (2011).