Abstract
In this study, I report the results of a pretest-posttest, control group experiment in which some of my more than 500 respondents were exposed to factual information about celebrity support for political parties and some were not. I proceed from the assumption that celebrity political activity is more likely to influence citizens’ views of political parties than it is to affect either citizens’ vote choices or views of individual candidates. I make this assumption based on the work of Green, Palmquist, and Schickler, who posit that party identification is a social identity. The results provide support for this notion. Specifically, they show that celebrity political activity can indeed influence some citizens’ views of political parties. The results show also that celebrity political activity can affect citizens’ views of politically active celebrities.
Celebrities are part of American politics. Some celebrities become politicians; Ronald Reagan, Arnold Schwarzenegger, and Jesse Ventura are the most obvious examples. Others endorse candidates. For example, during the 2008 presidential election, George Clooney, Samuel L. Jackson, Brad Pitt, will.i.am, Oprah Winfrey, and a slew of other entertainment heavyweights endorsed Senator Barack Obama. Concurrently, Jon Cryer, Clint Eastwood, Lee Greenwood, Chuck Norris, Hank Williams Jr., and several others endorsed Senator John McCain. Still, other celebrities engage in lower profile political activity, especially contributing money to political parties and candidates. A search of Federal Election Commission (FEC) disclosure reports shows that literally hundreds of Hollywood celebrities, sports stars, and other famous people contribute money to candidates and political parties. Most of the time these contributions occur neither with fanfare nor with publicity. Does all this celebrity political activity matter? Does it affect the way ordinary citizens view candidates and/or parties? Does it affect the way ordinary citizens view politically active celebrities? In this study, I address these questions via a randomized pretest-posttest, control group laboratory experiment in which some of my 503 respondents (i.e., respondents in the two treatment groups) were exposed to factual information about celebrity support for political parties and some were not. I proceed from the assumption that celebrity political activity is capable of affecting citizens’ views of political parties. I make this assumption based on the work of Green, Palmquist, and Schickler (2002), who posit that party identification (in America and probably elsewhere) is a social identity—one that is affected by individuals’ views of which groups are associated with each party. Overall, my data indicate that celebrity political activity can indeed influence some citizens’ views of political parties. The results show also that celebrity political activity can affect citizens’ views of politically active celebrities.
Background and Hypotheses
The American presidential election of 2008 was unique in many ways, but prodigious celebrity activism was not one of them. The media paid particular attention to Oprah Winfrey when she said of Barack Obama, “he’s the one,” but her public endorsement was hardly novel. Examples of celebrity endorsements are manifold. For example, W.C. Fields weighed in for Wendell Willkie in 1940, Wilt Chamberlin endorsed Richard Nixon in 1968, and Muhammad Ali endorsed Ronald Reagan for reelection in 1984. Moreover, celebrities of all types have in one way or another supported candidates and/or parties since virtually the beginning of the republic. Other political activists of the last century include John Wayne (active Republican), Gregory Peck (staunch Democrat), Frank Sinatra (friend of the presidents of both parties), and Marilyn Monroe (registered Democrat). In short, celebrities and American politics have always mixed.
Celebrities: Do They Matter?
Politically active celebrities are most noticeable during election time, when media outlets breathlessly report candidate endorsements and show pictures of candidates mugging with movie and music and sports stars. Do these endorsements matter? Few political scientists have addressed this question, but the limited extant evidence is suggestive. Garthwaite and Moore (2008), for example, explore the effects of Oprah Winfrey’s high-profile endorsement of Barack Obama during the 2008 Democratic presidential primary, and conclude that it “had statistically and politically significant effects on Obama’s” eventual selection as the Democratic nominee (p. 39). They find that in areas where there were greater concentrations of Oprah Winfrey fans, Barack Obama enjoyed a higher level of electoral success than he would have absent the endorsement. Pease and Brewer (2008) also find an “Oprah effect.” Using experimental data, they find that although Oprah’s endorsement had no impact on how much respondents liked Barack Obama, it did affect their views of Obama’s viability as a candidate. Specifically, respondents who learned of Oprah’s endorsement viewed Obama as more likely to win than did respondents who did not learn of Oprah’s endorsement. Perhaps more importantly, the endorsement also led exposed respondents to say that they were more likely to vote for Obama. However, this effect was limited to those who had a strong positive opinion of Oprah. Other studies show celebrity effects as well. In a recent inventive quasi-experimental study, political scientist David Jackson (2007) finds that celebrity musicians are capable of affecting their fans’ attitudes and opinions on political issues. Celebrities with a well-known history of political activism are particularly influential. In another study, Jackson and Darrow (2005) show that popular celebrities can influence young people’s attitudes about political issues. However, not all the evidence points toward celebrity effects. In a large-scale survey-based study of first-time voters in the 2004 presidential election, two marketing professors find that celebrity endorsements have very little impact, as voters consistently cite personal and family relationships as the most important determinants of their voting preferences (Wood & Herbst, 2007).
Communications scholars have studied celebrity endorsements to a greater extent than have political scientists, and their findings suggest that celebrity endorsements are far from insignificant. For example, in a study of young adults in the United Kingdom, Veer, Becirovic, and Martin (2010) find that among poorly informed, apathetic respondents, celebrity endorsements affect party evaluations. Other studies suggest that celebrities can, as Inthorn and Street (2011, p. 3) state, “connect citizens with a political cause,” as many voters view celebrities as more trustworthy and credible than politicians (but see Frizzell, 2011). Finally, Austin, Van de Vord, Pinkleton, and Epstein (2008) show that celebrities can positively affect levels of political engagement among young voters (see also Lopez, Marcelo, & Kirby, 2007; Street, Hague, & Savigny, 2008).
In short, the limited evidence that exists, although not definitive, suggests the following: Celebrity endorsements matter. Where does this influence come from and how does it work? While political scientists do not have a particularly compelling answer to this question, marketing and advertising scholars do. Numerous marketing studies show that celebrity endorsements of products can increase sales (see, for example, Agrawal & Kamakura, 1995; Butler, Cowan, & Nilsson, 2005; Erdogan, Baker, & Tagg, 2001; Farrell, Karels, Montfort, & McClatchey, 2000). The most compelling explanation for how and why this is the case comes from anthropologist and economist Grant McCracken. His “meaning transfer theory” holds that a celebrity increases the attractiveness of a product by transferring his or her meaning to that product (see McCracken, 1986, 1989). The celebrity in question develops a public persona through his or her activities (e.g., the roles he or she plays on television or in movies), and, over time, he or she is imbued with a specific meaning (e.g., he or she is seen as tough, responsible, competent, beautiful, or sophisticated). When a celebrity endorses a product, this meaning is then transferred to that product. Influence occurs to the extent that people like the celebrity, like what he or she represents, and proceed to like the product he or she endorses.
Celebrities and Party Identification
Here, I contend that celebrities can influence people’s views of political parties just as they influence people’s views of products. A celebrity who is politically active has a persona. When this celebrity endorses a party, people who like the celebrity—that is, people who identify with the celebrity and are open to guidance from him or her—transfer the qualities of the celebrity to the political party the celebrity endorses or supports. This view fits in nicely with Green et al.’s (2002) compelling model of partisanship. According to this model, partisanship is a form of social group identification. A person acquires a party identification based on his or her perceptions of the social groups that comprise each party. Partisanship is the result of a psychological process of “self-categorization and group evaluation” (p. 13). This process works like this. First, people figure out which groups they belong to. Second, people learn which groups are aligned with which party. Finally, people choose the party that appears to them to contain the group or groups they belong to. Summarizing the whole process, Green et al. state, “people ask themselves two questions: What kinds of social groups come to mind as I think about Democrats, Republicans, and Independents? Which assemblage of groups (if any) describes me?” (p. 8). This view is at odds with most previous views of party identification. It contradicts rational choice treatments (e.g., Fiorina, 1981), for example, in its assertion that neither retrospective judgments of candidates nor party performance or policy concerns substantially affect party identification. This view also contradicts the Michigan school’s conceptualization of party identification as a more or less inherited heuristic tool or time-saving cue (see Campbell, Converse, Miller, & Stokes, 1960). And contrary to The American Voter and its scholarly progeny, Green et al. find that partisanship does not act as a “perceptual filter” that leads to biased learning.
When and how does a person form his or her party identification? Green and his colleagues (2002) conclude, “Partisan attachments form relatively early in adulthood” (p. 4). It is in early adulthood that people learn which groups identify with and comprise the two parties. Young adults learn from a variety of sources including parents and friends. Here, I contend that young adults learn also from the media that convey information about celebrity political activity. I contend that celebrity support for political parties represents valuable information about which groups are aligned with each party. To the extent that people attend to this information, celebrity endorsements of political parties can affect citizens’ views of the parties. How exactly? As per Green et al., a person asks himself or herself, “What kinds of social groups come to mind as I think about Democrats, Republicans, and Independents?” (p. 8). Information about celebrity endorsements helps people address this question. For example, a media report that Ben Affleck is a die-hard Democrat conveys to voters that suave, young, “face men” are Democrats. Similarly, a media report that Mike Ditka is a strong Republican conveys that athletic, tough, “hard guys” are Republicans. Next, the person asks, which of these groups—that is, suave men, young people, “face men,” athletic people, tough people, or “hard guys”—do I most identify with? After addressing this question, a person adjusts his or her views of the parties accordingly.
In short, I contend that a celebrity’s endorsement of a political party can affect how people view and evaluate the political parties. I am certainly not saying that celebrity endorsements are either all important or determinative. Instead, I am saying that information about celebrity endorsements is one type of information among many types that people may use in their evaluations of political parties.
Hypotheses
Celebrities do not “endorse” political parties per se. They endorse candidates to be sure, but they seldom publicly endorse entire political parties by saying things such as, “I like the Republican Party and think you should vote Republican.” Thus, here I do not explore the impact of explicit celebrity endorsements of political parties. Rather, I explore the impact of information about celebrity support for political parties. Specifically, in what follows, I explore the impact of two pieces of information. The first is information about Hollywood movie star Jennifer Aniston’s financial support for Democrats. The second is information about National Football League star Peyton Manning’s financial support for Republicans. (I will say much more about my specific research design in a moment.) In what follows, I test eight simple hypotheses about the effects of information on celebrity support for political parties on people’s views of the two major parties in the United States. My specific hypotheses—which flow from the extant research—are as follows:
Hypothesis 1: For respondents who view Jennifer Aniston positively, evaluations of the Democratic Party will become more positive with exposure to information about her financial support for Democrats and the Democratic Party.
Hypothesis 1a: For respondents who view Peyton Manning positively, evaluations of the Republican Party will become more positive with exposure to information about his financial support for Republicans and the Republican Party.
Hypothesis 2: For respondents who view Jennifer Aniston negatively, evaluations of the Democratic Party will become less positive with exposure to information about her financial support for Democrats and the Democratic Party.
Hypothesis 2a: For respondents who view Peyton Manning negatively, evaluations of the Republican Party will become less positive with exposure to information about his financial support for Republicans and the Republican Party.
Hypothesis 3: For respondents who view Peyton Manning positively, evaluations of the Democratic Party will become less positive with exposure to information about his financial support for Republicans and the Republican Party.
Hypothesis 3a: For respondents who view Jennifer Aniston positively, evaluations of the Republican Party will become less positive with exposure to information about her financial support for Democrats and the Democratic Party.
Hypothesis 4: For respondents who view Peyton Manning negatively, evaluations of the Democratic Party will become more positive with exposure to information about his financial support for Republicans and the Republican Party.
Hypothesis 4a: For respondents who view Jennifer Aniston negatively, evaluations of the Republican Party will become more positive with exposure to information about her financial support for Democrats and the Democratic Party.
Of course, I am primarily interested in how information about celebrity support for political parties affects people’s views of the political parties. However, as I have the data and the question is an interesting one, I will also examine how information about celebrity political support for political parties affects people’s views of the celebrities themselves. Specifically, I test the following four hypotheses:
Hypothesis 5: For respondents who view the Republican Party positively, evaluations of Peyton Manning will become more positive with exposure to information about his financial support for Republicans and the Republican Party.
Hypothesis 5a: For respondents who view the Republican Party negatively, evaluations of Peyton Manning will become less positive with exposure to information about his financial support for Republicans and the Republican Party.
Hypothesis 6: For respondents who view the Democratic Party positively, evaluations of Jennifer Aniston will become more positive with exposure to information about her financial support for Democrats and the Democratic Party.
Hypothesis 6a: For respondents who view the Democratic Party negatively, evaluations of Jennifer Aniston will become less positive with exposure to information about her financial support for Democrats and the Democratic Party.
Data and Method
To test these hypotheses, I conducted a randomized pretest-posttest, control group experiment (see Shadish, Cook, & Campbell, 2002, pp. 115-123). My subject pool comprised 503 university students who participated in exchange for extra course credit. I conducted the experiment during a 2-week period in the spring of 2010. I randomly assigned participants to one of three groups—the control group (n = 177), the Aniston treatment group (n = 145), or the Manning treatment group (n = 181). 1 On the first day of the experiment, each student (no matter the group) was administered a brief questionnaire (i.e., the pretest questionnaire) comprising 18 survey items. The questionnaire contained several questions about each respondent’s political attitudes and affiliations (e.g., 2008 vote choice, ideology, party identification) as well as his or her opinions of nine politically active celebrities, four contemporary political figures, the “tea party” movement, and four political parties. 2 The survey also contained a few questions about respondents’ personal characteristics (e.g., academic major, age, family income, home state, and race/ethnicity). A copy of the survey can be found on my website at http://web.utk.edu/~anownes/. Prior to completing a pretest survey, respondents were told that they were participating in a university-wide study designed to gauge the political attitudes and opinions of young people. Each respondent was given the following project description:
Project Description
This research is intended to continue the social scientific study of political attitudes and behavior in the United States. Specifically, it will seek to add to our growing knowledge of the development of attitudes in young people—that is, how young people form the opinions they have. Specifically, in this study we are seeking to determine how young people form their attitudes and opinions about political figures, UT personalities, and celebrities.
On the second day of the experiment, respondents in the three groups were treated differently. Students in the control group took the identical survey again. Students in the Aniston treatment group were given the posttest survey with a cover sheet that looked like this:
Obama and McCain: Both Pet Lovers
Barack Obama and John McCain have something in common: They both are pet lovers. The country’s “First Dog” is named Bo Obama. He is a Portuguese water dog. John McCain is also an animal lover. Reports indicate that he has more than 20 pets, including many saltwater fish, four dogs, a cat, and two turtles.
Jennifer Aniston Among Democratic Donors
Recent data released by the Federal Election Commission (FEC) indicate that actress Jennifer Aniston contributed heavily to the Democratic Party and President Barack Obama during the 2008 campaign. Aniston, best known for her work on the long-running television program Friends, contributed $2,300 to Barack Obama’s campaign committee in 2008, and another $26,200 to the Democratic National Committee.
Respondents in the Manning treatment group were given the posttest survey with a cover sheet that looked like this:
Obama and McCain: Both Pet Lovers
Barack Obama and John McCain have something in common: They both are pet lovers. The country’s “First Dog” is named Bo Obama. He is a Portuguese water dog. John McCain is also an animal lover. Reports indicate that he has more than 20 pets, including many saltwater fish, four dogs, a cat, and two turtles.
Peyton Manning Among Republican Donors
Recent data released by the Federal Election Commission (FEC) indicate that quarterback Peyton Manning has contributed heavily to Republican candidates in recent years. Manning, the former University of Tennessee quarterback and current signal caller for the Indianapolis Colts, contributed $1,000 to Senator Bob Corker in 2009. Manning also contributed $4,200 to Corker in 2006, $2,000 to the Bush/Cheney presidential campaign in 2004, and $2,300 to Fred Thompson’s political action committee (PAC) in 2007.
In short, respondents in the control group completed identical pre- and posttest questionnaires, respondents in the Aniston treatment group completed a questionnaire with a cover sheet containing information about Jennifer Aniston’s support for the Democratic Party and Democratic candidates, and respondents in the Manning treatment group completed a questionnaire with a cover sheet containing information about Peyton Manning’s support for Republican candidates. The cover sheets presented to respondents in the two treatment groups also contained an innocuous story about the two 2008 presidential contenders’ love for pets in addition to the actual stimulus. I included this innocuous information to obscure the true purpose of the experiment.
This classic pretest-posttest, control group experiment allowed me to test the effects of celebrity political activity in a straightforward and simple manner. My basic strategy was to compare changes in respondents’ ratings of the political parties and the celebrities in question between the control group and the two treatment groups.
I worked hard to maximize internal validity. First, I randomly assigned respondents to the three experimental groups. (This helps with external validity as well.) As I mentioned above, I asked respondents a few questions about themselves, and the groups were very similar in composition. For example, there were no statistically significant differences between the three groups in mean age, mean income level (using a 4-point scale), and percentage from the state of Tennessee (the state in which I conducted the study). Second, I did not reveal the true nature of the experiment to participants until after the experiment was complete. I did this to ensure that respondents did not succumb to demand characteristics. Third, to overcome evaluation apprehension (i.e., the tendency for some respondents to be afraid of testing or evaluation), I told respondents that their participation was completely voluntary and that they could receive extra course credit even if they did not participate. I also told respondents that their answers would be both confidential and anonymous. Finally, to ensure that I had enough respondents to allow for meaningful statistical analyses, I started with a sample of 600, knowing that there would be substantial attrition (which there was). In the end, despite attrition, the sample size is large enough to allow for statistical tests.
I also took several steps to maximize external validity. First, to make the experiment as realistic as possible, as stimuli, rather than using fake candidates or doctored news reports (as some laboratory experiments do), I used real-life factual information (obtained from the Federal Election Commission) concerning the political activities of Peyton Manning and Jennifer Aniston. Second, as participants, I chose only potential or actual adult voters. Third, in this section, I provide a detailed description of my data collection methods and analysis strategies. This, I hope, will allow for replication as well as follow-up studies. I realize that we can only fully understand the impact of celebrity political activity by conducting multiple studies. This is the first, I hope, of many others.
Before moving on to my findings, I feel compelled to address two additional issues. First, there is the “college sophomore problem”—the perceived threat to external validity posed by conducting the study on a sample of college students. Does my reliance on college students substantially threaten this article’s external validity? Probably not. As Druckman and Kam (2011) point out, college student samples are surprisingly similar in many ways to nonstudent samples. Moreover, the variables on which students are likely to differ from nonstudents (e.g., education, family income) are in no way purported here to affect the size or direction of the treatment effect. In addition, the basic theoretical framework on which this study is based speaks directly to the impact of information on young adults. And for the most part, college students are young adults. Second, I will address the question of why I chose the specific celebrities for this study. Why did I select Peyton Manning and Jennifer Aniston? First of all, I believed that a thorough test of my hypotheses necessitated the selection of one Republican and one Democrat. I began with a list of politically active Democrats and Republicans, and Aniston fell into the former category whereas Manning fell into the latter. Second, to make my findings more broadly applicable, I selected celebrities from two distinctly different parts of the entertainment world. Manning is a football player, whereas Aniston is an actor. Third, to make sure my respondents were familiar with the celebrities in question, I chose two celebrities with very high “q scores” (which indicate the level of the public’s familiarity with celebrities). There are many politically active celebrities, but Manning and Aniston are two of the best known.
Results
To test Hypotheses 1 to 4 and 1a to 4a, I used two dependent variables. They are Democratic Party posttest thermometer rating and Republican Party posttest thermometer rating. To test Hypotheses 5, 6, 5a, and 6a, I used two additional dependent variables. They are Peyton Manning posttest thermometer rating and Jennifer Aniston posttest thermometer rating. All four of these variables are based on the standard feeling thermometer survey item from the American National Election Studies (ANES). Democratic Party posttest thermometer rating has a range of 0-100, a mean of 49.0, and a standard deviation of 28.0. Republican Party posttest thermometer rating has a range of 0-100, a mean of 56.9, and a standard deviation of 26.0. Peyton Manning posttest thermometer rating has a range of 0-100, a mean of 81.0, and a standard deviation of 21.2. Jennifer Aniston posttest thermometer rating has a range of 0-100, a mean of 63.5, and a standard deviation of 24.
The Independent Variables
My primary independent variables of interest are Aniston treatment and Manning treatment. Aniston Treatment takes a value of 1 if a respondent fell into the Aniston treatment group, and 0 otherwise. Manning treatment takes a value of 1 if a respondent fell into the Manning treatment group, and 0 otherwise. The baseline category here is the control group. To test my hypotheses, I also used several other independent variables, all also based on the standard ANES feeling thermometer question. First, there is Democratic Party pretest thermometer rating which has a mean of 49.0 and a standard deviation of 27.2. Second, there is Republican Party pretest thermometer rating which has a mean of 56.6 and a standard deviation of 26.2. Third, there is Jennifer Aniston pretest thermometer rating which has a mean of 65.6 and a standard deviation of 23.4. Fourth, there is Peyton Manning Pretest Thermometer Rating which has a mean of 82.3 and a standard deviation of 20.4. Finally, I created four interaction terms, each of which is self-explanatory: (a) Aniston treatment × Jennifer Aniston pretest thermometer rating, (b) Manning treatment × Peyton Manning pretest thermometer rating, (c) Aniston treatment × Democratic Party pretest thermometer rating, and (d) Manning treatment × Republican Party pretest thermometer rating.
Table 1 contains summary information on the dependent variables and their corresponding independent variables. I include the table primarily to provide information about the structure of the data.
Summary Statistics on Dependent Variables and Their Corresponding Independent Variables
Source: Author’s data.
Note: Ns vary between variables owing to missing data.
The Effects of Treatment on Party Thermometer Ratings
I began by testing Hypotheses 1 to 4. Specifically, I built a simple ordinary least squares (OLS) regression model in which the dependent variable is Democratic Party posttest thermometer rating. The model (Model 1) included the following independent variables: Aniston treatment, Manning treatment, Democratic Party Pretest Thermometer Rating, Jennifer Aniston Pretest Thermometer Rating, Aniston treatment × Jennifer Aniston pretest thermometer rating; and Manning treatment × Peyton Manning pretest thermometer rating. I included the two treatment variables in the model for obvious reasons. I included Democratic Party Pretest Thermometer Rating to control for respondents’ pretest views of the Democratic Party. Finally, the two interaction terms are included because as my hypotheses state, I expect the treatments to have different effects on people depending upon how they view the celebrities in question.
Aniston and the Democrats
The second column in Table 2 presents the results of this analysis. Of most interest here is the significant coefficient on Aniston treatment (p < .05, two-tailed test). This indicates that the information about Jennifer Aniston’s support for Democrats had some impact on exposed respondents’ views of the Democratic Party. However, it is impossible to determine the precise nature of this impact without taking a closer look at the interaction term, Aniston treatment × Jennifer Aniston pretest thermometer rating (which is also significant at p < .05, two-tailed test; see Brambor, Clark, & Golder, 2006). The coefficient on Aniston treatment (–9.10) represents the treatment effect for respondents who scored 0 on Jennifer Aniston pretest thermometer rating. This indicates that among respondents who gave Jennifer Aniston a thermometer rating of 0, exposure to the information about her political giving had a strong reductive impact on the Democratic Party’s thermometer rating. To fully understand the effects of the treatment (i.e., the information about Jennifer Aniston’s support for Democrats) on respondents’ views of the Democratic Party, I produced Figure 1. The solid sloping line in the figure shows how the Aniston treatment affects the values on Democratic Party posttest thermometer rating and how the effect of the treatment changes with a respondent’s score on Jennifer Aniston pretest thermometer rating. The confidence intervals around the line indicate the conditions under which the treatment has a statistically significant (p < .05, two-tailed test) impact on values of Democratic Party posttest thermometer rating. Figure 1 shows that for respondents who gave Jennifer Aniston a pretest thermometer rating of 0-61, exposure to information about Aniston’s political giving had a negative impact on values of Democratic Party posttest thermometer rating. However, this negative effect decreases as Aniston’s pretest ratings increase, and then disappears altogether for respondents who gave Jennifer Aniston a pretest thermometer rating of 62 or more.
The Effects of Treatment on posttest thermometer ratings of the Political Parties
Source: Author’s Data.
Note: Ordinary least squares estimation with standard errors in parentheses.
p < .10, two-tailed. **p < .05, two-tailed.
Marginal effect of Aniston treatment on Democratic Party posttest thermometer rating as Jennifer Aniston Pretest Thermometer Rating changes
This first set of results provides no support for Hypothesis 1. In sum, it was not the case that respondents who viewed Jennifer Aniston positively rated the Democratic Party higher after being exposed to information about Aniston’s support for Democrats. However, the data provide strong support for Hypothesis 2. Figure 1 clearly shows that people who viewed Jennifer Aniston negatively (i.e., gave her a thermometer rating of 0-49) viewed the Democratic Party less favorably after being exposed to information about Aniston’s donations to Democrats. In addition, Figure 1 shows that people who felt lukewarm about Jennifer Aniston (i.e., who gave her a pretest thermometer rating of between 50 and 61) viewed the Democratic Party less favorably after being exposed to information about her donations to Democrats.
Manning and the Democrats
Next I turn to Hypotheses 3 and 4, which concern the effect of the Manning treatment on views of the Democratic Party. The coefficient on Manning treatment in Model 1 is not statistically significant. However, again, because the model contains an interaction term, care must be taken in interpreting its results. If it were significant, the coefficient on Manning treatment (–2.6) would represent the treatment effect on Democratic Party posttest thermometer rating for respondents who scored 0 on Peyton Manning Pretest Thermometer Rating. I conducted a number of follow-up analyses (the results of which are not presented here but are available from me) and found that exposure to information about Peyton Manning’s political support for Republicans had no effect on respondents’ views of the Democratic Party. This finding leads me to reject Hypotheses 3 and 4.
Next, I turn to Hypotheses 1a to 4a. To test these hypotheses (all of which concern the effects of treatment on respondents’ views of the Republican Party), I built an OLS regression model (Model 2) in which Republican Party posttest thermometer rating is the dependent variable. The model contained the following independent variables: Aniston treatment, Manning treatment, Republican Party Pretest Thermometer Rating, Jennifer Aniston pretest thermometer rating, Aniston treatment × Jennifer Aniston pretest thermometer rating; and Manning treatment × Peyton Manning pretest thermometer rating. Again, I included the two treatment variables in the model for obvious reasons. I included Republican Party pretest rating to control for respondents’ pretest views of the Republican Party. Finally, I included the two interaction terms because I expect the treatments to have different effects on people depending upon how they view the celebrities in question.
Manning and the Republicans
The third column in Table 2 presents the results of this analysis. The coefficient on Manning treatment is not statistically significant, but this tells us little. To understand the effect of the Manning treatment, we must examine the interaction variable Manning treatment × Peyton Manning pretest thermometer rating. To fully explore the effects of the treatment (i.e., the information about Peyton Manning’s support for Republicans) on respondents’ views of the Republican Party, I produced Figure 2. Again, the solid sloping line shows how the Manning treatment affects values on Republican Party posttest thermometer rating, and how the effect of the treatment changes with a respondent’s score on Peyton Manning pretest thermometer rating. The confidence intervals around the line show that for respondents who gave Peyton Manning a pretest thermometer rating of 69-84, exposure to information about Manning’s political giving had a positive impact on values of Republican Party posttest thermometer rating. Contrary to expectations, this positive effect decreases (very slightly) as Manning’s pretest ratings increase, and then disappears altogether for respondents who gave Peyton Manning a pretest thermometer rating of 85 or more. It is worth noting, however, that if I use a one-tailed significance test instead of a two-tailed test (which is defensible in light of my directional hypothesis), the confidence intervals would stretch the upper range of values for which the Manning treatment has a positive effect to 92.
Marginal effect of Manning treatment on Republican Party posttest thermometer rating as Peyton Manning Pretest Thermometer Rating changes
These results provide mixed support for Hypothesis 1a. On one hand, for many people who view Peyton Manning positively (that is, give him a thermometer rating of between 69 and 84), exposure to the Manning treatment does indeed lead to a more positive view of the Republican Party. Moreover, the solid sloping line in Figure 2 remains in positive territory for the entire range of Peyton Manning Pretest Thermometer Rating, which is suggestive of a relatively constant and positive “Manning effect.” On the other hand, contrary to Hypothesis 1a, significance tests indicate that we must reject the notion that among people with the highest opinions of Peyton Manning—that is, people who gave him a thermometer rating of between 85 and 100—exposure to the Manning treatment leads to an enhanced view of the Republican Party. The results provide no support for Hypothesis 2a. In other words, there is no evidence here that people who view Peyton Manning negatively view the Republican Party more negatively after they are exposed to information about Manning’s giving to Republicans.
Aniston and the Republicans
Next, I turn to Hypotheses 3a and 4a which concern the effect of the Aniston treatment on views of the Republican Party. The coefficient on Aniston treatment in Model 2 is not significant. However, because the model contains an interaction term care must be taken in interpreting its results. If it were significant, the coefficient on Aniston treatment (1.37) would represent the treatment effect on Republican Party Posttest Rating for respondents who scored 0 on Jennifer Aniston pretest thermometer rating. I conducted a number of follow-up analyses (the results of which are not presented here but are available from me) and found that exposure to information about Jennifer Aniston’s political support for Democrats had no effect whatsoever on any respondents’ views of the Republican Party. This finding suggests that we should reject Hypotheses 3a and 4a.
The Effects of Treatment on Celebrity Thermometer Ratings
Does information about celebrities’ political giving affect how people view them? To answer this question and test Hypotheses 5, 6, 5a, and 6a, I cast two OLS regression models. In the first, Peyton Manning posttest thermometer rating is the dependent variable, and Manning treatment, Republican Party pretest thermometer rating, Peyton Manning Pretest Thermometer Rating, and Manning treatment × Republican Party pretest thermometer rating are the independent variables. In the second, Jennifer Aniston posttest thermometer rating is the dependent variable and Aniston treatment, Democratic Party pretest thermometer rating, Jennifer Aniston pretest thermometer rating, and Aniston treatment × Democratic Party pretest thermometer rating are the independent variables.
The results of both models are found in Table 3. I will begin with Model 1, the results of which are found in the first column of numbers in Table 3. The significant coefficient on Manning treatment (p < .05, two-tailed test) indicates that the information about Peyton Manning’s support for Republicans influenced exposed respondents’ views of Manning. To determine the precise nature of this impact, we must examine the interaction term, Manning treatment × Republican Party pretest thermometer rating (which is also significant at p < .05, two-tailed test). The coefficient on Manning treatment (–6.84) represents the treatment effect for respondents who scored 0 on Republican Party pretest thermometer rating. In short, among respondents who dislike the Republican Party the most, exposure to information about Peyton Manning’s support for Republicans had a strong reductive impact on Manning’s thermometer rating. Figure 3 illustrates the effects of the treatment on respondents’ views of Peyton Manning. The solid sloping line shows how the Manning treatment affects values on Peyton Manning posttest thermometer rating and how the effect of the treatment changes with a respondent’s score on Republican Party pretest thermometer rating. The confidence intervals indicate the conditions under which the treatment has a statistically significant (p < .05, two-tailed test) impact on values of Peyton Manning posttest thermometer rating. Figure 3 shows that for respondents who gave the Republican Party a pretest thermometer rating of 0-53, exposure to information about Manning’s political giving had a reductive impact on values of Peyton Manning posttest thermometer rating. However, this reductive effect decreases as the Republican Party’s pretest ratings increase, and disappears for respondents with a Republican Party Pretest Thermometer Rating of 54 or more. These results provide no support for Hypothesis 5, but strong support for Hypothesis 5a.
The Effects of Treatment on posttest thermometer ratings of Celebrities
Source: Author’s Data.
Note: Ordinary least squares estimation with standard errors in parentheses.
p < .10, two-tailed. **p < .05, two-tailed.
Marginal effect of Manning treatment on Peyton Manning posttest thermometer rating as Republican Party Pretest Thermometer Rating changes
Now I will turn to Model 2. The second column of numbers in Table 3 presents the results of this analysis. The significant coefficient on Aniston treatment (p < .05, two-tailed test) indicates that the information about Jennifer Aniston’s support for Democrats influenced exposed respondents’ views of Aniston. To determine the precise nature of this impact, we must look closer at the interaction term, Aniston treatment × Democratic Party pretest thermometer rating (which is also significant at p < .05, two-tailed test). The coefficient on Aniston treatment (–13.27) represents the treatment effect for respondents who scored 0 on Democratic Party pretest thermometer rating. In short, among respondents who dislike the Democratic Party the most, exposure to information about Jennifer Aniston’s support for Democrats had a strong reductive impact on Aniston’s thermometer rating. To fully comprehend the effects of the treatment, I produced Figure 4. The solid sloping line shows how the Aniston treatment affects values on Jennifer Aniston posttest thermometer rating and how the effect of the treatment changes with a respondent’s score on Democratic Party pretest thermometer rating. The confidence intervals indicate the conditions under which the treatment has a statistically significant (p < .05, two-tailed test) impact on values of Jennifer Aniston posttest thermometer rating. Figure 4 shows that for respondents who gave the Democratic Party a pretest thermometer rating of 0-41, exposure to information about Aniston’s political giving had a reductive impact on values of Jennifer Aniston posttest thermometer rating. However, this reductive effect decreases as the Democratic Party’s pretest ratings increase. In addition, Figure 4 shows that for respondents who gave the Democratic Party a pretest thermometer rating of 70-100, exposure to information about Aniston’s political giving had an additive effect on Jennifer Aniston posttest thermometer rating. In all, the results of Model 2 provide strong support for both Hypotheses 6 and 6a.
Marginal effect of Aniston treatment on Jennifer Aniston posttest thermometer rating as Democratic Party Pretest Thermometer Rating changes
Conclusion
My results support the general notion that celebrity giving to political parties and their candidates affects peoples’ views of the parties. Specifically, my experimental results show the following. First, when people who dislike Jennifer Aniston are exposed to information about her support for Democrats, they report liking the Democratic Party less. Second, when some people who like Peyton Manning are exposed to information about Manning’s support for Republicans, they report liking the Republican Party more. Together, these results suggest that meaning transfer theory has some utility for scholars of party identification. Specifically, if we think of the political parties as “brands,” these results suggest that information about which celebrities “use” each brand can affect people’s attitudes about the brands. This said, the data I examine here suggest that if celebrities want to help the parties they support by “going public” with this support, they may have a hard time doing so. I say this because Jennifer Aniston’s impact on people’s views of the Democratic Party was only negative. In addition, Peyton Manning’s effects on respondents’ views of the Republican Party were positive, but quite modest.
Of course, I realize that my results are far from the last word on the subject. Because this was a “one-shot” experiment, I cannot speak to the question of how long the treatment effects last. Do treatment effects linger over time? Or are they ephemeral? What types of celebrities are capable of affecting people’s views of the parties? Only further study can answer these questions.
As for the celebrities themselves, the data clearly show that political activity affects how people view them. The data show that people who are not particularly fond of Republicans are turned off by Peyton Manning’s support for the Grand Old Party (GOP) and adjust their opinions of him accordingly. Similarly, people who dislike the Democratic Party view Jennifer Aniston more negatively after learning about her support for Democrats. Aniston, however, unlike Peyton Manning, gets something good from her political activity; the Democratic Party’s biggest fans like her more after learning about her support for Democratic candidates.
In the end, my results suggest that information about celebrities’ political activities can influence how people think about the political parties. It is too early to say whether party identification itself is open to change due to exposure to information about celebrity political activity, but the data are suggestive.
Footnotes
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
The author received no financial support for the research, authorship, and/or publication of this article.