Polls and Elections. Editorial Cartoons 2.0: The Effects of Digital Political Satire on Presidential Candidate Evaluations

Polls and Elections Editorial Cartoons 2.0: The Effects of Digital Political Satire on Presidential Candidate Evaluations JODY C. BAUMGARTNER East Carolina University While the number of full-time editorial cartoonists has declined in the past few decades, several have taken their craft online in the form of animated Flash cartoons. In this article I test the effects of one of the more popular animated editorial cartoons on presidential candidate evaluations of 18- to 24-year-olds. A posttest-only experimental design was used to survey students from several universities in six states. The results from this online experiment suggest that these editorial cartoons have a negative effect on candidate evaluations. However, viewing the clip did not change candidate preferences and an analysis of the control group suggests that viewership of online humor may have a positive effect on political participation. On December 12, 2005, the Association of American Editorial Cartoonists held a protest they referred to as Black Ink Monday in response to the cut of the editorial cartoon staff at the Los Angeles Times and the Baltimore Sun by the Tribune Company (Worcester 2007). While the protest was precipitated by the Tribune s cutback, it was also aimed at drawing attention to the more general trend of corporate downsizing of the profession. According to one source, there are now fewer than 60 full-time editorial cartoonists working in the United States today, down from over 200 in the 1980s (Summers 2006; Danjoux 2007). Many cartoonists, however, have turned their attention to the Internet as an alternative and/or supplementary venue (Lordan 2005, 166-69), suggesting that reports of the death of this particular genre may be premature. In fact, political humor in general flourishes on the Internet, and this is perhaps no more evident than during a presidential campaign. For example, a two-minute YouTube video featuring John Edwards having his hair styled for a media appearance, set to the Jody C. Baumgartner is an assistant professor of political science at East Carolina University. Presidential Studies Quarterly 38, no. 4 (December) 735 2008 Center for the Study of the Presidency

736 PRESIDENTIAL STUDIES QUARTERLY / December 2008 tune of I Feel Pretty, has been viewed over one million times since it was posted in November of 2006. The search term political humor returns several millions of pages, depending on the search engine used, that feature cartoons, comics, videos, jokes, satire, and more. Most of these sites go unnoticed by the majority of people, but some are well visited. In 2004 the popular This Land video by JibJab.com was seen by over 10 million people in the first month after its release, three times the number of visitors the Bush and Kerry Web sites combined attracted during the same time period (Lohr 2004; Political Networking 2004). Although it was the most popular video of the campaign season (Darr and Barko 2004), other sites parodied candidates Web sites (Cornfield 2004), offered high quality videos made for the Web (Darr and Barko 2004), or poked fun at the candidates by way of jokes, cartoons, parody, or satire. But for all of its popularity, we know very little about the effects of online political humor. While it is possible that it simply entertains audiences, it is also possible that it might change the political attitudes of those who view it. This study builds on a small amount of research that examines the effect of various types of political humor. In it, I present the results of an online experiment conducted in late November of 2007 testing the effects of an online animated editorial cartoon on the presidential candidate evaluations of 18- to 24-year-old youth. The cartoon, created by Pulitzer Prize winning cartoonist Walt Handelsman, lampoons the six candidates who were leading in their respective party s race for the nomination by presenting them as participants in a faux reality television show. Consistent with the elaboration likelihood model (ELM) of persuasion, I find that the cartoon lowered overall evaluations of the candidates. Extending the analysis, I show that the clip did not change the candidate preferences of respondents, and analysis of the control group suggests that viewership of online humor may have a positive effect on political participation. The research is important for several reasons. First, political humor during the pre-primary stage of the presidential campaign has greater potential to affect people s attitudes and evaluations for the simple reason that most presidential aspirants are still relatively unknown at this stage. And, although the regular audience for online political humor may be relatively small, the potential audience is much larger. Especially funny political humor has the potential to spread virally (Darr and Barko 2004). In addition, in the midst of a presidential campaign many can find themselves at the various decoy, spoof, or parody sites that use domain names similar to those of the candidates (Cornfield 2004; Crummy 2007). The study also speaks to how youth are socialized into politics in the digital age. This is important because the mass media play an important role in shaping political attitudes, beliefs, and ultimately participation among youth (Chaffee, Ward, and Tipton 1970). However, the media habits of young people today are dramatically different than those of even a generation ago (Wattenberg 2006). Youth are increasingly turning to alternative sources of news for their understanding of the political world. For example, in 2007, 46% of 18- to 29-year-olds reported they got most of their election news from the Internet, an increase of 25% from 2004, while the percentage of those mentioning television as their main source dropped 15%. A full 27% say they got some information about the campaign from social networking sites (Pew 2008). This shift in media habits

Baumgartner / EDITORIAL CARTOONS 2.0 737 is still only dimly understood, but some research exists that suggest that the Internet might serve as an effective mobilization tool, increasing political engagement, interest, and participation among younger citizens (Shah, Kwak, and Holbert 2001). This, in turn, is important because this age group appears to be more cynical about politics than other age groups and consistently ranks lowest in terms of traditional measures of political participation (Longo and Meyer 2006). Because political humor generally casts political leaders and institutions in a negative light, Internet political humor has the potential to exacerbate cynicism among youth, which may cause them to further disengage from the world of politics. Worse, any potential adverse effects could be lasting, as political attitudes are mainly formed in an individual s younger years (Sears 1983). In the next section I briefly discuss research on editorial cartoons and place it in context of the small amount of research on the effects of political humor. Following that, I discuss the methodology of the research and present my findings. In the final section I discuss the implications of this study and suggest areas for further work. Editorial Cartoons and the Effects of Political Humor Evidence has been found suggesting that the political cartoon has been part of the human experience since at least 1360 B.C., when an unknown artist drew a caricature of the unpopular father-in-law of the Egyptian leader Tutankhamen (Danjoux 2007). The rise of editorial cartoons in America dates back to the mid-eighteenth century, attributable to increased population and literacy rates in the colonies, the growth of print media, and, of course, the political turbulence of the times (Lordan 2005). Throughout the nineteenth century editorial cartoons, with their simplification of the political world, played a significant role in the appeal of the daily newspaper. The political power of editorial cartoons during this period gained almost mythical status, as evidenced by the oft-repeated reaction (stop them damn pictures ) of Boss Tweed to Thomas Nast s cartoons (Danjoux 2007, 246). It is believed that political cartoons have the ability to leave an indelible stain on [the] public image of leaders (Danjoux 2007, 246). The reaction to the controversy in September of 2005 surrounding the cartoons depicting Mohammad in Danish newspapers suggests many people think editorial cartoons do matter (ibid.). It is certainly true that most cartoons...[are] unflattering to political leaders and institutions (Buell and Maus 1988, 856), that in general, negative images prevail (Edwards 2001). In fact, as long as editorial cartoons have caricatured politicians, politicians have feared for their public image (Buell and Maus 1988, 847). Do they in fact have the power to influence public opinion? Unfortunately, there is almost no behavioral research examining this question. There are a number of excellent collections of editorial cartoons (Hess and Northrop 1996; Lamb 2004; Trostle 2004; Lordan 2005), but most of the research on the subject examines the history and forms of the genre and the political messages, representations, symbols, and metaphors that they contain (Worcester 2007). In fact, there is some reason to believe that editorial cartoons might not affect public opinion. First, editorial cartoons, like most political humor, builds on preexisting

738 PRESIDENTIAL STUDIES QUARTERLY / December 2008 negative images that many people already hold about the target of the humor (Niven, Lichter, and Amundson 2003; Young 2004a). This makes a certain amount of sense, because the foundation of successful humor (a joke, satire, etc.) must be understood by the audience (Nilsen and Nilsen 2008). Second, it is not entirely clear that all editorial cartoons are interpreted the way they are intended by the artist, or even in the same way by different individuals (Carl 1968, 1970). In other words, even if they do have the potential to change opinion, the effects may vary from person to person. This said, there is a small body of evidence that suggests that editorial cartoons may have some effect on political attitudes. An early study found that editorial cartoons, especially if accompanied with an editorial, had some ability to change opinions about the target being lampooned (Brinkman 1968). Other research (mainly in the fields of psychology and marketing) into the effects of humorous messages more generally suggest that they are both memorable and persuasive (Scott, Klein, and Bryant 1990; Schmidt 1994; Gruner 1996; Berg and Lippman 2001; Lyttle 2001). One early study suggested that the reason that humorous messages might have the potential to affect attitudes is that the receiver s tendency to critically examine the message (or counterargue) is reduced if a message is delivered with humor (Sternthall and Craig 1973). This might be, they speculated, due to the positive mood that results from receiving a humorous message or from increased likeability of the source of the humor. This notion that persuasion and attitude change might result from a reduction in counterargument has subsequently found form in what is known as the ELM (Petty and Cacioppo 1986; Lyttle 2001). In this model of persuasion, communication is thought to be processed along what are known as central or peripheral routes. Central route processing is roughly akin to critical reasoning, involving high elaboration. If a message is processed along the central route, its persuasive power is limited by the power of the argument itself, as well as other factors, including the individual s predisposition. In the peripheral route, message processing involves less cognition and is more dependent on contextual and affective considerations. There is, in other words, less cognition involved, making it more likely the individual will process the message without respect to the substance of the argument. Subsequent research suggests that humor may be processed along the peripheral route, giving humorous messages the power to persuade or influence attitudes(zhang 1996; Young 2004b). The nature or direction of attitude change seems to be dependent on the message contained in the humor or the type of humor. For example, research on the effects of satirical comedysuggestthatviewersmaybeledtoagreewiththedirectmessage(thatwhichisbeing lampooned) rather than with the more subtle indirect message. Studies of the popular situation comedy All in the Family suggested that the program may have influenced or reinforced preexisting racial stereotypes held by viewers (Vidmar and Rokeach 1974; Brigham 1975; but see also Surlin and Tate 1976; Brigham and Giesbrecht 1976). A recent study of the effects of The Colbert Report found that Stephen Colbert s satirical treatment of conservative talk show hosts led respondents to agree with the conservative viewpoints he was implicitly poking fun at (Baumgartner and Morris 2007). On the other hand, self-deprecating humor seems to increase the likeability of the source. This is consistent with an understanding that many politicians have about the

Baumgartner / EDITORIAL CARTOONS 2.0 739 effect of telling jokes at their own expense (Katz 2004). For example, in 1982, Nancy Reagan successfully transformed her image as an elitist by singing Secondhand Clothes at the 1982 White House correspondents dinner (Muir and Mooney 2004). In a recent experimental study, Baumgartner (2007) demonstrated that exposure to an online Flash video parody from Jibjab.com featuring an animated President George W. Bush singing about his second term actually increased subjects evaluations of him. When, however, political leaders or institutions are the target of jokes made by others, there seems to be a negative effect on respondents evaluations of them. For example, Young (2004a) demonstrated that viewership of The Late Show with David Letterman or The Tonight Show with Jay Leno during the campaign of 2000 had a small but significant and negative effect on viewers evaluations of the presidential candidates. In a similar study, Baumgartner and Morris (2006) showed that exposure to The Daily Show with Jon Stewart during the campaign of 2004 lowered respondents evaluations of the candidates, government, and the media. Baumgartner s (2007) study found that the same online Flash cartoon from Jibjab.com that raised evaluations of President Bush had the effect of lowering respondents evaluations of political institutions. Based on the ELM, as well as research investigating the effects of humor (political and otherwise) on attitude change, I expect that exposure to Handelsman s animated editorial will lower viewers evaluations of the candidates being lampooned. In the next section I discuss the data and methodology of the study. Data and Methods Presidential candidates and campaigns are almost endlessly targeted by political humorists on television (Niven, Lichter, and Amundson 2003), in newspaper editorials (Conners 2007), and on the Internet (Darr and Barko 2004; Baumgartner 2007). In early 2006, nationally syndicated, award-winning, editorial cartoonist Walt Handelsman joined other cartoonists in directing part of his creative energy to producing animated Flash cartoons for the Web ( Walt Handelsman n.d.). One of these cartoons won him his second Pulitzer Prize in 2007, and the cartoons, which appear on Newsday.com, helped the paper become one of the five most visited online newspapers in November and December of 2007 ( Joshi 2008). The specific cartoon I selected was Handelsman s Political Reality Show, which first appeared in early 2007. The clip takes the form of a fake reality television show, titled So You Wanna Be America s Next Top Contender to Survive the Amazing Race for President! As political humor goes, it is not terribly critical. At a systems level, it takes aim at the presidential election process itself, suggesting that it is too long and may not draw or produce competent leaders. Each of six candidates are presented in turn as starring in various other fake television programs, selected to highlight and lightly lampoon some characteristic of the candidate or their candidacy. In this sense it is quite similar to traditional editorial cartoons, which provide a continual series of snapshot portraits that evaluate candidates on their leadership qualities (Edwards 2001). Table 1 summarizes the portrayal of each candidate (see Appendix for full transcript of the clip).

740 PRESIDENTIAL STUDIES QUARTERLY / December 2008 TABLE 1 Portrayal of 2008 Presidential Candidates in Political Reality Show Candidate(s) (Fake) TV Program Portrayal of Candidate Barack Obama Apprentice, Los Angeles Inexperienced yet charismatic John McCain Survivor, Vietnam A (campaign) survivor: brutalized, yet keeps going Hillary Clinton Trading Spouses Hillary as successor to Bill John Edwards and America s Next Top Model Attractive to woman, and not much else Mitt Romney Rudy Giuliani American Idol Hypocritical conservative To test the effects of the clip on respondents evaluations of the presidential candidates, I employed a posttest-only experimental design (Campbell and Stanley 1963). The experiment took place in late November of 2007, the period of time in the presidential election cycle that many are starting to pay attention to the positioning and candidacies of the main presidential contenders (Pew 2007). This particular research has three advantages that answer common concerns about the validity of experimental research. First, the subjects are college students, an age group that corresponds to the target population (18- to 24-year-old youth). This means that the results of the research are more generalizable than many studies that use college-age students (Sears 1986). Second, the experimental stimulus and survey were administered online, at the computer of the subjects choosing. Because the research is concerned with the effects of viewing online editorial humor, the design incorporates a high degree of experimental realism, allowing us to be more confident about the results (Kinder and Palfrey 1993). In short, the design minimizes problems with both internal and external validity. A final advantage of this design is that this particular age group is becoming increasingly difficult to contact via telephone (Pew 2006). Of those who responded to the survey, only 25.6% reported that they had traditional land-line telephones in the residence where they live, meaning that a traditional telephone survey would likely reach a nonrepresentative sample of these youth. Participation in the survey was solicited via e-mail, and the e-mail address sample was obtained as follows. In the summer and fall of 2007 I selected 10 public universities from each of 10 different states. I then sent each university s registrar a letter requesting a directory of all undergraduate student e-mail addresses. Included in the mailing was a brief description of the research, a copy of the survey, and the institutional review board approval from my university. I received directories from eight institution (in five different states), 1 most for a nominal processing fee. After combining the directories, I randomly selected approximately 15,000 e-mail addresses. In addition, I received the e-mail directory from my home institution, a large public university in the southeast, adding another 19,711 addresses to the mailing list. Following this I randomly assigned addresses from each directory (the broader sample and my home institution) into control or experimental conditions. 1. States included were MN, NC, NJ, WI, and TX.

Baumgartner / EDITORIAL CARTOONS 2.0 741 The invitation e-mail contained the words, [Institution] Political Attitudes Survey 2007, in the subject line. A brief introductory paragraph asked students to participate in research I was conducting about the political attitudes and behavior of college students. I informed them in the short letter that their participation was voluntary and that the survey would take approximately 10 minutes. As an incentive, each individual who completed the survey would become eligible to win one of ten $25 gift cards from Target. For this, their e-mail address was required, meaning that participation was not anonymous. However, I stressed the fact that survey results were completely confidential and that I was the only person that would have access to the data. The e-mail also included my contact information (including my e-mail address). Finally, the invitation contained a link to one of two Web sites, based on the subject s assignment to either the control or the experimental group. Those in the control group were directed immediately to the survey, while those in the experimental group were linked first to a page containing the animated editorial. After viewing the editorial, subjects were then directed to the survey. Both groups completed the same survey, although those in the experimental group were asked a few additional questions about the animated clip. I sent 250 invitations e-mails (125 to each condition) on November 24, 2007, in order to test the survey and the stimulus. Respondents were asked in this initial wave to make note of the time it took them to complete the survey and note any trouble they may have had understanding questions, navigating the survey, and so on. After verifying that the survey could indeed be completed in about 10 minutes and making a few minor adjustments, e-mails to the entire sample were sent on November 26, 2007. A follow-up letter was sent two days later, and the survey was concluded on November 30, 2007. In all, there were 2,938 responses, for a response rate of 8.5%. The relatively low response rate can be explained by several factors. First, the invitations were sent to the students campus e-mail addresses. However, most have other e-mail accounts (e.g., free Web-based e-mail) as well and may not check their university accounts as often (Sax, Gilmartin, and Bryant 2003). This explanation has more weight when we factor in the fact that the survey was only active for four days. Second, universities, like all e-mail providers, are becoming more aggressive in blocking what they perceive as spam. In other (similar) research I received any number of bounced e-mails from various spam-blocking applications, as well as reports from students who found my e-mail in their junk e-mail folder. In addition, institutional researchers report that response rates among college students continue to decline, in part due to survey fatigue (Porter and Whitcomb 2003; Porter, Whitcomb, and Weitzer 2004). Web-based surveys have made it easier for academic departments, units, groups, and individuals on college campuses to survey the student population, and they are doing so with greater frequency. While the response rate was lower than expected, there is reason to be confident that the sample is relatively representative of college students. Samples were drawn from public universities located in several regions of the country and, as such, are probably fairly representative of the larger population of university students. Confidence in the sample is bolstered by the fact that some recent scholarship on nonrespondents suggests

742 PRESIDENTIAL STUDIES QUARTERLY / December 2008 that, while there may be demographic differences between respondents and nonrespondents in survey research, these differences may not be attitudinal (Teitler, Reichman, and Sprachman 2003). Dropping cases that were either under the age of 18 or over the age of 24 yielded a sample of 2,370. In addition to basic demographic and partisan affiliation information, I asked respondents in the experimental group several questions about the stimulus, the first of which was whether they had ever previously seen the clip. Those respondents who reported having seen the clip prior to the experiment were excluded from the analysis. 2 Because the more politically engaged would be less inclined to be swayed by the humorous messages contained in the clip (Miller and Krosnick 1996), I also measured political knowledge and political interest. Political knowledge was measured by constructing an index from six questions that probed respondents basic knowledge of each candidate (see Appendix for question wording). For each question a score of 1 was recorded for the correct answer and a 0 was recorded for all incorrect answers. Scores were added such that the resultant variable ranged in values from 0 to 6 (mean = 4.3, s.d. = 1.6). In a similar way, a political interest index was constructed by summing the scores from two questions: On a scale of 0 (never) to 4 (regularly), how often would you say you talk about politics with family, friends, co-workers, fellow students, or others and On a scale of 0 (not closely at all) to 4 (very closely), how closely would you say you have been following news about the 2008 presidential elections? Scores on this political interest variable ranged from 0 to 8 (mean = 3.3, s.d. = 2.1). Because I was measuring candidate evaluations I wanted to ensure that I was not measuring the nonattitudes of those who had not heard of a particular candidate. Therefore, I asked the following question of each candidate: On a scale of 1 (almost nothing or nothing) to 5 (a great deal), how much would you say you have heard about the following candidates, either from family, friends, neighbors, and so on, or from various media reports? If the respondent answered 1 (almost nothing or nothing) for any candidate, I recoded all candidate evaluation scores for that candidate as missing for that case. Table 2 shows the distribution of the independent variables by experimental condition for the resulting sample. As Table 2 illustrates, the sample (like many national samples; O Rourke and Lakner 1989) overrepresents females and whites to a lesser extent. In all, however, the sample is not a homogenous, freakish, collective (Nelson and Oxley 1999, 1044). It also shows a fair amount of variation in how much respondents had heard about each of the candidates. Noteworthy is the fact that, of all of the candidates, Mitt Romney seemed to be by far the least well known. In fact, 51.6% (1,223) reported having heard almost nothing or nothing about him. This resulted in a much lower n for the analyses of his evaluations. Candidate evaluations were measured in several different ways. I first used a standard thermometer score to measure respondents overall evaluation of each candidate. The question asked, On a scale of 1-10, how do you feel about the following candidates? not. 2. I also asked if respondents enjoyed viewing the online clip, and only 9.1% reported that they did

Baumgartner / EDITORIAL CARTOONS 2.0 743 TABLE 2 Sample Descriptive Statistics Variable Control Group Experimental Group Entire Sample Race White 85.9% 82.1% 84.1%*** Non-white 14.1 17.9 15.9 Gender Female 69.4% 69.9%*** 69.6% Male 30.6 30.1 30.4 Party Identification Democrat 32.1% 32.7% 32.4%* Republican 21.0 18.7 19.9 Independent 35.4 35.6 35.5 (Don t know) 11.6 13.1 12.3 Mean age 20.30 (st.d. 1.6) 19.92 (st.d. 1.7) 20.12 (st.d. 1.7) Political interest (mean) 3.37 (st.d. 2.1) 3.29 (st.d. 2.1) 3.33 (st.d. 2.1) Political knowledge (mean) 4.33 (st.d. 1.6) 4.20 (st.d. 1.6) 4.27 (st.d. 1.6)** Heard about candidates (mean): All 19.15 (st.d. 5.1) 19.23 (st.d. 5.1) 19.18 (st.d. 5.1) Clinton 4.24 (st.d. 1.0) 4.34 (st.d. 1.0) 4.29 (st.d. 1.0)** Obama 3.88 (st.d. 1.2) 3.93 (st.d. 1.3) 3.9 (st.d. 1.3) Edwards 3.28 (st.d. 1.2) 3.25 (st.d. 1.2) 3.27 (st.d. 1.2) McCain 2.62 (st.d. 1.2) 2.65 (st.d. 1.2) 2.63 (st.d. 1.2) Romney 1.97 (st.d. 1.2) 1.90 (st.d. 1.2) 1.94 (st.d. 1.2) Giuliani 3.16 (st.d. 1.4) 3.17 (st.d. 1.4) 3.16 (st.d. 1.4) Institution Home institution 69.7% 74.1% 71.8*** Other institutions 30.3 25.9 28.2 N 53.5% 46.5% 2,370 Range for political interest was 0-8; range for political knowledge was 0-6; range for heard about (all) candidates was 6-30. Difference of means tests: * p <.1; ** p <.05; *** p <.01. The higher the number, the more favorable you feel; the lower the number, the less favorable you feel. Following this, evaluations of individual candidates were probed in more depth, with the following question battery: On a scale of 1 (not well at all) to 5 (extremely well), how well do the words or phrases listed below describe (CANDIDATE NAME). Words or phrases included Really cares about people like me, Honest, Inspiring, Knowledgeable, Provides strong leadership, and Competent. I then added these scores together to get a summary evaluation for each candidate, with scores ranging from 6 to 30. Differences of means tests suggest that there were significant differences between the samples gathered from the different institutions on such key variables as gender, race, (p <.01), political knowledge (p <.05), and partisan identification (p <.1). For this reason, and because the sample overrepresents my home institution, I included a dummy institution variable (1 = home institution, 0 = other institutions) in the models. In the following section I present the findings.

744 PRESIDENTIAL STUDIES QUARTERLY / December 2008 Findings Because this research was experimental in nature, I first present mean scores, by experimental condition, of the various summary measures of candidate evaluations for each candidate. As seen in Table 3, Barack Obama and Rudy Giuliani rank the highest in terms of overall evaluations, while Mitt Romney ranks the lowest. Romney is also the candidate who was least well known by respondents. Each of the measures employed for candidate evaluations move in the expected direction (negative) except the thermometer scores for Obama. However, with the exception of the thermometer scores of John Edwards, Rudy Giuliani, and the combined candidate thermometer scores, none of the differences in means are significant. In order to isolate the effect of the clip on candidate evaluations I ran ordinary least squares regression on these evaluations scores. Included in the analysis are the various demographic and political engagement variables discussed above, as well as a variable controlling for the respondent s institution. 3 Table 4 presents the results of the analysis for the thermometer scores for each of the six candidates, as well as a combined index of all candidate thermometer scores (range: 6-60). Table 4 reveals several expected results. First, partisan identification is significant and in the expected direction for each candidate. The effect on all three Democratic candidates is negative, while the effect on all three Republican candidates is positive. Relatedly, the effect of race is significant for all three Democratic candidates and in the expected (negative) direction. The effect of having heard more about the candidate had a positive effect for each of the candidates, as well as for the combined field of all six candidates. This too was not unexpected, as some research exists that suggests that the effects of humor (specifically, late night comedy) on evaluations of lesser-known candidates is greater (Moy, Xenos, and Hess 2006). One interpretation of this finding is that, all other things being equal, familiarity may help breed more positive evaluations of a given candidate. The effect of institutional affiliation is also significant in the case of several candidates, suggesting that it s inclusion in the model was important. The findings with respect to the effect of viewing the online editorial clip were mixed. The effects of the clip were in the expected direction in all cases, but were only significant in the case of the three Republican candidates as well as the entire field. This suggests that, while I cannot disconfirm the null hypothesis that viewing an online humorous editorial cartoon has no effect on candidate evaluations, neither can I rule out the possibility that there might be an effect. Table 5 presents the analysis of the effects of the clip on combined candidate evaluation scores. In many respects the results in Table 5 mirror those in Table 4. Race and party identification are significant and in the expected direction for each party s respective candidates. As in the previous analysis, having heard about the candidate also seems to 3. In addition to testing the effects of the institution variable as a dichotomous home institution other institutions variable, I created dummy variables for each institution. Tests of these models (using different institutions as reference groups) show no significant differences in the results. Because of this (and because there was a fair amount of variation in the n for each of the other institutions) I present the more parsimonious models with the home-other dummy variable results.

Baumgartner / EDITORIAL CARTOONS 2.0 745 TABLE 3 Summary Evaluation Scores by Experimental Condition Control: Mean (st.d.) Experimental: Mean (st.d.) Sample: Mean (st.d.) Clinton Thermometer 4.58 (3.2) 4.50 (3.2) 4.54 (3.2) Index 17.33 (6.8) 17.07 (6.9) 17.21 (6.8) N 1,234 1,085 2,319 Obama Thermometer 5.63 (3.0) 5.71 (2.9) 5.67 (3.0) Index 19.94 (6.7) 19.69 (6.8) 19.82 (6.7) N 1,179 1,017 2,196 Edwards Thermometer 5.09 (2.6) 5.04 (2.5)** 5.06 (2.6) Index 17.63 (6.5) 17.36 (6.5) 17.51 (6.5) N 1,163 990 2,153 All Democrats Thermometer 15.35 (7.2) 15.34 (7.0) 15.35 (7.1) Index 55.04 (16.5) 54.16 (16.5) 54.64 (16.5) N 1,100 943 2,043 McCain Thermometer 5.14 (2.4) 4.83 (2.4) 4.99 (2.4) Index 18.58 (5.7) 17.68 (5.7) 18.16 (5.7) N 975 853 1,828 Romney Thermometer 4.56 (2.4) 4.02 (2.4) 4.31 (2.4) Index 16.92 (5.6) 15.87 (5.7) 16.43 (5.7) N 616 531 1,147 Giuliani Thermometer 5.75 (2.6) 5.63 (2.7)** 5.69 (2.6) Index 19.83 (5.8) 19.43 (5.8) 19.65 (5.8) N 1,055 909 1,964 All Republicans Thermometer 15.85 (5.6) 14.58 (6.0) 15.26 (5.8) Index 55.00 (13.2) 53.60 (13.6) 54.88 (13.4) N 561 487 1,048 All Candidates Thermometer 30.66 (8.1) 29.66 (8.7)** 30.20 (8.4) Index 109.25 (21.4) 106.75 (22.3) 108.08 (21.8) N 543 471 1,014 Individual candidate Thermometer scores ranged from 1 to 10; Index scores ranged from 6 to 30. Combined Democratic and Republican candidates Thermometer scores ranged from 3 to 30; Index scores ranged from 18 to 90. Combined candidate Thermometer scores ranged from 6 to 60; Index scores ranged from 36 to 180. ** p <.05. exert a positive influence on candidate evaluations. The results differ, however, in the effects seen in the main variable of interest, namely, viewership of the online animated editorial. Here, the effect is negative and significant on the evaluations of each of the candidates. It should be noted that these effects are relatively small, but hold up even

746 PRESIDENTIAL STUDIES QUARTERLY / December 2008 TABLE 4 Effects of Online Editorial on Candidate Thermometer Scores Variable Clinton Obama Edwards McCain Romney Giuliani All candidates Viewed editorial -0.18 (.11) -0.07 (.11) -0.14 (.11) -0.27 (.11)** -0.47 (.14)*** -0.18 (11)* -1.04 (.35)*** Heard about candidate 0.43 (.07)*** 0.65 (.06)*** 0.59 (.06)*** 0.69 (.06)*** 0.62 (.07)*** 0.78 (.06)*** 1.10 (.05)*** Political knowledge -0.10 (.04)** 0.13 (.04)*** -.03 (.04) 0.90 (.04)** -0.04 (.05) -0.07 (.04) 0.56 (.13)*** Political interest -0.00 (.03) -0.04 (.03) -0.15 (.03)*** -0.54 (.03)* 0.00 (.04) -0.17 (.03)*** -0.20 (.11)* Partisan identification -1.58 (.05)*** -1.19 (.05)*** -0.91 (.05)*** 0.49 (.05)*** 0.75 (.06)*** 0.72 (.05)*** -1.91 (.16)*** Gender -0.32 (.12)*** -0.07 (.13) 0.21 (.12)* 0.41 (.12)*** 0.02 (.14) 0.01 (.12) 0.53 (.38) Age 0.13 (.03)*** -0.02 (.03) 0.1 (.03) -.01 (.03) 0.02 (.04) -0.10 (.03)*** 0.06 (.11) Race -1.58 (.16)*** -1.03 (.16)*** -0.42 (.16)*** 0.27 (.17) 0.07 (.20) 0.37 (.17)** -1.95 (.50)*** Institution -0.08 (.13) -0.64 (.13)*** -0.23 (.13)* 0.37 (.13)*** 0.16 (.15) 0.50 (.13)*** -0.36 (.40) Constant 6.80 (.78)*** 8.01 (.77)*** 6.74 (.71)***.74 (.74) -0.01 (.88) 2.96 (.74)*** 9.28 (2.28)*** Adjusted R 2.405.327.212.169.212.225.373 n 2,049 1,955 1,924 1,661 1,068 1,770 2,079 Note: Partisan identification was coded 1 = strong Democrat to 5 = strong Republican; race was coded 0 = Non-White 1 = White; gender was coded 0 = female 1 = male; institution was coded 1 = home institution, 0 = other institutions. Cell entries are unstandardized ordinary least squares coefficients, with standard errors in parentheses. * p <.1; ** p <.05; *** p <.01 (one-tailed).

Baumgartner / EDITORIAL CARTOONS 2.0 747 TABLE 5 Effects of Online Editorial on Combined Candidate Evaluation Scores Variable Clinton Obama Edwards McCain Romney Giuliani Viewed editorial -0.49 (.25)* -0.62 (.26)** -0.41 (.27) -0.85 (.25)*** -0.88 (.32)*** -0.50 (.26)* Heard about candidate 1.04 (.16)*** 1.50 (.15)*** 1.24 (.14)*** 1.65 (.14)*** 1.44 (.17)*** 1.72 (.14)*** Political knowledge -0.12 (.09) 0.18 (.10)* -0.11 (.10) 0.38 (.10)*** -0.23 (.13)* -0.11 (.10) Political interest -0.09 (.07) -0.13 (.08)* -0.23 (.08)*** 0.13 (.07)* 0.06 (.09) -0.26 (.07)*** Partisan identification -3.11 (.12)*** -2.52 (.12)*** -2.12 (.13)*** 0.91 (.12)*** 1.40 (.15)*** 1.07 (.12)*** Gender -1.55 (.28)*** -0.74 (.29)** -0.38 (.30) 0.99 (.27)*** -0.60 (.34)* -0.43 (.28) Age 0.28 (.08)*** -0.05 (.08) 0.03 (.08) -0.02 (.08) -0.02 (.10) -0.24 (.08)*** Race -2.61 (.36)*** -2.40 (.38)*** -1.08 (.40)*** 1.09 (.39)*** 0.02 (.49) 1.25 (.39)*** Institution -0.05 (.28) -1.00 (.30)*** -0.06 (.32) 0.27 (.29) 0.49 (.37) 1.03 (.29)*** Constant 20.35 (1.73)*** 25.22 (1.78)*** 21.62 (1.81)*** 7.33 (1.68)*** 9.37 (2.11)*** 15.05 (1.71)*** Adjusted R 2.351.292.182.189.159.163 n 2,049 1,955 1,924 1,661 1,068 1,770 Cell entries are unstandardized ordinary least squares coefficients, with standard errors in parentheses. * p <.1; ** p <.05; *** p <.01 (one-tailed).

748 PRESIDENTIAL STUDIES QUARTERLY / December 2008 TABLE 6 Effects of Online Editorial on Candidate Preference Candidate Control Group Experimental Group Difference Obama 25.1% 26.0% +0.9% Giuliani 18.5 19.7 +1.2% Clinton 17.0 16.2-0.8% Edwards 10.2 11.3 +1.1% McCain 8.5 7.2-1.3% Romney 4.1 3.3-0.8% None of the above 16.6 16.3-0.3% n 1,268 1,102 with the inclusion of other powerful predictors of political attitudes (party identification, political knowledge and interest, race, etc.). Of course, having determined that online political humor in the form of an animated political editorial has negative effects on the evaluation of political candidates, at least two questions remain, both of which speak to the so what question. First, do these effects substantively alter the nature of respondents final evaluations? In other words, do they affect candidate preference or vote choice? The latter question is outside the scope of the present design, but the former can be determined. Although the survey was conducted relatively early in the campaign season, I asked the standard question about vote choice: If the presidential election was being held today, which candidate would you vote for? Respondents were presented with the choice of all six candidates, as well as an option to select none of these candidates (or other). While the editorial attacked all candidates, it could be the case that respondents perceived the attack on one or more candidates as being harsher than on others. It could also be the case that support for one or more candidates was softer than for other candidates. In either case, the effects of the clip could potentially alter their preference for one candidate over another. The results from this question were crosstabulated by experimental group and are presented in Table 6. As seen in Table 6, the rank order of candidates does not change as the result of exposure to the clip. While the difference between groups favoring certain candidates is in some cases positive and in others negative, in no case does the difference exceed 1.3% in absolute values. Chi-square tests of these differences reveal that none are significant (p =.65). Of course, a stricter test of candidate preference might be head-to-head matchups, but the results do tentatively suggest that the negative effects of online humor may not affect candidate preferences. This is especially the case when we consider the fact that on the whole, the body of political humor during a presidential campaign takes aim at each candidate roughly equally. What about other possible consequences of the negative effects of online political humor? A recent article by Baumgartner and Morris (2006) on the effects of political humor on political attitudes demonstrated that viewership of The Daily Show lowered evaluations of presidential candidates and lowered trust in political institutions. This

Baumgartner / EDITORIAL CARTOONS 2.0 749 TABLE 7 Effects of Online Humor Viewership on Political Participation (Control Group Only) Variable Online humor viewership 0.21 (.05)*** Heard about candidates 0.01 (.01) Political knowledge -0.01 (.03) Political interest 0.24 (.02)*** Partisan intensity -0.06 (.04) Gender 0.03 (.08) Age 0.05 (.02)** Race -0.12 (.10) Institution -0.22 (.07)*** Constant -0.85 (.45)* Adjusted R 2.211 n 1,268 Cell entries are unstandardized ordinary least squares coefficients, with standard errors in parentheses. * p <.1; ** p <.05; *** p <.01 (one-tailed). seemed to suggest a certain cynicism toward politics and the political system. They further speculated that this cynicism may translate into lowered political participation, although they did not test for this. In order to get a preliminary sense of the relationship between the effects of online humor and political participation, I asked respondents about their political activity and online humor viewing habits. Participation was measured by presenting a battery of eight political activities and asking respondents to check all of the ones that they had done in the past 12 months (see Appendix for question wording). The variable ranged in values from 0 to 8. To measure online humor viewing habits, respondents were asked, On a scale of 1 (never) to 5 (regularly), how often do you visit JibJab.com or other websites devoted to or containing political humor? This variable captures, in a rough way, exposure to online humor. In addition to these two variables, I controlled for political interest, knowledge, how much the respondents heard about the candidates (an additive index for all candidates), the demographic variables, and the dummy institutional variable. I recoded the existing measure of partisanship to measure capture partisan intensity, as a good deal of scholarship exists that suggest that strong partisans are more likely to participate politically. Strong Democrats and Republicans were coded as 3, Democrats and Republicans were coded as 2, independent or neither were coded as 1, and those we answered I don t know, or I haven t given it much thought were coded as 0. In order to eliminate the effects of having viewed the online clip, and because viewing the clip could have no immediate effect on political participation, I tested this model on respondents in the control group only. Results are in Table 7. Table 7 shows that the effects of political interest and age are both positive and significant, and this was expected. While I did not present a formal hypothesis concern-

750 PRESIDENTIAL STUDIES QUARTERLY / December 2008 ing the effects of online humor viewership on political participation, I theorized that it might have a negative effect. This expectation appears to have been wrong. While controlling for an array of variables that are related to political participation, online humor viewership appears to have a positive effect on participation. Of course, it could be the case that only the politically engaged are those that view political humor online, but this analysis provides partial evidence that refutes Baumgartner and Morris s (2006) speculation. Discussion Compared to its availability even a generation ago, political humor is ubiquitous. One can see humorists (and others) poking fun at political leaders and institutions on any variety of talk shows (cable or otherwise), television cartoons (e.g., The Simpsons ), talk radio, newspapers, and now, increasingly, the Internet. One of the more interesting developments in the world of political humor in the past decade or so has been the decline of the number of full time editorial cartoonists. Some few have adapted to this new reality by supplementing their work by moving online. Humorists like Mark Fiore and Walt Handelsman, as well as others (e.g., JibJab.com) produce high-quality and clever Flash animations that continue the long tradition of editorial cartoons. There is little work that examines the effects of political humor on political attitudes, and almost none that focus on the effects of this new genre of editorial. In this article I tested the proposition that exposure to these online animated editorials has a negative effect on presidential candidate evaluations. The analysis illustrates that this may be the case. Holding other factors that influence political attitudes constant, the effects of the clip on the combined candidate evaluation scores (Table 5) was negative for all six candidates. In the case of the thermometer scores, the results were less clear. For three of six candidates, as well as the combined thermometer scores for all candidates, the effect was negative. For the other three candidates the clip moved attitudes in the expected direction, but the effect was not significant. Interestingly, the three candidates that the clip seemed to have no effect on were all Democrats, but why this might have been the case is unclear, as partisan identification was controlled for. In addition, the clip seemed to have been equally harsh on each candidate. It might also have been the case that the effect of the clip was moderated by how much respondents had heard about the candidate, but this too was controlled for. While the research was framed in terms of the ELM of attitude change, the results speak more broadly to other literature on the priming effects of the news media on candidate evaluations (Miller and Krosnick 1996). Other media effects research suggests that negative messages in political advertisements may increase negative attitudes toward politics and political actors (Ansolabehere and Iyengar 1995; Cappella and Jamieson 1997). Following the presentation of the effects of the clip of candidate evaluations, I extended the analysis to test the effect of the clip on candidate preferences and online political humor on political participation. In the case of the former, I cross tabulated

Baumgartner / EDITORIAL CARTOONS 2.0 751 candidate preferences by experimental group. Results suggested that the clip had no effect: preferences remained constant, the difference in the percentage favoring individual candidates was small and not significant. However, because the effects of the clip were negative, we might expect that, if the body of political humor focused more on a certain candidate more than another (or all others), this might affect the rank ordering of candidate preferences, especially if the race was close. Following this, I regressed the effects of online humor viewership on political participation. Using responses from the control group only, and controlling for other factors that influence political participation, I showed that online humor viewership may have a positive effect on participation. Of course, it could be the case that the politically active are more likely to view political humor, but the results do suggest the possibility that political humor may engage and mobilize youth. So what to make of the fact that these results appear to be contradictory? Effects of the clip on candidate evaluations were negative, but had no substantive effect on candidate preferences and appeared to have a positive effect on political participation. The findings that the clip had a negative effect on candidate evaluations is consistent with the small body of scholarship that illustrates the negative effect of political humor on political attitudes. What this body of research has not addressed is the substantive effect of political humor on political participation. While some have posited that the effect of political humor on various measures of political participation is negative, this analysis suggests that the effects might be benign (as in the case of candidate preferences) or positive (in the case of political participation). These findings were unexpected, but do not contradict any existing research. The contribution the research makes, therefore, is to further confirm that political humor has a negative effect of attitudes and to suggest that effects beyond attitudinal may be minimal or even positive. Further research is clearly needed in this area. This research does, of course, have its limitations. The most obvious of these has to do with the ability to generalize from the sample. First, the survey was confined to university students only. However, testing media effects on university students is fairly standard practice (Druckman 2001; Nelson and Oxley 1999; Nelson, Oxley, and Clawson 1997). It might also be suggested that Internet usage has not attained the nearuniversality that television has. But while a digital divide still exists, recent estimates suggest that as many as 80% of 18- to 24-year-olds have access to and use the Internet (Fox 2005). Although not strictly representative of the total population of 18- to 24-year-olds, the sample was less unrepresentative than many other nonprobability samples. Finally, while these Internet users may not be representative of all youth, recent research examining results from Internet and more traditional telephone surveys suggest that, while the samples may differ in certain demographic characteristics, the underlying psychological mechanisms underlying common political decisions of Internet users do not differ much from the population (Best, Krueger, Hubbard, and Smith 2001). Of course, the effects of media stimuli introduced during an experiment may be short lived. However, because political humor draws on and in constructed from preexisting images, it may be the case that humor reinforces already existing attitudes. Given the small amount of scholarship that exists on the effects of political humor, as well as the seemingly contradictory results of this analysis, it is clear that further research is warranted.