Overview
Q: Can you please give us an overview of the study?
A: Well, this project is a result of a kind of a happy coincidence of interests between myself and my co-author, Becky, who at the time was actually an undergraduate at the University of Wisconsin, and Becky was interested in the impact of the media on children; a political socialization kind of interest, and I’d been interested in the impact of emotions on political thinking and political behavior, so, putting those two interests together, we decided that we would study the impact of negative campaigning on children’s political attitudes, which is a research area that really has been underdeveloped. There’s been some studies, as you know, on the impact of negative campaigning on adult’s political orientations and there’s some controversy about whether negative campaigning is in fact bad, but there really hasn’t been any studies on children and we were interested in looking at children in part because we know that there’s some generational differences in political attitudes and voting turnout, in particular, among younger generations and there’s a suspicion that the [???] of political discourse is part of the answer to that question of why do young people not turn out and so we decided that we would study this question experimentally and so we put together and so we put together some positive political ads from the 1988 presidential campaign and a set of negative political ads from the 1988 presidential campaign and we used the ’88 presidential campaign because we didn’t want the children to be familiar with the political candidates that were featured in the ads. The most interesting aspect of the actual administration was trying to recruit the children and when you do research on children the bar is set a little bit higher in terms of getting approval from, what is known as the Institutional Review Board or the Human Subjects Committee and we had to figure out a way to find children with their parents, because parents have to sign the consent form for the children because they’re not of legal age and therefor can’t make their own informed consent and so we decided that we would go to the fair. And so we took the project to the Dane County Fair in the summer of 1994 and my co-author, Becky, went up and down the show barns, where the kids were showing cows and sheep and all the 4-H kinds of things, and asked the kids if they would like to participate in the study and then if their parents agreed, we showed them some political ads. Before we showed them the ads we measured their initial orientations and then we showed them the ads, either the positive set of ads or the set of negative ads, and then we measured their orientations afterwards, so it’s what is known in the experimental literature as a pre/post test design. And, our essential findings didn’t surprise us; the ads did affect the ways children felt about the country—positive ads made them feel good, negative ads made them feel bad, but what is of more importance is that we were able to demonstrate through a set of analysis that the feelings that the children had about the country in turn affected their judgments about government. In particular, how competent they thought government was and how much they sort of, liked government in general. And so, our conclusion is that in fact, negative political advertising can have some negative consequences for people’s belief about government so that they think government’s less competent, they’re less willing to trust it, and should these attitudes persist into adulthood the implication is that we’re creating a generation of future citizens who enter political majority with fairly cynical attitudes already and then the other interesting finding and I think, of most concern for people who study negative advertising is that, while we didn’t find that the negative ads in general diminished the children’s enthusiasm for voting, what we did find is that what is known as interaction affect, so that children who lacked psychological resources; that is, they weren’t sure that they could make a difference the negative ads actually demobilized them. But for children who felt that they were fairly [???] in fact, the ads stimulated their interest in voting. So you end up getting a V-shape like this where some children are demobilized and some children are actually stimulated, so the net effect is actually, if you just look at main effects, there’s no main effect, but there’s this very interesting interaction, the implication of which is that negative political campaigning can exacerbate differences in psychological inequalities in the electorate and I think that that’s a fairly interesting finding with some interesting implications for reform, in terms of political discourse.
Research Questions
Q: What was the state of the literature on political advertising when you began your study?
A: Well, the research literature on political advertising is fairly mixed. There’s a very important book by Steven [???] and [???] [???] which argues that negative campaigning has a number of consequences for voters attitudes about the candidates but also their attitudes toward political participation, but there have been a number of criticisms of that basic finding largely for people who are doing survey research, the [???] and [???] book is an experimental study and the survey research does not show the same findings, does not support their findings that negative campaigning is demobilizing . So, when there’s controversy in the literature, that’s an invitation for people to jump in and try to clarify the controversy and what we do is we sort of split the difference because we find, in our study we find, no main effects for political advertising but we do find an interaction. So that, if you just look at turn out levels over time, you probably won’t see any impact of negative advertising but if you look at individual differences you do find that campaign advertising can have a demobilizing effect on some voters.
Q: A key concept was the public mood. Can you define that for us and explain why it was important to craft that concept in the way you did?
A: The concept of public mood is designed to get at the sense that because people are members of a national political community, a country, that’s an important identity for them, whether they recognize it or not, and as an identity it can have an emotional consequences for them when they hear information that bears on that identity. And so, the concept of public mood is designed to capture the notion that, not only do we experience emotional reactions because of personal events, if your son brings home an A on his report card that makes you happy, but we also experience emotional reactions because of these collective identities that we share, so when we hear for example that the crime rate has fallen for six consecutive years, that in general, would make most Americans feel good, or a budget surplus makes most Americans feel good. But an event like the shooting at the Capitol where two police officers were killed is an event that would make most of us feel bad, even though in fact, it’s not an event that we really personally experience but rather something that we learn about through mediated information systems. So, the notion of public mood is designed to capture those emotional reactions we have as a consequence of our citizenship status, our shared citizenship status.
Q: Could you describe the four hypotheses that motivated your study?
A: Well, we were interested in the impact of campaign advertising and we developed four hypotheses about this. The first hypothesis was the obvious one, which is that political advertising will affect people’s emotional reactions, in this case the children’s emotional reactions about the country so that they would feel more positive about the country after seeing positive ads and negative after seeing negative ads. This is essentially in the experimental literature, what’s known as a manipulation check. The second hypothesis was that there would be some consequences of those feelings for more specific political judgments about government. In other words, the impact of campaign advertising on political attitudes would be mediated through the changes in feelings that were evoked as a result of the ads. And then the third and fourth hypotheses we investigated were related to this notion about the impact of campaign advertising on political participation. And the simple, we call it our simple, hypothesis, is that children who had viewed negative political advertising would be demobilized, that is, show less enthusiasm for voting. And our interactive hypothesis was that, rather than seeing a main effect of campaign advertising on voting enthusiasm, we would see an interaction, so that negative campaign advertising would be demobilizing for children who either lacked political [???] or had lower levels of political knowledge, but for children who had high levels of [???] or had a lot of information about the political system, that negative feelings would actually stimulate them to participation and so we found support for that hypothesis rather than the simple demobilization hypothesis.
Q: The interaction hypothesis in not quite as intuitive as the others. Is that something you anticipated in advance or something that emerged from the data analysis?
A: Well, because of the work that I’ve done in the past, on the role of emotions in political information and political behavior, the work that I’ve done on adults on negative emotions does suggest that their impact is more differentiated, that for some people, negative emotions can be stimulating, especially if you think you have the capacity to cope with whatever is causing that negative feeling state, so, for example, if you meet a bear in the woods, almost all of us are going to feel fear, but whether we run away or stand pat depends a lot on if we have a gun, and so you would predict different things based on knowing something about the individual’s capacities, and so given the mixed state of the literature, where some people were finding effects and other people were not finding effects, that also alerts you that you might be actually in a situation where there are other interactions with other variables.
Design Sampling
Q: Could you give us an overview of the research design used in your study?
A: Well, we decided in our research design for this study to pursue an experimental strategy so you can have control over the independent variable. And, then the main decision facing us once we decided to take an experimental approach, was whether we would used what’s known as a post-test only control group design or a pre-test/post-test design. And there are arguments for and against each of these kinds of designs. A post-test only design does not really permit you to talk about change, all you can say is that one group is different from the other, on the other hand, a pre-test/post-test design where you can measure change has the disadvantage of reactivity because you’re measuring people twice and so there is some suspicion that by measuring them one time that’s going to have carryover effects for the measurement at that time too. We ultimately decided on a pre-test/post-test design in part because the pre-test would allow us to measure attributes of the children that later we would need to control for, so if there were any differences--since the study sample is fairly small, it’s about 70 kids—if there were differences for example, if kids who had seen the negative ads were higher in political knowledge, for example, just by chance, than by kids who had seen positive ads, you would want to able to control for those differences and we were able to measure that then, in the pre-test and also we could control for any individual differences in initial levels of either political attitudes or the kid’s initial feelings about the country prior to seeing the ads.
Q: What ages were the children?
A: The children that we used in our study, the youngest child was eight and the oldest child was 14, so middle-schoolers, mostly, and the mean age was about ten. Most of the kids were between fourth and fifth grade, which actually, given the political socialization literature, is a fairly good time to study children because they do have a fair amount of knowledge about the political system and by about fourth grade, most children, some studies suggested about 90 percent of American children, understand the role of elections and know about elections and so that was important for us, that they had some basic understanding about elections and campaigns and voting.
Q: 164 Where were the experiments conducted?
A: Well, one of the things when you study children as we did in our study, that you need to take into account is that you need to get the parents’ consent for the children’s participation and we initially thought that we would work through the Madison, Wisconsin public schools, but it turns out that it’s very, very difficult bureaucratically to work through the public school system and so we decided instead that we would try to work through existing children’s organizations as a way to tap an existing set of children and then also maybe able to get their parents’ permission at the same time. And so we contacted the Dane County 4-H association and we got their membership list and initially what we did was took a random sample of children’s names and sent a letter to their parents asking them whether they would allow their child to participate in our study. Unfortunately, our response rate was very low, in terms of that solicitation, and so we decided we actually had to take the experiment to the kids. And so, we showed up at the Dane County Fair in the summer of 1994 in Madison and my co-author, Becky [???], who’s very cute and very charming, went up and down the cattle barns and the sheep barns and the horse barns trying to find children who were showing their animals who also had their parents their with them. So she would ask parents and then the children if they would participate in this study and we actually had a small inducement for the children; the Dane County 4-H Association gave us coupons to give the kids for a free yogurt cone. So if they came to participate in our little 15 minute study, then they could go scamper off to the concession stand and get a free yogurt cone and we then gave a small donation to the 4-H Association for every child who participated.
Q: Where did you actually conduct the experiments? There on the fairgrounds?
A: Yes, the experiments was conducted on-site in a building that the 4-H Association has on the Dane County Fairgrounds.
Q: Could you please describe the group structure in the design, the pre-test, the post-test, the control group?
A: Well, we didn’t have the classic control group that doesn’t get any stimulus. Instead, what we had was we had the positive ad group and the negative ad group and each served as a control for the other. In other words, if you see differences between the two groups, you can attribute it to political advertising. If you had a control group that was shown say, regular kinds of commercials, that would allow you then to look at the notion of how reactive the kids were to just being in the setting in general, but we didn’t have that control group, so there are some limitations to the design. It would have been nice to have a more fancy kind of experiment where the commercials were actually embedded into real broadcast, but children who are 8,9, 10 don’t have the longest attentions spans in the world, so we just decided to run the experiment straight up, it’s just, we want you to look at these ads and then we’re going to ask you some questions about them.
Q: How long did the experiment take for each child participant?
A: The experiment was relatively brief. We did want to keep it brief so their attention didn’t wander. And so, it probably took them 5-10 minutes to fill out the pre-test and then the ads themselves were about 3 minutes and then another 5 minutes to fill out the post-test so it was about a 15 minute session.
Q: There are two key concepts in your study, political knowledge and public mood, how are they operationalized?
A: Well, the definition of public mood is that it’s a diffuse emotional state that people experience as a consequence of their membership in a nation-state and so when I’ve looked at this concept in studies of adults, we simply asked them, when you think about the United States, to what extent do you feel x?, where x is a series of emotions terms and when working with children, I had to do some research into children’s understandings of emotion labels, because in studies of adults we used words like "enthusiastic" and "frustrated" and we were a little bit concerned that those emotion terms might not be as familiar to especially the younger children in a study and so I found some research that suggested that by about the age of five, all children pretty much know the concepts of anger, happiness, and sadness, so we asked about those three emotions. Political knowledge, we measured in pretty much the same way it’s measured in studies of adults--just factual knowledge about the U.S. political system. So we asked them where the nation’s capital is located, it was an open-ended question, we then had photographs of President Clinton and Vice-President Gore and asked the children to identify who those people were, we asked them if they knew the presidents before Clinton and there were two questions if they knew the President and the Vice-Presidents political party membership, and those two were the toughest questions, only about 30 to 40 percent of the children knew that Clinton and Gore were Democrats.
Q: How did you randomly assign the children to the positive and negative advertising groups?
A: Yeah, there’s, well no we didn’t flip a coin, we should have flipped a coin, but we just went every other one.
Measurement & Data Collection
Q: There were several questions you used to measure political knowledge and public mood. How were they scaled to create the variables used in your analysis?
A: To create the political knowledge variable was very easy. We simply summed the number of correct answers that the children had to our six political knowledge questions, so the variable range is zero, that is, they flunked, or six, they answered all of them correctly. In terms of measuring public mood, what we did was we averaged across all three emotions, so we subtracted the negative emotions from the one happy emotion and then divided by three and so that variable range is from very negative to very positive feelings. And the reliable of the public mood variable is .80, which is a pretty nice reliability statistic.
Q: Were there any problems in administering a written questionnaire to children that you might not have with adults?
A: Well, you know it’s a common question we get asked about this study is "can you administer a pencil and paper test to children?" and one response is "they take these tests all the time in school". By third or fourth grade they’re taking Iowa tests of basic skills so they’re very comfortable with the questionnaire format, but you do have to calibrate the questions and so we pre-tested our questionnaire, first we stole from questions from existing political socialization studies that were done on children about the same age as the children in our study, but then in addition we pre-tested our questionnaire on Becky’s younger sister and her friends and made some small changes in question wording after that pre-test, so we did have an expert group judge the quality of our questionnaire.
Q: Were there problems with missing data in a study like this?
A: Well, we did, for almost all the questions on the questionnaire, there was a don’t know response, so if the children didn’t know, they told us that they didn’t know, so a lot of missing data is actually captured in the questionnaire, but most children were able to check the appropriate box. It wasn’t a long questionnaire, it was big type, we didn’t have a huge number of response options, like for example, the question that asked the children how much they liked government, it was the response options were "I really like it", "I sort of like it", "I don’t really like it", so it wasn’t a very complicated task that the children had to execute on the questionnaire.
Q: Were the parents allowed to assist the children?
A: No, the parents were not in the rooms with the children, so most of the parents stayed in the barns where Becky had found them because the 4-H building is very close to the show barns and so sometimes the parents waited because it was a very short session, sometimes the parents just waited for the children.
Q: The dependent variable on your regression analysis then was a public mood variable?
A: Actually no, well, it was for one of the hypothesis, for one of the hypothesis the dependent variable was the post-test measure of public mood, for the other hypotheses, the dependent variable was either their attitudes about government, or in the case of participation, it was their enthusiasm for voting. A simple question, "would you like to vote?", yes, maybe, no, and that was the dependent variable in the final set of analysis.
Q: And did you actually have real live ads that you showed on video to the kids?
A: To our experimental children, yes, we did have four positive ads from the 1988 campaign and four negative ads from the 1988 campaign and so the positive ads included one that Dukakis ran that featured little children playing on little toy pianos and had a big American flag and very inspiring music and one of the negative ads was Bush’s infamous furlough ad, prison furlough ad, and also the Boston Harbor ad, which had very, very scary music and lots of gunk in Boston Harbor. It was a very ominous appearing commercial.
Q: Did you show all four ads to each of the children?
A: Yes, each set of ads was shown to each child, so it’s a pretty heavy dose of political advertising, but it’s actually not, during a heavy campaign season, it’s not unusually to see four or five campaign ads in a row because a lot of campaign ads are 15 second spots or 30 second spots and a lot of them are run right around the news and so you can see a lot of ads in a row, so we didn’t think it was that unusual to feature that many ads.
Data Analysis
Q: What statistical analysis did you do on the data?
A: In terms of the analysis in the study, we used some very simple mean differences to see whether in fact we could find some differences in the children who had seen positive and negative ads. And then for testing some of the complicated hypotheses we used regression analysis and for the analysis of interest in participation we used regression analysis with interaction terms and so we were able to get tests of statistical significance because we get coefficients and standard errors and the regression analysis is nice because you also get a sense of how big the effect is and the impact of the ads on the children’s feelings was really quite enormous, the coefficient was very large, the impact of the ads on the attitudes was smaller but we were able to show in fact that that impact was mediated through the feelings because once we controlled for the children’s public mood state in fact, then the coefficient on the advertising variable went to non-significance.
Interpretation & Dissemination
Q: It’s oftentimes argued that variables of those kinds are even the more the more complicated public mood variable that they’re essentially ordinal and regression at least [???], has some limitations in analyzing those data but yet it has the advantage of interpretability, how did you approach those trade-offs?
A: Well, most of the analysis that you read in the literature that have sort of simple OLS versus more complicated and more appropriate methods really find that the interpretive significance doesn’t change. In other words, you’re not going to get different results, the coefficients that are important and big are important either way. And we did want this study to be able to reach an audience that frankly, is not all that methodologically sophisticated, you know, they might be comfortable with OLS, but the other techniques that would be appropriate, [???] or something, so we went for ease of interpretation and kind of the comfort factor.
Q: Could you describe, once again for us, the essential findings?
A: Well, to summarize what I think are the take-home points from the study are first, that ads do affect children’s feelings, and perhaps adults’ feelings as well, but we were able to show that positive and negative ads have what you would expect to see in terms of people’s emotional responses. These ads in turn, also affected attitudes, but those attitudes were affected because the ads changed the feelings first. So, there is this intermediate step that suggests that feelings are quite important. And then finally, we were able to show with the study that while negative political advertising does not have a simple demobilizing impact on children’s enthusiasm for voting, it does have an interesting interaction that, one implication of which is that it can exacerbate underlying inequalities among people in terms of their participatory resources.
Q: It would be useful for us I think if you just reflected a moment again for us on the decision to go with an experimental design as opposed to a survey instrument. The trade-offs between controlling the independent variable versus the external validity of the results, could you say something about that?
A: Part of our decision to pursue an experimental strategy instead of a survey-based strategy is that it’s very hard when you study media effects to be able to measure accurately people’s level of exposure to the media in which you’re interested. And this has plagued survey studies of media effects for a very long time and so eventually concluded that if you really want to make some precise causal statements about the impact of media, you do need to use a more controlled setting where you can make sure that people have constant exposure levels and that the differences that you see among people who see the media that you’re interested in versus not aren’t due to underlying characteristics in the people themselves, that’s always a problem in survey research, that well, "do heavy TV viewers have less political knowledge because they watch TV or is it because people with less education tend to watch TV more than people with less education?" and so we were able to at least make sure that the findings we were able to show were due to differences in the media environment and not some other third variable that is affecting children’s exposure to political ads.
Q: There are a very wide variety of journals out there in the area of politics and the media to choose from and you choose to go to Political Communication. Why did you choose that outlet for this work?
A: I chose to submit the article to Political Communication for a number of reasons. One is that the editor of the journal, Doris Graber, has been working very hard to improve the quality of the journal and the quality of the journal has gone up quite a bit under her editorship. Another reason for selecting the journal is that, this is a fairly, what you might think of as a boutique study I guess, something that’s probably not suitable for a general readership journal in political science but would be of interest to people who are interested in the role of the media in politics. In addition, the readership of Political Communication is not just political scientists, but it’s also people in mass communication and we did want to reach that audience as well, so those are some of the factors that went into our decision to send it to that journal.
Q: Where does this work go from here?
A: Where does this work go from here? Well, Becky is busy being a lawyer, and getting married, so she’s onto some other pursuits. I’m still interested in looking at the role of emotions in people’s feeling about the country and their attitudes so I’m going back to some survey research where I have some longitudinal data, cross sections who are interviewed every day, and I’m trying to develop a content analysis scheme to measure kind of the positivity or negativity of news discourse and political discourse, so I have a study that I did in 1996 and then I have a smaller study that was done during the 1988 campaign that was piggybacked on to the 1998 national elections study, so I’ve got some good quality data there and having my research assistant code campaign discourse and looking at the impact, so it’s designed to in effect, replicate our experimental study but in a context of a real campaign using adults and also a representative sample of adults.
Q: Perhaps it’s because I’m not a political scientist, but who was running in the 1988 election?
A: These are state level races, so the survey in 1988, the national election studies actually went into three states: Georgia, Illinois, and California, so we’re looking at state races.
Q: But you didn’t need to randomize the order of the…
A: We didn’t randomize, well we probably should have randomized the order of the ads, but we didn’t.
Q: Do you randomize the assignment of the children to the positive and negative conditions?
A: Yes, yes.
Q: I have a question about your audience and your intended audience and whether or not the results are used by candidates or you hope that they’ll be used by candidates to reduce their level of negative advertising.
A: Well, I would, the purpose of the study, I think, for both Becky and me, was to suggest that there is a kind of collective bad that’s produced by negative political advertising. It’s in each candidate’s interest, especially if he or she is a challenger to run negative ads because the research does show that they do tend to work in terms of affecting people’s attitudes about the candidates. But, if every candidate runs negative political ads then we’ve got this climate of political discourse that is potentially demobilizing and it’s hard to know exactly how to regulate campaign discourse or even whether you should regulate campaign discourse ‘cause each candidate should have the opportunity to make his or her case before the electorate. On the other hand, the collectivity has an interest in making sure that the socialization of it’s future citizens is done in such a way that they are interested in participating in the democratic process, so that’s kind of the two sides that you have to balance in terms of regulating campaign discourse—the candidate’s free speech interest versus the collectivity’s interest in terms of making sure that it’s citizens want to participate.
Q: And because of those balancing considerations you’re taking more of an academic perspective on the implications of the results?
A: You mean, on the one hand/on the other? Yes, my own belief, actually, is that the free speech interest trump the collectivity’s interest, on the other hand I do think that there can be some campaign learning that goes one when there are candidates who run successful positive campaigns, like Paul Wellstone, or like Russ Feingold, because then candidates learn that hey, it’s not gonna be a loser if I run positive ads and the electorate apparently seems to reward candidates who run positive ads, so that’s the incentive that then will operate on candidates and will ultimately change their behavior., because I think, that trying to regulate the content of campaign discourse flies in the face of the first amendment and perhaps would deprive people of needed information to inform their decisions.